diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,148819 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9949450373104389,
+  "eval_steps": 500,
+  "global_step": 24800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9999999999999997e-06,
+      "loss": 0.6082,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.999999999999999e-06,
+      "loss": 0.6457,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.999999999999999e-06,
+      "loss": 0.5707,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1999999999999999e-05,
+      "loss": 0.9886,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 0.9204,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7999999999999997e-05,
+      "loss": 1.0933,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1e-05,
+      "loss": 0.4468,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3999999999999997e-05,
+      "loss": 1.5412,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6999999999999996e-05,
+      "loss": 0.7863,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 0.944,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2999999999999996e-05,
+      "loss": 1.3362,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5999999999999994e-05,
+      "loss": 0.4761,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.9e-05,
+      "loss": 1.0441,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2e-05,
+      "loss": 0.8828,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 0.9583,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 0.4751,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.1e-05,
+      "loss": 0.8238,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.399999999999999e-05,
+      "loss": 0.5497,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 1.4629,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 1.0918,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.299999999999999e-05,
+      "loss": 0.2344,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.599999999999999e-05,
+      "loss": 0.8805,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.9e-05,
+      "loss": 0.7408,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.199999999999999e-05,
+      "loss": 0.3671,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.5e-05,
+      "loss": 0.5042,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.8e-05,
+      "loss": 0.8721,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.1e-05,
+      "loss": 0.8938,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.4e-05,
+      "loss": 1.0697,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.699999999999999e-05,
+      "loss": 1.5208,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 1.1255,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.3e-05,
+      "loss": 0.8054,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.599999999999999e-05,
+      "loss": 1.2554,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.9e-05,
+      "loss": 0.339,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000102,
+      "loss": 1.6215,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010499999999999999,
+      "loss": 0.2207,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00010799999999999998,
+      "loss": 0.7652,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011099999999999999,
+      "loss": 0.8505,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 0.3423,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000117,
+      "loss": 0.7405,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 2.8693,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012299999999999998,
+      "loss": 0.4785,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00012599999999999997,
+      "loss": 0.1813,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000129,
+      "loss": 1.2773,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00013199999999999998,
+      "loss": 0.4066,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000135,
+      "loss": 1.0392,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000138,
+      "loss": 0.6844,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 1.027,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00014399999999999998,
+      "loss": 0.4649,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000147,
+      "loss": 0.6427,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015,
+      "loss": 0.5877,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00015299999999999998,
+      "loss": 0.0119,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000156,
+      "loss": 0.5225,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000159,
+      "loss": 2.098,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000162,
+      "loss": 0.5874,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000165,
+      "loss": 0.3626,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000168,
+      "loss": 0.9532,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017099999999999998,
+      "loss": 0.5811,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017399999999999997,
+      "loss": 5.7897,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017699999999999997,
+      "loss": 0.5894,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.4869,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018299999999999998,
+      "loss": 0.4649,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000186,
+      "loss": 0.6504,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00018899999999999999,
+      "loss": 1.0623,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00019199999999999998,
+      "loss": 0.2459,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000195,
+      "loss": 1.4967,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000198,
+      "loss": 0.0267,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000201,
+      "loss": 1.2445,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000204,
+      "loss": 0.4555,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020699999999999996,
+      "loss": 0.8448,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00020999999999999998,
+      "loss": 0.0172,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021299999999999997,
+      "loss": 0.793,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021599999999999996,
+      "loss": 0.4777,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00021899999999999998,
+      "loss": 0.1839,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022199999999999998,
+      "loss": 0.5815,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022199999999999998,
+      "loss": 2.0219,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000225,
+      "loss": 2.3776,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 0.8549,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023099999999999998,
+      "loss": 0.4429,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000234,
+      "loss": 0.1514,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000237,
+      "loss": 2.9324,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 0.3932,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000243,
+      "loss": 0.5469,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00024599999999999996,
+      "loss": 1.1362,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000249,
+      "loss": 0.8098,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00025199999999999995,
+      "loss": 0.2951,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 0.8272,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000258,
+      "loss": 1.0487,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000261,
+      "loss": 1.3986,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00026399999999999997,
+      "loss": 1.5231,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000267,
+      "loss": 0.9686,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00027,
+      "loss": 1.085,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00027299999999999997,
+      "loss": 0.3078,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000276,
+      "loss": 0.9379,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000279,
+      "loss": 0.9469,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 0.5037,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000285,
+      "loss": 1.5305,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00028799999999999995,
+      "loss": 0.5242,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 2.1693,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000294,
+      "loss": 0.534,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029699999999999996,
+      "loss": 1.166,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0003,
+      "loss": 0.4421,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999999879898754,
+      "loss": 0.4234,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999999519595024,
+      "loss": 0.5496,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999999891908881,
+      "loss": 0.8172,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999999807838013,
+      "loss": 0.3801,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999996997468994,
+      "loss": 0.8991,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999995676355414,
+      "loss": 0.47,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999999411503942,
+      "loss": 1.6896,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999992313521027,
+      "loss": 0.3518,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999990271800263,
+      "loss": 0.734,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999987989877175,
+      "loss": 0.4002,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999985467751795,
+      "loss": 0.1489,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999998270542415,
+      "loss": 0.5509,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000299999797028943,
+      "loss": 0.8304,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999976460162286,
+      "loss": 0.6136,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999972977228157,
+      "loss": 1.107,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999969254091975,
+      "loss": 0.5259,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000299999652907538,
+      "loss": 2.1946,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999996108721369,
+      "loss": 0.3648,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999995664347172,
+      "loss": 0.6461,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999995195952794,
+      "loss": 0.5963,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999994703538246,
+      "loss": 0.5154,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002999994187103533,
+      "loss": 0.8321,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00029999936466486646,
+      "loss": 0.9619,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999930821736495,
+      "loss": 0.9212,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999992493678495,
+      "loss": 0.8118,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999918811632136,
+      "loss": 1.0119,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999912446278124,
+      "loss": 0.526,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999990584072303,
+      "loss": 0.7132,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999989899496695,
+      "loss": 0.6287,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999891909010003,
+      "loss": 0.1722,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299998845828523,
+      "loss": 1.661,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999987701649395,
+      "loss": 0.9841,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999986920993509,
+      "loss": 0.764,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999986116317583,
+      "loss": 0.5235,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999985287621632,
+      "loss": 0.8351,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999984434905666,
+      "loss": 0.3025,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999983558169701,
+      "loss": 0.4589,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999982657413751,
+      "loss": 0.5024,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999981732637829,
+      "loss": 3.0238,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999980783841951,
+      "loss": 0.7635,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999979811026132,
+      "loss": 0.384,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999788141903876,
+      "loss": 0.4209,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999977793334734,
+      "loss": 0.8944,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999767484591866,
+      "loss": 0.7077,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999756795637625,
+      "loss": 0.5496,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999745866484794,
+      "loss": 0.6808,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999734697133543,
+      "loss": 0.3555,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999723287584055,
+      "loss": 0.55,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999971163783651,
+      "loss": 0.4174,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999699747891096,
+      "loss": 0.2511,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999687617748,
+      "loss": 0.3682,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999967524740741,
+      "loss": 0.9861,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999966263686955,
+      "loss": 0.4334,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999964978613459,
+      "loss": 0.3246,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999963669520275,
+      "loss": 0.4127,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999623364074246,
+      "loss": 1.2994,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999960979274928,
+      "loss": 0.3928,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999959598122808,
+      "loss": 0.5447,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999581929510854,
+      "loss": 1.3063,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999567637597845,
+      "loss": 0.7071,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999955310548926,
+      "loss": 0.7096,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999953833318535,
+      "loss": 0.3107,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999523320686344,
+      "loss": 1.7141,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999950806799249,
+      "loss": 0.4814,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999949257510401,
+      "loss": 0.4892,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999947684202118,
+      "loss": 0.5439,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999460868744234,
+      "loss": 0.5369,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999444655273434,
+      "loss": 0.5828,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999942820160904,
+      "loss": 0.8133,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999941150775131,
+      "loss": 0.723,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999394573700523,
+      "loss": 2.4967,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999937739945694,
+      "loss": 0.6977,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999935998502084,
+      "loss": 0.0939,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299993423303925,
+      "loss": 0.9066,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299993244355722,
+      "loss": 0.8641,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999306300560233,
+      "loss": 0.3496,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999928792535688,
+      "loss": 0.4615,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999269309962447,
+      "loss": 0.9519,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999925045437723,
+      "loss": 0.5484,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999923135860153,
+      "loss": 5.6815,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999921202263564,
+      "loss": 0.2777,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999192446479887,
+      "loss": 0.0841,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999917263013457,
+      "loss": 0.1023,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999152573600025,
+      "loss": 0.5286,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999132276876554,
+      "loss": 0.8989,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999111739964496,
+      "loss": 1.5437,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999909096286417,
+      "loss": 1.1663,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029999069945575913,
+      "loss": 0.4474,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999904868810006,
+      "loss": 0.0132,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999902719043696,
+      "loss": 0.1196,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999900545258694,
+      "loss": 0.8682,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998983474550364,
+      "loss": 0.7771,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999896125632758,
+      "loss": 0.5106,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999893879791893,
+      "loss": 2.946,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299989160993248,
+      "loss": 1.9552,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999889316054553,
+      "loss": 1.8721,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299988699815815,
+      "loss": 0.0067,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998846562433076,
+      "loss": 1.5086,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999882290310064,
+      "loss": 0.6052,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999879900358456,
+      "loss": 0.0507,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999877486388522,
+      "loss": 1.2328,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998750484003017,
+      "loss": 0.8417,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998725863938333,
+      "loss": 0.3224,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999870100369156,
+      "loss": 0.5401,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998675903263104,
+      "loss": 0.804,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998650562653367,
+      "loss": 4.3887,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998624981862745,
+      "loss": 0.0483,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998599160891657,
+      "loss": 0.6673,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998573099740514,
+      "loss": 0.2473,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999854679840973,
+      "loss": 0.6671,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998520256899734,
+      "loss": 0.2406,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998493475210937,
+      "loss": 0.4182,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999846645334378,
+      "loss": 2.0499,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999843919129869,
+      "loss": 0.4257,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999841168907611,
+      "loss": 0.0019,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998383946676474,
+      "loss": 0.7959,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998355964100235,
+      "loss": 0.1089,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999832774134783,
+      "loss": 2.9715,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998299278419715,
+      "loss": 0.7391,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998270575316345,
+      "loss": 0.7319,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998241632038183,
+      "loss": 3.2964,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998212448585686,
+      "loss": 0.3525,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999818302495933,
+      "loss": 0.8063,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998153361159583,
+      "loss": 0.2223,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999812345718692,
+      "loss": 0.0074,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999809331304182,
+      "loss": 0.5793,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029998062928724757,
+      "loss": 0.5266,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999803230423623,
+      "loss": 0.5553,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999803230423623,
+      "loss": 1.6031,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999800143957672,
+      "loss": 0.7125,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999797033474673,
+      "loss": 0.6833,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999793898974675,
+      "loss": 0.5578,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997907404577295,
+      "loss": 0.521,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999787557923886,
+      "loss": 0.424,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997843513731947,
+      "loss": 0.3912,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997811208057085,
+      "loss": 1.1264,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997778662214784,
+      "loss": 1.017,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999774587620556,
+      "loss": 0.7859,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997712850029953,
+      "loss": 0.7413,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999767958368848,
+      "loss": 2.7864,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999764607718168,
+      "loss": 0.0075,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997612330510086,
+      "loss": 0.5556,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997578343674236,
+      "loss": 3.1132,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997544116674676,
+      "loss": 1.5589,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999750964951196,
+      "loss": 1.1716,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999747494218663,
+      "loss": 0.0102,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997439994699246,
+      "loss": 0.7373,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997404807050374,
+      "loss": 0.0425,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997369379240567,
+      "loss": 0.6373,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299973337112704,
+      "loss": 0.3752,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997297803140437,
+      "loss": 0.5111,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999726165485126,
+      "loss": 0.0532,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999722526640345,
+      "loss": 0.2523,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997188637797585,
+      "loss": 0.6098,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999715176903425,
+      "loss": 0.4431,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997114660114036,
+      "loss": 0.2207,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997077311037535,
+      "loss": 0.8102,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029997039721805347,
+      "loss": 0.4433,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999700189241808,
+      "loss": 0.7388,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999696382287633,
+      "loss": 1.1607,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999692551318072,
+      "loss": 0.4863,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999688696333185,
+      "loss": 0.3484,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999684817333035,
+      "loss": 0.5244,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996809143176826,
+      "loss": 0.552,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999676987287191,
+      "loss": 0.289,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999673036241623,
+      "loss": 0.4618,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996690611810425,
+      "loss": 0.2009,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996650621055127,
+      "loss": 0.452,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996610390150977,
+      "loss": 2.1568,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999656991909861,
+      "loss": 3.054,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999652920789869,
+      "loss": 0.4962,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999648825655186,
+      "loss": 0.8852,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999644706505878,
+      "loss": 0.5776,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299964056334201,
+      "loss": 0.4406,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996363961636496,
+      "loss": 0.7579,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996322049708623,
+      "loss": 1.0634,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996279897637163,
+      "loss": 0.5465,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999623750542279,
+      "loss": 0.8095,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999619487306617,
+      "loss": 0.6515,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996152000568,
+      "loss": 0.5721,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999610888792896,
+      "loss": 0.1836,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999606553514975,
+      "loss": 0.654,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029996021942231056,
+      "loss": 1.8457,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999597810917357,
+      "loss": 0.2581,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995934035978,
+      "loss": 0.3528,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995889722645054,
+      "loss": 0.4445,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995845169175444,
+      "loss": 0.5048,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995800375569873,
+      "loss": 1.1177,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999575534182907,
+      "loss": 0.489,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995710067953746,
+      "loss": 0.7172,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999566455394463,
+      "loss": 0.524,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999561879980246,
+      "loss": 0.9177,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999557280552795,
+      "loss": 0.7532,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999552657112185,
+      "loss": 0.1295,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299954800965849,
+      "loss": 0.6342,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999543338191784,
+      "loss": 0.7693,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999538642712142,
+      "loss": 0.6009,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995339232196386,
+      "loss": 0.5431,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995291797143506,
+      "loss": 0.4771,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999524412196353,
+      "loss": 2.1071,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999519620665722,
+      "loss": 0.7119,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995148051225354,
+      "loss": 0.968,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995099655668694,
+      "loss": 0.0052,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999505101998802,
+      "loss": 0.9679,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029995002144184103,
+      "loss": 0.4108,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999495302825774,
+      "loss": 3.0098,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994903672209697,
+      "loss": 2.4187,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994854076040783,
+      "loss": 0.5288,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999480423975178,
+      "loss": 1.483,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994754163343494,
+      "loss": 0.8515,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994703846816724,
+      "loss": 0.9621,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999465329017228,
+      "loss": 0.2986,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999460249341096,
+      "loss": 0.6224,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999455145653359,
+      "loss": 0.7774,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994500179540977,
+      "loss": 0.6289,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999444866243395,
+      "loss": 1.4427,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994396905213335,
+      "loss": 0.6656,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994344907879954,
+      "loss": 0.5839,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994292670434643,
+      "loss": 0.9333,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999424019287824,
+      "loss": 2.0752,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999418747521158,
+      "loss": 0.9285,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999413451743551,
+      "loss": 0.457,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029994081319550877,
+      "loss": 1.1892,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999402788155854,
+      "loss": 1.6005,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999397420345935,
+      "loss": 2.2713,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999392028525416,
+      "loss": 0.4316,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993866126943843,
+      "loss": 0.4706,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993811728529267,
+      "loss": 0.3741,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993757090011295,
+      "loss": 0.5472,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299937022113908,
+      "loss": 0.3128,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993647092668675,
+      "loss": 1.3186,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999359173384579,
+      "loss": 0.0113,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993536134923036,
+      "loss": 1.0267,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993480295901305,
+      "loss": 0.3336,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993424216781485,
+      "loss": 0.3183,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993367897564475,
+      "loss": 0.3732,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993311338251193,
+      "loss": 0.0128,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999325453884252,
+      "loss": 0.7211,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999319749933938,
+      "loss": 0.964,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993140219742686,
+      "loss": 2.0126,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993082700053355,
+      "loss": 0.5211,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029993024940272304,
+      "loss": 0.4686,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992966940400455,
+      "loss": 0.569,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992908700438747,
+      "loss": 0.6212,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992850220388105,
+      "loss": 1.2915,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999279150024947,
+      "loss": 0.8984,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999273254002378,
+      "loss": 1.6295,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999267333971198,
+      "loss": 1.1624,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992613899315016,
+      "loss": 0.3895,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999255421883384,
+      "loss": 0.5453,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992494298269407,
+      "loss": 0.4575,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992434137622684,
+      "loss": 0.0689,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992373736894624,
+      "loss": 0.7966,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000299923130960862,
+      "loss": 0.4863,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992252215198386,
+      "loss": 0.1806,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999219109423215,
+      "loss": 1.1082,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999212973318847,
+      "loss": 0.598,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992068132068337,
+      "loss": 0.2244,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029992006290872734,
+      "loss": 0.6915,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991944209602645,
+      "loss": 2.0011,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991881888259073,
+      "loss": 0.6528,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999181932684301,
+      "loss": 0.5699,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991756525355465,
+      "loss": 0.2,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991693483797434,
+      "loss": 1.2829,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999163020216993,
+      "loss": 0.0038,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999156668047397,
+      "loss": 0.2096,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991502918710574,
+      "loss": 0.2529,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999143891688075,
+      "loss": 0.0098,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999137467498554,
+      "loss": 0.3423,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999131019302595,
+      "loss": 1.5642,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00029991245471003037,
+      "loss": 0.7957,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002999118050891782,
+      "loss": 0.3336,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999111530677135,
+      "loss": 0.321,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999104986456466,
+      "loss": 0.783,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999098418229881,
+      "loss": 0.8318,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999091825997485,
+      "loss": 0.7515,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029990852097593827,
+      "loss": 0.6569,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999078569515681,
+      "loss": 0.0101,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999071905266486,
+      "loss": 1.3934,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029990652170119034,
+      "loss": 0.3622,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029990585047520415,
+      "loss": 0.2884,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029990517684870077,
+      "loss": 0.4,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029990450082169096,
+      "loss": 0.2324,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999038223941855,
+      "loss": 0.1775,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999031415661953,
+      "loss": 0.0027,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029990245833773127,
+      "loss": 1.7972,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999017727088043,
+      "loss": 0.3842,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002999010846794255,
+      "loss": 0.0025,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029990039424960575,
+      "loss": 0.3474,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998997014193562,
+      "loss": 0.6299,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029989900618868786,
+      "loss": 0.3241,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998983085576119,
+      "loss": 0.6135,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998976085261395,
+      "loss": 0.9595,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029989690609428187,
+      "loss": 0.0035,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029989620126205023,
+      "loss": 0.8464,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998954940294559,
+      "loss": 0.5995,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029989478439651024,
+      "loss": 0.0016,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998940723632245,
+      "loss": 0.14,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029989335792961024,
+      "loss": 2.0168,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029989335792961024,
+      "loss": 0.2094,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998926410956787,
+      "loss": 0.5108,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029989192186144156,
+      "loss": 0.8928,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998912002269102,
+      "loss": 0.3447,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998904761920962,
+      "loss": 1.2306,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029988974975701127,
+      "loss": 0.6787,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029988902092166687,
+      "loss": 0.2462,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029988828968607477,
+      "loss": 3.6597,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998875560502467,
+      "loss": 0.5929,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998868200141943,
+      "loss": 1.0922,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998860815779295,
+      "loss": 0.8976,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299885340741464,
+      "loss": 0.5819,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998845975048097,
+      "loss": 0.0466,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998838518679786,
+      "loss": 1.0847,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029988310383098247,
+      "loss": 0.9847,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998823533938334,
+      "loss": 0.7482,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998816005565434,
+      "loss": 0.4283,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029988084531912447,
+      "loss": 0.7431,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029988008768158875,
+      "loss": 0.2175,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998793276439484,
+      "loss": 0.2961,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029987856520621555,
+      "loss": 0.7113,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998778003684024,
+      "loss": 0.2944,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998770331305212,
+      "loss": 0.6408,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998762634925843,
+      "loss": 0.1935,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998754914546039,
+      "loss": 0.0936,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998747170165925,
+      "loss": 0.803,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998739401785624,
+      "loss": 2.3781,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998731609405261,
+      "loss": 1.0266,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299872379302496,
+      "loss": 1.708,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998715952644848,
+      "loss": 3.5539,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998708088265048,
+      "loss": 0.385,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029987001998856873,
+      "loss": 2.2036,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029986922875068926,
+      "loss": 0.5798,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299868435112879,
+      "loss": 0.5154,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998676390751507,
+      "loss": 3.527,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029986684063751704,
+      "loss": 0.6486,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029986603979999086,
+      "loss": 0.9148,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029986523656258497,
+      "loss": 0.3921,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998644309253122,
+      "loss": 1.2489,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998636228881855,
+      "loss": 1.4371,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998628124512178,
+      "loss": 0.8653,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998619996144221,
+      "loss": 0.6369,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998611843778114,
+      "loss": 0.5806,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998603667413987,
+      "loss": 0.2169,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985954670519715,
+      "loss": 1.8133,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985872426921986,
+      "loss": 0.6107,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985789943348,
+      "loss": 0.4822,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998570721979908,
+      "loss": 0.0069,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998562425627655,
+      "loss": 0.754,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998554105278174,
+      "loss": 0.3175,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985457609315973,
+      "loss": 0.0749,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985373925880596,
+      "loss": 1.0707,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998529000247694,
+      "loss": 1.9348,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985205839106366,
+      "loss": 0.404,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029985121435770207,
+      "loss": 0.8832,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998503679246981,
+      "loss": 0.6141,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029984951909206543,
+      "loss": 0.3446,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998486678598176,
+      "loss": 0.6046,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998478142279683,
+      "loss": 0.6271,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029984695819653106,
+      "loss": 0.5219,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998460997655197,
+      "loss": 0.2009,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029984523893494797,
+      "loss": 0.3517,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998443757048296,
+      "loss": 0.8978,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029984351007517843,
+      "loss": 0.761,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029984264204600835,
+      "loss": 0.3242,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029984177161733325,
+      "loss": 1.0781,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299840898789167,
+      "loss": 0.4241,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998400235615237,
+      "loss": 0.2888,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983914593441726,
+      "loss": 0.9145,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983826590786184,
+      "loss": 1.1606,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998373834818714,
+      "loss": 0.3215,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983649865646014,
+      "loss": 0.3304,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998356114316423,
+      "loss": 1.107,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983472180743193,
+      "loss": 0.2646,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983382978384336,
+      "loss": 0.2637,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983293536089094,
+      "loss": 0.9527,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983203853858886,
+      "loss": 0.4542,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029983113931695155,
+      "loss": 0.7127,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998302376959935,
+      "loss": 0.1353,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982933367572897,
+      "loss": 0.3287,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982842725617253,
+      "loss": 0.5632,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982751843733875,
+      "loss": 1.5377,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982660721924206,
+      "loss": 0.2818,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982569360189715,
+      "loss": 0.2437,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982477758531864,
+      "loss": 0.5831,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982385916952107,
+      "loss": 3.5795,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982293835451935,
+      "loss": 1.0205,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029982201514032805,
+      "loss": 0.5634,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998210895269621,
+      "loss": 0.4732,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998201615144362,
+      "loss": 0.819,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998192311027653,
+      "loss": 0.697,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029981829829196424,
+      "loss": 0.4723,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029981736308204795,
+      "loss": 1.2741,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998164254730315,
+      "loss": 0.4337,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029981548546492975,
+      "loss": 1.2639,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998145430577579,
+      "loss": 1.2521,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299813598251531,
+      "loss": 0.6589,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998126510462641,
+      "loss": 0.5091,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998117014419725,
+      "loss": 0.6194,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998107494386713,
+      "loss": 2.2732,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980979503637577,
+      "loss": 1.1198,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980883823510127,
+      "loss": 0.426,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299807879034863,
+      "loss": 0.1394,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998069174356764,
+      "loss": 1.2332,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998059534375568,
+      "loss": 0.3686,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998049870405197,
+      "loss": 0.629,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980401824458063,
+      "loss": 0.7355,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980304704975495,
+      "loss": 1.0268,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998020734560583,
+      "loss": 0.9582,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002998010974635063,
+      "loss": 0.4177,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029980011907211455,
+      "loss": 0.3177,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979913828189874,
+      "loss": 0.3716,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979815509287446,
+      "loss": 0.6625,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997971695050576,
+      "loss": 0.6055,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979618151846385,
+      "loss": 0.3549,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997951911331091,
+      "loss": 1.0981,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979419834900913,
+      "loss": 0.9803,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997932031661799,
+      "loss": 0.3675,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979220558463737,
+      "loss": 0.0049,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029979120560439743,
+      "loss": 0.7795,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997902032254762,
+      "loss": 0.5777,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997891984478896,
+      "loss": 0.3353,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997881912716538,
+      "loss": 0.1844,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029978718169678493,
+      "loss": 0.4885,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029978616972329913,
+      "loss": 0.6982,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029978515535121264,
+      "loss": 1.782,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029978413858054167,
+      "loss": 0.999,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029978311941130254,
+      "loss": 0.4867,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997820978435115,
+      "loss": 0.6495,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029978107387718496,
+      "loss": 0.3001,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029978004751233937,
+      "loss": 0.4585,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029977901874899104,
+      "loss": 0.3027,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029977798758715657,
+      "loss": 0.1915,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029977695402685233,
+      "loss": 0.3148,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299775918068095,
+      "loss": 0.5644,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997748797109012,
+      "loss": 0.3074,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997738389552874,
+      "loss": 0.6409,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997727958012704,
+      "loss": 0.6965,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997717502488668,
+      "loss": 1.0783,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997707022980934,
+      "loss": 0.4707,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029976965194896695,
+      "loss": 0.9336,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997685992015044,
+      "loss": 0.3597,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997675440557224,
+      "loss": 0.5038,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299766486511638,
+      "loss": 0.356,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997654265692681,
+      "loss": 0.6105,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997643642286296,
+      "loss": 0.7791,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029976329948973963,
+      "loss": 0.9433,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029976223235261514,
+      "loss": 0.2018,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997611628172733,
+      "loss": 1.031,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997600908837312,
+      "loss": 0.0523,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029975901655200594,
+      "loss": 0.9312,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997579398221148,
+      "loss": 0.6132,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029975686069407513,
+      "loss": 1.3225,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299755779167904,
+      "loss": 0.5417,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029975469524361876,
+      "loss": 0.559,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029975360892123693,
+      "loss": 0.3718,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997525202007757,
+      "loss": 2.4617,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997514290822527,
+      "loss": 0.5204,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029975033556568526,
+      "loss": 0.5522,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029974923965109094,
+      "loss": 0.518,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997481413384873,
+      "loss": 0.3818,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029974704062789195,
+      "loss": 0.5278,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997459375193224,
+      "loss": 0.2671,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997448320127965,
+      "loss": 0.5331,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997437241083318,
+      "loss": 2.2319,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997426138059461,
+      "loss": 0.2925,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997415011056571,
+      "loss": 0.4191,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029974038600748276,
+      "loss": 0.3273,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997392685114409,
+      "loss": 0.7598,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029973814861754933,
+      "loss": 0.4249,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299737026325826,
+      "loss": 0.6454,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299735901636289,
+      "loss": 0.5924,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997347745489562,
+      "loss": 0.7876,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029973364506384576,
+      "loss": 0.4472,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029973251318097567,
+      "loss": 0.527,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029973137890036407,
+      "loss": 0.9478,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997302422220292,
+      "loss": 0.2056,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997291031459892,
+      "loss": 0.8,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997279616722624,
+      "loss": 0.6914,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997268178008669,
+      "loss": 0.3286,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997256715318212,
+      "loss": 0.5214,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029972452286514345,
+      "loss": 0.5975,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997233718008523,
+      "loss": 0.3803,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299722218338966,
+      "loss": 0.8148,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997210624795031,
+      "loss": 1.3607,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029971990422248204,
+      "loss": 0.9083,
+      "step": 586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029971874356792146,
+      "loss": 0.3893,
+      "step": 587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029971758051583984,
+      "loss": 0.4363,
+      "step": 588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029971641506625594,
+      "loss": 0.931,
+      "step": 589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997152472191883,
+      "loss": 0.4755,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997140769746557,
+      "loss": 0.8721,
+      "step": 591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029971290433267677,
+      "loss": 3.0693,
+      "step": 592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029971172929327044,
+      "loss": 0.6307,
+      "step": 593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997105518564554,
+      "loss": 0.3584,
+      "step": 594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997093720222506,
+      "loss": 1.1073,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029970818979067486,
+      "loss": 1.9136,
+      "step": 596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029970700516174714,
+      "loss": 0.2816,
+      "step": 597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997058181354865,
+      "loss": 0.0981,
+      "step": 598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997046287119117,
+      "loss": 0.3192,
+      "step": 599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000299703436891042,
+      "loss": 1.1676,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029970224267289653,
+      "loss": 0.3294,
+      "step": 601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002997010460574942,
+      "loss": 2.1328,
+      "step": 602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996998470448544,
+      "loss": 0.4742,
+      "step": 603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969864563499614,
+      "loss": 0.406,
+      "step": 604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969744182793874,
+      "loss": 0.4066,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969623562370144,
+      "loss": 0.3417,
+      "step": 606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969502702230366,
+      "loss": 0.3874,
+      "step": 607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969381602376464,
+      "loss": 1.0648,
+      "step": 608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969260262810383,
+      "loss": 0.1198,
+      "step": 609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969138683534066,
+      "loss": 0.5071,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029969016864549457,
+      "loss": 2.503,
+      "step": 611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996889480585851,
+      "loss": 0.938,
+      "step": 612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996877250746317,
+      "loss": 0.3216,
+      "step": 613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996864996936541,
+      "loss": 3.4333,
+      "step": 614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996852719156719,
+      "loss": 0.8181,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996840417407047,
+      "loss": 0.3687,
+      "step": 616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029968280916877214,
+      "loss": 0.3899,
+      "step": 617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029968157419989406,
+      "loss": 0.3349,
+      "step": 618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996803368340902,
+      "loss": 0.2812,
+      "step": 619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996790970713804,
+      "loss": 0.4847,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002996778549117846,
+      "loss": 0.6741,
+      "step": 621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029967661035532246,
+      "loss": 1.2943,
+      "step": 622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00029967536340201405,
+      "loss": 0.0219,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029967411405187937,
+      "loss": 0.6679,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029967286230493835,
+      "loss": 0.7483,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029967160816121106,
+      "loss": 0.4377,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996703516207176,
+      "loss": 1.035,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996690926834781,
+      "loss": 0.4486,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996678313495127,
+      "loss": 0.1569,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996665676188416,
+      "loss": 1.0024,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299665301491485,
+      "loss": 0.5543,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996640329674633,
+      "loss": 0.5298,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996627620467966,
+      "loss": 0.4987,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996614887295055,
+      "loss": 3.366,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029966021301561024,
+      "loss": 0.865,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029965893490513124,
+      "loss": 1.4343,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299657654398089,
+      "loss": 0.0179,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996563714945041,
+      "loss": 0.8609,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996550861943969,
+      "loss": 0.2364,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029965379849778817,
+      "loss": 0.7264,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029965250840469846,
+      "loss": 0.0133,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996512159151484,
+      "loss": 0.6661,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029964992102915873,
+      "loss": 1.0984,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996486237467502,
+      "loss": 0.8999,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996473240679435,
+      "loss": 0.6427,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996460219927595,
+      "loss": 0.4339,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299644717521219,
+      "loss": 0.3558,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299643410653343,
+      "loss": 0.2202,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029964210138915237,
+      "loss": 0.6986,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029964078972866804,
+      "loss": 0.7068,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996394756719111,
+      "loss": 0.3719,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996381592189024,
+      "loss": 0.9851,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996368403696632,
+      "loss": 0.6485,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029963551912421467,
+      "loss": 0.4309,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029963419548257784,
+      "loss": 0.3915,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029963286944477386,
+      "loss": 0.5389,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996315410108241,
+      "loss": 1.9492,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996302101807498,
+      "loss": 0.3634,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029962887695457224,
+      "loss": 0.8859,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029962754133231275,
+      "loss": 1.3406,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029962620331399275,
+      "loss": 0.6705,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029962486289963365,
+      "loss": 0.9486,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996235200892569,
+      "loss": 0.2311,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996221748828841,
+      "loss": 0.2675,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029962082728053667,
+      "loss": 0.2972,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996194772822363,
+      "loss": 0.7348,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996181248880045,
+      "loss": 0.828,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029961677009786303,
+      "loss": 0.0713,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029961541291183345,
+      "loss": 0.4655,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996140533299376,
+      "loss": 1.0571,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996126913521972,
+      "loss": 0.4444,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029961132697863414,
+      "loss": 2.4133,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029960996020927017,
+      "loss": 0.0452,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029960859104412725,
+      "loss": 0.2483,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029960721948322724,
+      "loss": 0.1143,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996058455265922,
+      "loss": 0.1648,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299604469174244,
+      "loss": 0.3379,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029960309042620484,
+      "loss": 0.6503,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029960170928249665,
+      "loss": 0.0833,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002996003257431416,
+      "loss": 1.0383,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995989398081619,
+      "loss": 0.5664,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995975514775796,
+      "loss": 0.5495,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995961607514171,
+      "loss": 1.0748,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995947676296966,
+      "loss": 0.7079,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029959337211244043,
+      "loss": 0.4719,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029959197419967084,
+      "loss": 0.4273,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029959057389141033,
+      "loss": 0.6066,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029958917118768133,
+      "loss": 0.5683,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029958776608850614,
+      "loss": 0.5021,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995863585939075,
+      "loss": 0.4476,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029958494870390773,
+      "loss": 1.8491,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029958353641852957,
+      "loss": 0.6671,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995821217377955,
+      "loss": 0.7913,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995807046617283,
+      "loss": 0.2723,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029957928519035057,
+      "loss": 0.4616,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995778633236851,
+      "loss": 0.2105,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029957643906175464,
+      "loss": 0.9581,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995750124045819,
+      "loss": 0.5218,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029957358335218993,
+      "loss": 0.2372,
+      "step": 699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995721519046015,
+      "loss": 0.2521,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029957071806183946,
+      "loss": 1.9224,
+      "step": 701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995692818239269,
+      "loss": 0.7851,
+      "step": 702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029956784319088673,
+      "loss": 0.7684,
+      "step": 703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029956640216274204,
+      "loss": 0.6064,
+      "step": 704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995649587395159,
+      "loss": 0.2881,
+      "step": 705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029956351292123134,
+      "loss": 0.493,
+      "step": 706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995620647079117,
+      "loss": 1.591,
+      "step": 707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029956061409957993,
+      "loss": 0.2488,
+      "step": 708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029955916109625946,
+      "loss": 0.5169,
+      "step": 709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029955770569797347,
+      "loss": 0.2644,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029955624790474526,
+      "loss": 0.2995,
+      "step": 711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995547877165982,
+      "loss": 0.6723,
+      "step": 712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029955332513355564,
+      "loss": 0.4755,
+      "step": 713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029955186015564106,
+      "loss": 0.9366,
+      "step": 714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029955039278287794,
+      "loss": 0.8933,
+      "step": 715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029954892301528965,
+      "loss": 0.0969,
+      "step": 716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995474508528998,
+      "loss": 0.6746,
+      "step": 717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299545976295732,
+      "loss": 1.277,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995444993438098,
+      "loss": 0.4479,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029954301999715686,
+      "loss": 0.0084,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029954153825579693,
+      "loss": 0.3678,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995400541197537,
+      "loss": 0.5088,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995385675890509,
+      "loss": 0.304,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995370786637124,
+      "loss": 0.0662,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299535587343762,
+      "loss": 0.422,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995340936292235,
+      "loss": 0.7488,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029953259752012103,
+      "loss": 0.1403,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029953109901647836,
+      "loss": 0.6124,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029952959811831956,
+      "loss": 1.3099,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029952809482566866,
+      "loss": 0.7573,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029952658913854973,
+      "loss": 1.0093,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995250810569869,
+      "loss": 1.0035,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029952357058100437,
+      "loss": 0.8918,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029952205771062614,
+      "loss": 0.2706,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029952054244587663,
+      "loss": 0.4327,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029951902478678,
+      "loss": 0.9155,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995175047333606,
+      "loss": 0.4515,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995159822856427,
+      "loss": 1.0384,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029951445744365085,
+      "loss": 0.9349,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995129302074093,
+      "loss": 0.5321,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995114005769425,
+      "loss": 3.8804,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029950986855227507,
+      "loss": 0.4829,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029950833413343146,
+      "loss": 0.7533,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995067973204363,
+      "loss": 0.3983,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029950525811331413,
+      "loss": 0.353,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995037165120896,
+      "loss": 0.5139,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002995021725167874,
+      "loss": 0.2974,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029950062612743236,
+      "loss": 0.5466,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029949907734404913,
+      "loss": 0.6046,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994975261666625,
+      "loss": 0.8182,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994959725952974,
+      "loss": 0.6434,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029949441662997865,
+      "loss": 0.6042,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994928582707312,
+      "loss": 0.4315,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029949129751757995,
+      "loss": 1.0256,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029948973437054994,
+      "loss": 0.5118,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994881688296662,
+      "loss": 0.5391,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994866008949537,
+      "loss": 0.4383,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994850305664377,
+      "loss": 0.4613,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994834578441433,
+      "loss": 0.4043,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994818827280957,
+      "loss": 0.3269,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029948030521832,
+      "loss": 0.9401,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994787253148416,
+      "loss": 0.3665,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994771430176858,
+      "loss": 0.4662,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994755583268778,
+      "loss": 0.3799,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029947397124244305,
+      "loss": 0.5422,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029947238176440704,
+      "loss": 0.3896,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029947078989279514,
+      "loss": 0.0436,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029946919562763287,
+      "loss": 0.3844,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029946759896894577,
+      "loss": 0.2607,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029946599991675937,
+      "loss": 1.2097,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994643984710993,
+      "loss": 1.1048,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994627946319912,
+      "loss": 0.1256,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994611883994608,
+      "loss": 1.306,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029945957977353373,
+      "loss": 0.5848,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994579687542358,
+      "loss": 1.3041,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029945635534159277,
+      "loss": 0.3233,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994547395356306,
+      "loss": 4.6411,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299453121336375,
+      "loss": 0.2731,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029945150074385204,
+      "loss": 0.1473,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994498777580875,
+      "loss": 1.4458,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994482523791075,
+      "loss": 0.5393,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029944662460693804,
+      "loss": 0.3147,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994449944416052,
+      "loss": 0.1741,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299443361883135,
+      "loss": 0.5847,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029944172693155364,
+      "loss": 0.4571,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994400895868873,
+      "loss": 0.3589,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994384498491623,
+      "loss": 1.2126,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994368077184047,
+      "loss": 0.674,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029943516319464093,
+      "loss": 1.154,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994335162778973,
+      "loss": 0.8745,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994318669682002,
+      "loss": 0.533,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299430215265576,
+      "loss": 0.6225,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994285611700511,
+      "loss": 0.9474,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994269046816521,
+      "loss": 0.8173,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029942524580040556,
+      "loss": 0.6842,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029942358452633786,
+      "loss": 0.5623,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994219208594757,
+      "loss": 0.4162,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994202547998458,
+      "loss": 0.3911,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994185863474747,
+      "loss": 0.7209,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994169155023892,
+      "loss": 0.3378,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994152422646161,
+      "loss": 0.212,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994135666341821,
+      "loss": 0.4292,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994118886111141,
+      "loss": 2.8947,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029941020819543887,
+      "loss": 1.5582,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029940852538718344,
+      "loss": 0.9103,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994068401863747,
+      "loss": 0.558,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994051525930396,
+      "loss": 0.8993,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029940346260720526,
+      "loss": 0.1624,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002994017702288987,
+      "loss": 0.5112,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029940007545814696,
+      "loss": 1.7019,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029939837829497725,
+      "loss": 0.6312,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993966787394167,
+      "loss": 0.5053,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993949767914926,
+      "loss": 0.8625,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029939327245123216,
+      "loss": 0.4833,
+      "step": 814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029939156571866265,
+      "loss": 0.4326,
+      "step": 815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993898565938114,
+      "loss": 0.4521,
+      "step": 816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029938814507670584,
+      "loss": 0.5834,
+      "step": 817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993864311673733,
+      "loss": 0.4002,
+      "step": 818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993847148658413,
+      "loss": 0.42,
+      "step": 819
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029938299617213724,
+      "loss": 1.0622,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029938127508628875,
+      "loss": 0.4922,
+      "step": 821
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029937955160832326,
+      "loss": 0.3832,
+      "step": 822
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993778257382685,
+      "loss": 0.3644,
+      "step": 823
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299376097476152,
+      "loss": 0.3039,
+      "step": 824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993743668220015,
+      "loss": 1.1071,
+      "step": 825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029937263377584464,
+      "loss": 0.7642,
+      "step": 826
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993708983377093,
+      "loss": 0.4299,
+      "step": 827
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993691605076232,
+      "loss": 0.4362,
+      "step": 828
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993674202856141,
+      "loss": 0.5481,
+      "step": 829
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029936567767171004,
+      "loss": 0.3224,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029936393266593876,
+      "loss": 0.9821,
+      "step": 831
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993621852683282,
+      "loss": 0.24,
+      "step": 832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993604354789065,
+      "loss": 0.3447,
+      "step": 833
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993586832977015,
+      "loss": 0.7007,
+      "step": 834
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029935692872474145,
+      "loss": 0.5945,
+      "step": 835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029935517176005427,
+      "loss": 0.2171,
+      "step": 836
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029935341240366816,
+      "loss": 0.587,
+      "step": 837
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993516506556113,
+      "loss": 2.3494,
+      "step": 838
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993498865159119,
+      "loss": 0.925,
+      "step": 839
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993481199845982,
+      "loss": 0.2795,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029934635106169857,
+      "loss": 0.4914,
+      "step": 841
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993445797472412,
+      "loss": 0.6274,
+      "step": 842
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029934280604125454,
+      "loss": 0.3978,
+      "step": 843
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299341029943767,
+      "loss": 0.7608,
+      "step": 844
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029933925145480693,
+      "loss": 0.2048,
+      "step": 845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993374705744029,
+      "loss": 1.1888,
+      "step": 846
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993356873025834,
+      "loss": 0.2022,
+      "step": 847
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299333901639377,
+      "loss": 0.2959,
+      "step": 848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993321135848123,
+      "loss": 1.2035,
+      "step": 849
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029933032313891796,
+      "loss": 0.7232,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029932853030172253,
+      "loss": 1.8751,
+      "step": 851
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993267350732548,
+      "loss": 0.8712,
+      "step": 852
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993249374535435,
+      "loss": 0.3461,
+      "step": 853
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993231374426175,
+      "loss": 0.7518,
+      "step": 854
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993213350405055,
+      "loss": 0.6311,
+      "step": 855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029931953024723646,
+      "loss": 0.3359,
+      "step": 856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993177230628392,
+      "loss": 1.1701,
+      "step": 857
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029931591348734276,
+      "loss": 0.7335,
+      "step": 858
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000299314101520776,
+      "loss": 0.2086,
+      "step": 859
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029931228716316806,
+      "loss": 0.4079,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993104704145478,
+      "loss": 0.0395,
+      "step": 861
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029930865127494454,
+      "loss": 0.266,
+      "step": 862
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993068297443873,
+      "loss": 4.7435,
+      "step": 863
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029930500582290524,
+      "loss": 1.2128,
+      "step": 864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002993031795105276,
+      "loss": 0.027,
+      "step": 865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029930135080728364,
+      "loss": 0.0161,
+      "step": 866
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002992995197132026,
+      "loss": 1.3555,
+      "step": 867
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029929768622831383,
+      "loss": 0.7428,
+      "step": 868
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029929585035264663,
+      "loss": 0.3911,
+      "step": 869
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002992940120862305,
+      "loss": 0.5341,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029929217142909476,
+      "loss": 0.7444,
+      "step": 871
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00029929032838126897,
+      "loss": 0.318,
+      "step": 872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029928848294278265,
+      "loss": 0.4635,
+      "step": 873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029928663511366533,
+      "loss": 0.1245,
+      "step": 874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992847848939466,
+      "loss": 0.2531,
+      "step": 875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992829322836561,
+      "loss": 2.031,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992810772828234,
+      "loss": 0.0135,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029927921989147835,
+      "loss": 0.8338,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992773601096506,
+      "loss": 0.2692,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029927549793737004,
+      "loss": 0.2163,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992736333746663,
+      "loss": 0.5727,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992717664215694,
+      "loss": 0.9672,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029926989707810924,
+      "loss": 0.4481,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992680253443156,
+      "loss": 0.0584,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029926615122021866,
+      "loss": 0.4092,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992642747058482,
+      "loss": 1.6374,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992623958012345,
+      "loss": 0.4533,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992605145064075,
+      "loss": 0.5229,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029925863082139745,
+      "loss": 0.0265,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992567447462343,
+      "loss": 0.0087,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992548562809485,
+      "loss": 1.8335,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029925296542557007,
+      "loss": 0.4818,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992510721801295,
+      "loss": 0.9951,
+      "step": 893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029924917654465695,
+      "loss": 1.8667,
+      "step": 894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992472785191828,
+      "loss": 0.5549,
+      "step": 895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992453781037375,
+      "loss": 0.0283,
+      "step": 896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992434752983515,
+      "loss": 0.1736,
+      "step": 897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029924157010305515,
+      "loss": 0.4723,
+      "step": 898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029923966251787907,
+      "loss": 0.5739,
+      "step": 899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992377525428538,
+      "loss": 0.0038,
+      "step": 900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029923584017800986,
+      "loss": 0.6895,
+      "step": 901
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029923392542337794,
+      "loss": 0.6953,
+      "step": 902
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029923200827898863,
+      "loss": 0.5351,
+      "step": 903
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992300887448727,
+      "loss": 2.4263,
+      "step": 904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992281668210609,
+      "loss": 0.2575,
+      "step": 905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029922624250758393,
+      "loss": 0.9257,
+      "step": 906
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029922431580447264,
+      "loss": 6.7776,
+      "step": 907
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992223867117579,
+      "loss": 0.7361,
+      "step": 908
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992204552294706,
+      "loss": 0.5994,
+      "step": 909
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992185213576416,
+      "loss": 0.0169,
+      "step": 910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029921658509630196,
+      "loss": 0.0048,
+      "step": 911
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992146464454826,
+      "loss": 1.7575,
+      "step": 912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992127054052147,
+      "loss": 1.124,
+      "step": 913
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992107619755293,
+      "loss": 0.5455,
+      "step": 914
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992088161564574,
+      "loss": 0.6768,
+      "step": 915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029920686794803023,
+      "loss": 0.8236,
+      "step": 916
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029920491735027904,
+      "loss": 0.1598,
+      "step": 917
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000299202964363235,
+      "loss": 0.0109,
+      "step": 918
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002992010089869294,
+      "loss": 0.1562,
+      "step": 919
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991990512213936,
+      "loss": 1.0518,
+      "step": 920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029919709106665884,
+      "loss": 0.2737,
+      "step": 921
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029919512852275666,
+      "loss": 0.001,
+      "step": 922
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029919316358971833,
+      "loss": 0.1649,
+      "step": 923
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029919119626757546,
+      "loss": 0.3942,
+      "step": 924
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029918922655635945,
+      "loss": 2.7463,
+      "step": 925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991872544561019,
+      "loss": 0.3174,
+      "step": 926
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029918527996683437,
+      "loss": 0.001,
+      "step": 927
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991833030885884,
+      "loss": 0.3442,
+      "step": 928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029918132382139584,
+      "loss": 2.5999,
+      "step": 929
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029917934216528824,
+      "loss": 0.008,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029917735812029733,
+      "loss": 0.7515,
+      "step": 931
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029917537168645494,
+      "loss": 0.5513,
+      "step": 932
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991733828637928,
+      "loss": 0.3925,
+      "step": 933
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029917139165234287,
+      "loss": 0.2674,
+      "step": 934
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029916939805213694,
+      "loss": 0.2861,
+      "step": 935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000299167402063207,
+      "loss": 0.6527,
+      "step": 936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000299165403685585,
+      "loss": 0.401,
+      "step": 937
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029916340291930287,
+      "loss": 1.0149,
+      "step": 938
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029916139976439275,
+      "loss": 2.3198,
+      "step": 939
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991593942208867,
+      "loss": 0.4219,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991573862888167,
+      "loss": 0.0033,
+      "step": 941
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029915537596821506,
+      "loss": 0.4326,
+      "step": 942
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029915336325911395,
+      "loss": 1.0358,
+      "step": 943
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991513481615456,
+      "loss": 0.5227,
+      "step": 944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991493306755422,
+      "loss": 1.0824,
+      "step": 945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991473108011361,
+      "loss": 1.4255,
+      "step": 946
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991452885383597,
+      "loss": 0.8158,
+      "step": 947
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029914326388724537,
+      "loss": 0.2127,
+      "step": 948
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991412368478254,
+      "loss": 0.5739,
+      "step": 949
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029913920742013247,
+      "loss": 0.3536,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029913717560419887,
+      "loss": 0.8242,
+      "step": 951
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991351414000573,
+      "loss": 0.8309,
+      "step": 952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029913310480774016,
+      "loss": 0.2524,
+      "step": 953
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029913106582728026,
+      "loss": 0.2781,
+      "step": 954
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991290244587101,
+      "loss": 0.6135,
+      "step": 955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029912698070206245,
+      "loss": 0.6617,
+      "step": 956
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029912493455737,
+      "loss": 0.5783,
+      "step": 957
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029912288602466556,
+      "loss": 1.0577,
+      "step": 958
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991208351039819,
+      "loss": 0.8206,
+      "step": 959
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029911878179535184,
+      "loss": 0.2499,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029911672609880824,
+      "loss": 0.3826,
+      "step": 961
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991146680143841,
+      "loss": 6.0591,
+      "step": 962
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029911260754211235,
+      "loss": 0.7188,
+      "step": 963
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029911054468202605,
+      "loss": 0.3507,
+      "step": 964
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029910847943415803,
+      "loss": 2.4967,
+      "step": 965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991064117985416,
+      "loss": 1.0494,
+      "step": 966
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991043417752097,
+      "loss": 0.1917,
+      "step": 967
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029910226936419554,
+      "loss": 0.4005,
+      "step": 968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002991001945655323,
+      "loss": 0.5091,
+      "step": 969
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029909811737925326,
+      "loss": 1.5878,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029909603780539157,
+      "loss": 0.8375,
+      "step": 971
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990939558439807,
+      "loss": 0.4423,
+      "step": 972
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029909187149505383,
+      "loss": 0.5571,
+      "step": 973
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029908978475864435,
+      "loss": 1.1678,
+      "step": 974
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029908769563478574,
+      "loss": 0.4139,
+      "step": 975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029908560412351146,
+      "loss": 0.687,
+      "step": 976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029908351022485494,
+      "loss": 1.3629,
+      "step": 977
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990814139388498,
+      "loss": 0.7035,
+      "step": 978
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990793152655295,
+      "loss": 1.2759,
+      "step": 979
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029907721420492774,
+      "loss": 0.6908,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029907511075707814,
+      "loss": 0.5393,
+      "step": 981
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029907300492201436,
+      "loss": 0.7518,
+      "step": 982
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990708966997701,
+      "loss": 1.3136,
+      "step": 983
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990687860903792,
+      "loss": 0.794,
+      "step": 984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029906667309387535,
+      "loss": 0.2208,
+      "step": 985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029906455771029253,
+      "loss": 0.6426,
+      "step": 986
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029906243993966445,
+      "loss": 1.1011,
+      "step": 987
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029906031978202516,
+      "loss": 0.6311,
+      "step": 988
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029905819723740853,
+      "loss": 0.1482,
+      "step": 989
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990560723058486,
+      "loss": 4.3845,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990539449873794,
+      "loss": 1.5175,
+      "step": 991
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990518152820349,
+      "loss": 0.3298,
+      "step": 992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990496831898493,
+      "loss": 0.2175,
+      "step": 993
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990475487108567,
+      "loss": 0.4262,
+      "step": 994
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029904541184509135,
+      "loss": 0.2252,
+      "step": 995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029904327259258737,
+      "loss": 0.3779,
+      "step": 996
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990411309533791,
+      "loss": 0.0824,
+      "step": 997
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990389869275007,
+      "loss": 1.3027,
+      "step": 998
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029903684051498674,
+      "loss": 0.3737,
+      "step": 999
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990346917158714,
+      "loss": 0.7352,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990325405301891,
+      "loss": 0.9568,
+      "step": 1001
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029903038695797433,
+      "loss": 0.6533,
+      "step": 1002
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990282309992616,
+      "loss": 0.6535,
+      "step": 1003
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990260726540854,
+      "loss": 0.1985,
+      "step": 1004
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990239119224804,
+      "loss": 0.2155,
+      "step": 1005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000299021748804481,
+      "loss": 0.0006,
+      "step": 1006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029901958330012194,
+      "loss": 0.5097,
+      "step": 1007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990174154094379,
+      "loss": 0.3575,
+      "step": 1008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029901524513246367,
+      "loss": 0.2923,
+      "step": 1009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029901307246923384,
+      "loss": 0.6536,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029901089741978334,
+      "loss": 0.3554,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990087199841469,
+      "loss": 0.3623,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029900654016235955,
+      "loss": 0.3944,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029900435795445596,
+      "loss": 0.3554,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002990021733604713,
+      "loss": 0.5789,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989999863804404,
+      "loss": 0.3797,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989977970143984,
+      "loss": 1.0534,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989956052623802,
+      "loss": 0.5563,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989934111244211,
+      "loss": 0.5911,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298991214600556,
+      "loss": 0.4437,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989890156908203,
+      "loss": 1.0279,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989868143952491,
+      "loss": 0.6299,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029898461071387766,
+      "loss": 1.194,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029898240464674124,
+      "loss": 4.9652,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029898019619387523,
+      "loss": 0.513,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298977985355315,
+      "loss": 0.1193,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029897577213109587,
+      "loss": 0.4635,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989735565212533,
+      "loss": 0.4456,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029897133852582286,
+      "loss": 0.2489,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029896911814484,
+      "loss": 0.6559,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989668953783403,
+      "loss": 0.8587,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989646702263593,
+      "loss": 4.1724,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029896244268893266,
+      "loss": 3.4047,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989602127660961,
+      "loss": 0.7557,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029895798045788527,
+      "loss": 0.6626,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989557457643359,
+      "loss": 0.7883,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989535086854839,
+      "loss": 1.3351,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298951269221365,
+      "loss": 0.5135,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298949027372015,
+      "loss": 1.8054,
+      "step": 1039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029894678313746993,
+      "loss": 0.4182,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989445365177656,
+      "loss": 0.8641,
+      "step": 1041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989422875129381,
+      "loss": 0.3784,
+      "step": 1042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989400361230234,
+      "loss": 0.9567,
+      "step": 1043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029893778234805753,
+      "loss": 0.533,
+      "step": 1044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029893552618807663,
+      "loss": 0.8951,
+      "step": 1045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029893326764311674,
+      "loss": 4.1852,
+      "step": 1046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029893100671321413,
+      "loss": 0.9375,
+      "step": 1047
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029892874339840496,
+      "loss": 0.8032,
+      "step": 1048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029892647769872544,
+      "loss": 0.368,
+      "step": 1049
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989242096142119,
+      "loss": 0.4057,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989219391449007,
+      "loss": 0.4079,
+      "step": 1051
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029891966629082804,
+      "loss": 2.0127,
+      "step": 1052
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029891739105203054,
+      "loss": 0.4465,
+      "step": 1053
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029891511342854446,
+      "loss": 0.5166,
+      "step": 1054
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029891283342040634,
+      "loss": 0.3368,
+      "step": 1055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029891055102765265,
+      "loss": 0.4449,
+      "step": 1056
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029890826625032,
+      "loss": 2.3287,
+      "step": 1057
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989059790884449,
+      "loss": 1.4501,
+      "step": 1058
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989036895420641,
+      "loss": 0.9448,
+      "step": 1059
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002989013976112141,
+      "loss": 0.4909,
+      "step": 1060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029889910329593174,
+      "loss": 0.5025,
+      "step": 1061
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988968065962537,
+      "loss": 0.1797,
+      "step": 1062
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988945075122168,
+      "loss": 0.2703,
+      "step": 1063
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029889220604385774,
+      "loss": 0.4091,
+      "step": 1064
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029888990219121355,
+      "loss": 0.4542,
+      "step": 1065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029888759595432095,
+      "loss": 3.7607,
+      "step": 1066
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298885287333217,
+      "loss": 0.5195,
+      "step": 1067
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029888297632793857,
+      "loss": 1.3322,
+      "step": 1068
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988806629385227,
+      "loss": 1.0276,
+      "step": 1069
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988783471650066,
+      "loss": 0.1306,
+      "step": 1070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029887602900742705,
+      "loss": 1.0508,
+      "step": 1071
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988737084658214,
+      "loss": 0.7646,
+      "step": 1072
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988713855402267,
+      "loss": 0.6347,
+      "step": 1073
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988690602306802,
+      "loss": 0.7973,
+      "step": 1074
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988667325372191,
+      "loss": 0.0358,
+      "step": 1075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988644024598807,
+      "loss": 2.6185,
+      "step": 1076
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029886206999870236,
+      "loss": 1.5649,
+      "step": 1077
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988597351537213,
+      "loss": 2.2712,
+      "step": 1078
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029885739792497506,
+      "loss": 0.2692,
+      "step": 1079
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298855058312501,
+      "loss": 1.6648,
+      "step": 1080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029885271631633655,
+      "loss": 1.5293,
+      "step": 1081
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988503719365192,
+      "loss": 0.7427,
+      "step": 1082
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988480251730866,
+      "loss": 0.2913,
+      "step": 1083
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029884567602607627,
+      "loss": 1.6609,
+      "step": 1084
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029884332449552574,
+      "loss": 1.779,
+      "step": 1085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029884097058147286,
+      "loss": 0.0348,
+      "step": 1086
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988386142839552,
+      "loss": 1.9334,
+      "step": 1087
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988362556030104,
+      "loss": 1.3197,
+      "step": 1088
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029883389453867645,
+      "loss": 0.3256,
+      "step": 1089
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298831531090991,
+      "loss": 1.4009,
+      "step": 1090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029882916525999196,
+      "loss": 0.5543,
+      "step": 1091
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988267970457172,
+      "loss": 0.0235,
+      "step": 1092
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029882442644820465,
+      "loss": 1.834,
+      "step": 1093
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988220534674923,
+      "loss": 0.4342,
+      "step": 1094
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029881967810361804,
+      "loss": 0.554,
+      "step": 1095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029881730035662,
+      "loss": 1.6521,
+      "step": 1096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029881492022653625,
+      "loss": 0.407,
+      "step": 1097
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988125377134049,
+      "loss": 0.5393,
+      "step": 1098
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988101528172641,
+      "loss": 0.9902,
+      "step": 1099
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298807765538152,
+      "loss": 0.5744,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002988053758761069,
+      "loss": 1.0854,
+      "step": 1101
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000298802983831167,
+      "loss": 0.6826,
+      "step": 1102
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029880058940337067,
+      "loss": 0.4425,
+      "step": 1103
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987981925927562,
+      "loss": 1.3798,
+      "step": 1104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029879579339936197,
+      "loss": 0.1355,
+      "step": 1105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987933918232265,
+      "loss": 1.2349,
+      "step": 1106
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029879098786438806,
+      "loss": 0.5253,
+      "step": 1107
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987885815228853,
+      "loss": 0.9236,
+      "step": 1108
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029878617279875674,
+      "loss": 1.9825,
+      "step": 1109
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987837616920409,
+      "loss": 1.589,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029878134820277637,
+      "loss": 0.5732,
+      "step": 1111
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987789323310019,
+      "loss": 0.5291,
+      "step": 1112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029877651407675605,
+      "loss": 0.2494,
+      "step": 1113
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029877409344007764,
+      "loss": 0.3808,
+      "step": 1114
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987716704210054,
+      "loss": 1.0081,
+      "step": 1115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029876924501957814,
+      "loss": 0.5692,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002987668172358347,
+      "loss": 0.8389,
+      "step": 1117
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029876438706981395,
+      "loss": 0.2524,
+      "step": 1118
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029876195452155477,
+      "loss": 0.3396,
+      "step": 1119
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029875951959109617,
+      "loss": 0.0554,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00029875708227847714,
+      "loss": 0.737,
+      "step": 1121
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987546425837367,
+      "loss": 0.0987,
+      "step": 1122
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029875220050691386,
+      "loss": 0.1778,
+      "step": 1123
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987497560480478,
+      "loss": 0.1349,
+      "step": 1124
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029874730920717763,
+      "loss": 1.0875,
+      "step": 1125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029874485998434257,
+      "loss": 0.4404,
+      "step": 1126
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987424083795818,
+      "loss": 1.71,
+      "step": 1127
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029873995439293454,
+      "loss": 0.8929,
+      "step": 1128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987374980244402,
+      "loss": 1.1732,
+      "step": 1129
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029873503927413804,
+      "loss": 0.3088,
+      "step": 1130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029873257814206744,
+      "loss": 0.4324,
+      "step": 1131
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987301146282678,
+      "loss": 0.0123,
+      "step": 1132
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987276487327786,
+      "loss": 0.8436,
+      "step": 1133
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987251804556393,
+      "loss": 0.6492,
+      "step": 1134
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029872270979688947,
+      "loss": 0.4837,
+      "step": 1135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029872023675656865,
+      "loss": 0.5273,
+      "step": 1136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987177613347164,
+      "loss": 0.9841,
+      "step": 1137
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987152835313724,
+      "loss": 0.0995,
+      "step": 1138
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029871280334657635,
+      "loss": 0.346,
+      "step": 1139
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987103207803679,
+      "loss": 1.2408,
+      "step": 1140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029870783583278693,
+      "loss": 0.4761,
+      "step": 1141
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987053485038731,
+      "loss": 0.7216,
+      "step": 1142
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029870285879366625,
+      "loss": 0.7437,
+      "step": 1143
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002987003667022063,
+      "loss": 1.1406,
+      "step": 1144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986978722295332,
+      "loss": 0.4197,
+      "step": 1145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029869537537568683,
+      "loss": 0.4418,
+      "step": 1146
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986928761407071,
+      "loss": 0.558,
+      "step": 1147
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029869037452463423,
+      "loss": 0.65,
+      "step": 1148
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986878705275081,
+      "loss": 0.9091,
+      "step": 1149
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986853641493689,
+      "loss": 0.6679,
+      "step": 1150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986828553902567,
+      "loss": 0.2689,
+      "step": 1151
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029868034425021173,
+      "loss": 0.3997,
+      "step": 1152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029867783072927423,
+      "loss": 0.0093,
+      "step": 1153
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986753148274843,
+      "loss": 1.4716,
+      "step": 1154
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986727965448824,
+      "loss": 0.7709,
+      "step": 1155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029867027588150884,
+      "loss": 0.123,
+      "step": 1156
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029866775283740387,
+      "loss": 0.0037,
+      "step": 1157
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298665227412608,
+      "loss": 0.0036,
+      "step": 1158
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029866269960716153,
+      "loss": 0.3454,
+      "step": 1159
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029866016942110515,
+      "loss": 0.3882,
+      "step": 1160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029865763685447917,
+      "loss": 1.2862,
+      "step": 1161
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986551019073243,
+      "loss": 0.2353,
+      "step": 1162
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986525645796811,
+      "loss": 1.309,
+      "step": 1163
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986500248715901,
+      "loss": 0.4305,
+      "step": 1164
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029864748278309207,
+      "loss": 1.2008,
+      "step": 1165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029864493831422773,
+      "loss": 0.9319,
+      "step": 1166
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029864239146503774,
+      "loss": 0.72,
+      "step": 1167
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298639842235563,
+      "loss": 2.755,
+      "step": 1168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029863729062584417,
+      "loss": 0.3581,
+      "step": 1169
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986347366359223,
+      "loss": 0.2871,
+      "step": 1170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029863218026583813,
+      "loss": 0.8202,
+      "step": 1171
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029862962151563265,
+      "loss": 0.4296,
+      "step": 1172
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986270603853469,
+      "loss": 0.3077,
+      "step": 1173
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029862449687502175,
+      "loss": 0.2855,
+      "step": 1174
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986219309846984,
+      "loss": 0.6558,
+      "step": 1175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986193627144179,
+      "loss": 0.4507,
+      "step": 1176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029861679206422126,
+      "loss": 0.5765,
+      "step": 1177
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986142190341498,
+      "loss": 2.1127,
+      "step": 1178
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029861164362424467,
+      "loss": 0.5063,
+      "step": 1179
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029860906583454704,
+      "loss": 0.5453,
+      "step": 1180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029860648566509834,
+      "loss": 0.2578,
+      "step": 1181
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029860390311593977,
+      "loss": 1.9338,
+      "step": 1182
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002986013181871127,
+      "loss": 0.4206,
+      "step": 1183
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985987308786585,
+      "loss": 1.0649,
+      "step": 1184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985961411906187,
+      "loss": 0.8388,
+      "step": 1185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985935491230347,
+      "loss": 0.3,
+      "step": 1186
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029859095467594805,
+      "loss": 0.2407,
+      "step": 1187
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029858835784940014,
+      "loss": 0.6752,
+      "step": 1188
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985857586434328,
+      "loss": 0.4638,
+      "step": 1189
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985831570580875,
+      "loss": 0.3337,
+      "step": 1190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029858055309340593,
+      "loss": 0.6802,
+      "step": 1191
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985779467494298,
+      "loss": 0.5261,
+      "step": 1192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985753380262008,
+      "loss": 0.6721,
+      "step": 1193
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029857272692376076,
+      "loss": 0.7317,
+      "step": 1194
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029857011344215146,
+      "loss": 0.4485,
+      "step": 1195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029856749758141484,
+      "loss": 0.484,
+      "step": 1196
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985648793415926,
+      "loss": 0.4263,
+      "step": 1197
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029856225872272683,
+      "loss": 0.4201,
+      "step": 1198
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985596357248595,
+      "loss": 1.1119,
+      "step": 1199
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029855701034803257,
+      "loss": 0.8075,
+      "step": 1200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029855438259228796,
+      "loss": 0.7074,
+      "step": 1201
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029855175245766793,
+      "loss": 1.0385,
+      "step": 1202
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029854911994421456,
+      "loss": 0.2883,
+      "step": 1203
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985464850519699,
+      "loss": 0.0522,
+      "step": 1204
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985438477809763,
+      "loss": 0.3093,
+      "step": 1205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985412081312758,
+      "loss": 0.2516,
+      "step": 1206
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029853856610291085,
+      "loss": 0.1572,
+      "step": 1207
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029853592169592374,
+      "loss": 0.4271,
+      "step": 1208
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985332749103567,
+      "loss": 1.0423,
+      "step": 1209
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985306257462522,
+      "loss": 0.2668,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985279742036526,
+      "loss": 0.615,
+      "step": 1211
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029852532028260047,
+      "loss": 0.6872,
+      "step": 1212
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029852266398313823,
+      "loss": 0.9604,
+      "step": 1213
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985200053053084,
+      "loss": 0.3855,
+      "step": 1214
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985173442491536,
+      "loss": 0.5392,
+      "step": 1215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985146808147164,
+      "loss": 0.0691,
+      "step": 1216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985120150020395,
+      "loss": 0.6627,
+      "step": 1217
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029850934681116553,
+      "loss": 0.0049,
+      "step": 1218
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029850667624213724,
+      "loss": 0.2971,
+      "step": 1219
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985040032949975,
+      "loss": 0.3956,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002985013279697889,
+      "loss": 0.3858,
+      "step": 1221
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029849865026655454,
+      "loss": 0.4768,
+      "step": 1222
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984959701853371,
+      "loss": 0.8525,
+      "step": 1223
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984932877261795,
+      "loss": 0.0505,
+      "step": 1224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029849060288912485,
+      "loss": 0.4015,
+      "step": 1225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298487915674216,
+      "loss": 0.3356,
+      "step": 1226
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298485226081496,
+      "loss": 0.4455,
+      "step": 1227
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029848253411100795,
+      "loss": 0.5801,
+      "step": 1228
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298479839762795,
+      "loss": 0.3539,
+      "step": 1229
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029847714303690026,
+      "loss": 0.0018,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029847444393336686,
+      "loss": 0.7412,
+      "step": 1231
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984717424522381,
+      "loss": 0.3175,
+      "step": 1232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984690385935572,
+      "loss": 0.6519,
+      "step": 1233
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984663323573675,
+      "loss": 0.5466,
+      "step": 1234
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029846362374371223,
+      "loss": 2.6799,
+      "step": 1235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029846091275263493,
+      "loss": 0.9054,
+      "step": 1236
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029845819938417885,
+      "loss": 1.1795,
+      "step": 1237
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029845548363838757,
+      "loss": 0.3124,
+      "step": 1238
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029845276551530446,
+      "loss": 0.4187,
+      "step": 1239
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029845004501497316,
+      "loss": 0.0025,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029844732213743714,
+      "loss": 0.4403,
+      "step": 1241
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984445968827401,
+      "loss": 0.4305,
+      "step": 1242
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984418692509255,
+      "loss": 0.4682,
+      "step": 1243
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984391392420373,
+      "loss": 3.5746,
+      "step": 1244
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029843640685611896,
+      "loss": 0.2206,
+      "step": 1245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984336720932144,
+      "loss": 0.2843,
+      "step": 1246
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029843093495336725,
+      "loss": 1.8343,
+      "step": 1247
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984281954366215,
+      "loss": 0.5017,
+      "step": 1248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298425453543021,
+      "loss": 2.1938,
+      "step": 1249
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984227092726096,
+      "loss": 0.4121,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029841996262543126,
+      "loss": 1.8785,
+      "step": 1251
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029841721360152995,
+      "loss": 0.1499,
+      "step": 1252
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984144622009497,
+      "loss": 0.4938,
+      "step": 1253
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029841170842373455,
+      "loss": 0.4759,
+      "step": 1254
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029840895226992866,
+      "loss": 0.4987,
+      "step": 1255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984061937395761,
+      "loss": 1.0962,
+      "step": 1256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029840343283272114,
+      "loss": 6.4129,
+      "step": 1257
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002984006695494079,
+      "loss": 0.2109,
+      "step": 1258
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029839790388968066,
+      "loss": 0.0588,
+      "step": 1259
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983951358535836,
+      "loss": 1.0189,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983923654411613,
+      "loss": 0.4313,
+      "step": 1261
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983895926524579,
+      "loss": 0.3112,
+      "step": 1262
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983868174875179,
+      "loss": 0.4846,
+      "step": 1263
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029838403994638567,
+      "loss": 0.3745,
+      "step": 1264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983812600291058,
+      "loss": 0.9863,
+      "step": 1265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029837847773572264,
+      "loss": 0.7129,
+      "step": 1266
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029837569306628094,
+      "loss": 0.7103,
+      "step": 1267
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983729060208252,
+      "loss": 0.4657,
+      "step": 1268
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029837011659939995,
+      "loss": 0.2826,
+      "step": 1269
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029836732480205007,
+      "loss": 0.157,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983645306288201,
+      "loss": 0.0155,
+      "step": 1271
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983617340797548,
+      "loss": 0.3915,
+      "step": 1272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298358935154899,
+      "loss": 0.5472,
+      "step": 1273
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029835613385429755,
+      "loss": 0.0081,
+      "step": 1274
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029835333017799525,
+      "loss": 0.5005,
+      "step": 1275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298350524126037,
+      "loss": 0.5163,
+      "step": 1276
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029834771569846774,
+      "loss": 0.4016,
+      "step": 1277
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983449048953325,
+      "loss": 0.5705,
+      "step": 1278
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029834209171667624,
+      "loss": 0.6577,
+      "step": 1279
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298339276162544,
+      "loss": 1.4493,
+      "step": 1280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983364582329808,
+      "loss": 0.6448,
+      "step": 1281
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983336379280319,
+      "loss": 0.667,
+      "step": 1282
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983308152477425,
+      "loss": 2.2483,
+      "step": 1283
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983279901921576,
+      "loss": 0.9235,
+      "step": 1284
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029832516276132254,
+      "loss": 0.368,
+      "step": 1285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029832233295528266,
+      "loss": 0.1225,
+      "step": 1286
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029831950077408313,
+      "loss": 0.3589,
+      "step": 1287
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983166662177695,
+      "loss": 2.0767,
+      "step": 1288
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298313829286387,
+      "loss": 0.0058,
+      "step": 1289
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983109899799811,
+      "loss": 0.6639,
+      "step": 1290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029830814829859726,
+      "loss": 0.368,
+      "step": 1291
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002983053042422811,
+      "loss": 0.6515,
+      "step": 1292
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029830245781107797,
+      "loss": 0.8121,
+      "step": 1293
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982996090050336,
+      "loss": 0.4888,
+      "step": 1294
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982967578241935,
+      "loss": 0.3614,
+      "step": 1295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029829390426860345,
+      "loss": 0.3878,
+      "step": 1296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029829104833830906,
+      "loss": 0.4714,
+      "step": 1297
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982881900333561,
+      "loss": 0.3619,
+      "step": 1298
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982853293537903,
+      "loss": 0.4123,
+      "step": 1299
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029828246629965755,
+      "loss": 1.2433,
+      "step": 1300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982796008710036,
+      "loss": 0.3795,
+      "step": 1301
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029827673306787435,
+      "loss": 0.3579,
+      "step": 1302
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982738628903158,
+      "loss": 0.266,
+      "step": 1303
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982709903383738,
+      "loss": 1.061,
+      "step": 1304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982681154120945,
+      "loss": 0.3992,
+      "step": 1305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982652381115238,
+      "loss": 0.2739,
+      "step": 1306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982623584367078,
+      "loss": 0.5786,
+      "step": 1307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029825947638769274,
+      "loss": 0.0227,
+      "step": 1308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982565919645246,
+      "loss": 2.3752,
+      "step": 1309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029825370516724963,
+      "loss": 1.6689,
+      "step": 1310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982508159959141,
+      "loss": 0.6507,
+      "step": 1311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982479244505642,
+      "loss": 0.3078,
+      "step": 1312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029824503053124636,
+      "loss": 0.6572,
+      "step": 1313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982421342380068,
+      "loss": 0.8039,
+      "step": 1314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982392355708919,
+      "loss": 0.464,
+      "step": 1315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029823633452994813,
+      "loss": 0.4842,
+      "step": 1316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298233431115222,
+      "loss": 0.2844,
+      "step": 1317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029823052532675987,
+      "loss": 0.6501,
+      "step": 1318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029822761716460835,
+      "loss": 0.4487,
+      "step": 1319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029822470662881404,
+      "loss": 0.746,
+      "step": 1320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029822179371942346,
+      "loss": 1.8234,
+      "step": 1321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982188784364833,
+      "loss": 3.481,
+      "step": 1322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002982159607800403,
+      "loss": 0.5231,
+      "step": 1323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029821304075014113,
+      "loss": 0.4446,
+      "step": 1324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029821011834683253,
+      "loss": 0.4891,
+      "step": 1325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029820719357016124,
+      "loss": 0.6432,
+      "step": 1326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029820426642017427,
+      "loss": 0.7765,
+      "step": 1327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029820133689691835,
+      "loss": 0.48,
+      "step": 1328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029819840500044047,
+      "loss": 2.749,
+      "step": 1329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029819547073078753,
+      "loss": 1.1266,
+      "step": 1330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029819253408800653,
+      "loss": 0.3576,
+      "step": 1331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029818959507214453,
+      "loss": 0.2134,
+      "step": 1332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029818665368324857,
+      "loss": 0.4379,
+      "step": 1333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981837099213657,
+      "loss": 0.4789,
+      "step": 1334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981807637865431,
+      "loss": 0.4601,
+      "step": 1335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298177815278828,
+      "loss": 4.96,
+      "step": 1336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029817486439826754,
+      "loss": 0.5905,
+      "step": 1337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029817191114490905,
+      "loss": 0.341,
+      "step": 1338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981689555187998,
+      "loss": 2.6038,
+      "step": 1339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000298165997519987,
+      "loss": 0.4313,
+      "step": 1340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981630371485182,
+      "loss": 0.4402,
+      "step": 1341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029816007440444073,
+      "loss": 0.6681,
+      "step": 1342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029815710928780196,
+      "loss": 0.2891,
+      "step": 1343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029815414179864945,
+      "loss": 1.1695,
+      "step": 1344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981511719370307,
+      "loss": 0.3235,
+      "step": 1345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981481997029933,
+      "loss": 0.6936,
+      "step": 1346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029814522509658486,
+      "loss": 5.4087,
+      "step": 1347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981422481178529,
+      "loss": 0.6898,
+      "step": 1348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981392687668452,
+      "loss": 1.197,
+      "step": 1349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981362870436095,
+      "loss": 0.7382,
+      "step": 1350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981333029481934,
+      "loss": 2.9387,
+      "step": 1351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029813031648064477,
+      "loss": 0.2913,
+      "step": 1352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029812732764101146,
+      "loss": 0.083,
+      "step": 1353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029812433642934134,
+      "loss": 0.5557,
+      "step": 1354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981213428456822,
+      "loss": 0.4749,
+      "step": 1355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981183468900821,
+      "loss": 0.705,
+      "step": 1356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029811534856258895,
+      "loss": 0.6068,
+      "step": 1357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981123478632508,
+      "loss": 0.8256,
+      "step": 1358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029810934479211566,
+      "loss": 0.6578,
+      "step": 1359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981063393492317,
+      "loss": 0.6963,
+      "step": 1360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029810333153464693,
+      "loss": 0.0394,
+      "step": 1361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002981003213484096,
+      "loss": 1.1988,
+      "step": 1362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029809730879056787,
+      "loss": 0.5193,
+      "step": 1363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029809429386117004,
+      "loss": 0.0128,
+      "step": 1364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029809127656026433,
+      "loss": 1.0269,
+      "step": 1365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980882568878991,
+      "loss": 0.5834,
+      "step": 1366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980852348441227,
+      "loss": 0.7885,
+      "step": 1367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029808221042898345,
+      "loss": 0.0701,
+      "step": 1368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029807918364252987,
+      "loss": 2.0845,
+      "step": 1369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002980761544848104,
+      "loss": 0.7173,
+      "step": 1370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029807312295587355,
+      "loss": 0.4817,
+      "step": 1371
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029807008905576786,
+      "loss": 0.9277,
+      "step": 1372
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980670527845419,
+      "loss": 0.2472,
+      "step": 1373
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980640141422443,
+      "loss": 0.2713,
+      "step": 1374
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029806097312892374,
+      "loss": 0.3595,
+      "step": 1375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029805792974462895,
+      "loss": 0.4002,
+      "step": 1376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980548839894086,
+      "loss": 0.6564,
+      "step": 1377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980518358633114,
+      "loss": 0.2357,
+      "step": 1378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980487853663864,
+      "loss": 0.5183,
+      "step": 1379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980457324986822,
+      "loss": 0.5629,
+      "step": 1380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029804267726024776,
+      "loss": 0.1545,
+      "step": 1381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000298039619651132,
+      "loss": 0.2156,
+      "step": 1382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029803655967138395,
+      "loss": 1.6445,
+      "step": 1383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980334973210526,
+      "loss": 0.533,
+      "step": 1384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980304326001869,
+      "loss": 0.75,
+      "step": 1385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029802736550883606,
+      "loss": 0.1853,
+      "step": 1386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029802429604704905,
+      "loss": 1.4368,
+      "step": 1387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980212242148751,
+      "loss": 0.6644,
+      "step": 1388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980181500123634,
+      "loss": 0.5216,
+      "step": 1389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980150734395632,
+      "loss": 0.6819,
+      "step": 1390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029801199449652375,
+      "loss": 1.3306,
+      "step": 1391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002980089131832943,
+      "loss": 0.2993,
+      "step": 1392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029800582949992426,
+      "loss": 0.2151,
+      "step": 1393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029800274344646293,
+      "loss": 0.2341,
+      "step": 1394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979996550229598,
+      "loss": 3.3148,
+      "step": 1395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979965642294644,
+      "loss": 0.789,
+      "step": 1396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029799347106602606,
+      "loss": 1.2306,
+      "step": 1397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979903755326944,
+      "loss": 0.4352,
+      "step": 1398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297987277629519,
+      "loss": 0.426,
+      "step": 1399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979841773565494,
+      "loss": 0.635,
+      "step": 1400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029798107471383533,
+      "loss": 0.3757,
+      "step": 1401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979779697014264,
+      "loss": 0.2998,
+      "step": 1402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979748623193724,
+      "loss": 0.3119,
+      "step": 1403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979717525677231,
+      "loss": 0.753,
+      "step": 1404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979686404465282,
+      "loss": 0.38,
+      "step": 1405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029796552595583756,
+      "loss": 0.2268,
+      "step": 1406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029796240909570114,
+      "loss": 0.9142,
+      "step": 1407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979592898661688,
+      "loss": 0.7947,
+      "step": 1408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979561682672905,
+      "loss": 1.0531,
+      "step": 1409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979530442991162,
+      "loss": 0.8839,
+      "step": 1410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979499179616959,
+      "loss": 0.1409,
+      "step": 1411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029794678925507975,
+      "loss": 1.7604,
+      "step": 1412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979436581793178,
+      "loss": 7.3418,
+      "step": 1413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029794052473446016,
+      "loss": 0.648,
+      "step": 1414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979373889205571,
+      "loss": 0.9539,
+      "step": 1415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029793425073765874,
+      "loss": 0.3229,
+      "step": 1416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979311101858154,
+      "loss": 0.8845,
+      "step": 1417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029792796726507736,
+      "loss": 0.6991,
+      "step": 1418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979248219754949,
+      "loss": 0.2007,
+      "step": 1419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979216743171184,
+      "loss": 0.6523,
+      "step": 1420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029791852428999834,
+      "loss": 0.3098,
+      "step": 1421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979153718941851,
+      "loss": 0.3459,
+      "step": 1422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029791221712972913,
+      "loss": 0.4898,
+      "step": 1423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029790905999668103,
+      "loss": 0.6201,
+      "step": 1424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002979059004950913,
+      "loss": 0.1855,
+      "step": 1425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029790273862501056,
+      "loss": 0.2419,
+      "step": 1426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978995743864894,
+      "loss": 0.4038,
+      "step": 1427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029789640777957857,
+      "loss": 0.2493,
+      "step": 1428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029789323880432866,
+      "loss": 0.2204,
+      "step": 1429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978900674607905,
+      "loss": 0.4744,
+      "step": 1430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029788689374901493,
+      "loss": 0.0413,
+      "step": 1431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029788371766905264,
+      "loss": 0.7745,
+      "step": 1432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029788053922095455,
+      "loss": 0.5911,
+      "step": 1433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978773584047716,
+      "loss": 0.3085,
+      "step": 1434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978741752205547,
+      "loss": 0.2638,
+      "step": 1435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978709896683548,
+      "loss": 0.4287,
+      "step": 1436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978678017482229,
+      "loss": 1.8328,
+      "step": 1437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029786461146021005,
+      "loss": 0.512,
+      "step": 1438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978614188043674,
+      "loss": 0.4144,
+      "step": 1439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029785822378074607,
+      "loss": 0.4314,
+      "step": 1440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978550263893971,
+      "loss": 2.8832,
+      "step": 1441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029785182663037186,
+      "loss": 0.6054,
+      "step": 1442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978486245037215,
+      "loss": 0.2769,
+      "step": 1443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029784542000949727,
+      "loss": 0.7809,
+      "step": 1444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978422131477505,
+      "loss": 2.2076,
+      "step": 1445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029783900391853265,
+      "loss": 0.4326,
+      "step": 1446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297835792321895,
+      "loss": 0.7844,
+      "step": 1447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297832578357889,
+      "loss": 0.9925,
+      "step": 1448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978293620265661,
+      "loss": 1.0259,
+      "step": 1449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029782614332797785,
+      "loss": 0.3537,
+      "step": 1450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978229222621758,
+      "loss": 0.3302,
+      "step": 1451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029781969882921146,
+      "loss": 0.0289,
+      "step": 1452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029781647302913654,
+      "loss": 0.185,
+      "step": 1453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978132448620026,
+      "loss": 0.4229,
+      "step": 1454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978100143278614,
+      "loss": 0.408,
+      "step": 1455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978067814267646,
+      "loss": 0.7042,
+      "step": 1456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029780354615876407,
+      "loss": 1.2514,
+      "step": 1457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002978003085239116,
+      "loss": 0.2353,
+      "step": 1458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977970685222589,
+      "loss": 0.2432,
+      "step": 1459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297793826153858,
+      "loss": 0.8877,
+      "step": 1460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977905814187609,
+      "loss": 0.5,
+      "step": 1461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977873343170193,
+      "loss": 5.0793,
+      "step": 1462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029778408484868536,
+      "loss": 0.4218,
+      "step": 1463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977808330138111,
+      "loss": 0.6598,
+      "step": 1464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029777757881244865,
+      "loss": 0.5762,
+      "step": 1465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029777432224465,
+      "loss": 0.9055,
+      "step": 1466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029777106331046744,
+      "loss": 0.954,
+      "step": 1467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029776780200995297,
+      "loss": 0.5355,
+      "step": 1468
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029776453834315895,
+      "loss": 1.0463,
+      "step": 1469
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977612723101376,
+      "loss": 0.4031,
+      "step": 1470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977580039109413,
+      "loss": 0.7447,
+      "step": 1471
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029775473314562224,
+      "loss": 1.9116,
+      "step": 1472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029775146001423294,
+      "loss": 1.4164,
+      "step": 1473
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977481845168257,
+      "loss": 0.3033,
+      "step": 1474
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029774490665345307,
+      "loss": 1.2505,
+      "step": 1475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029774162642416744,
+      "loss": 0.3085,
+      "step": 1476
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977383438290215,
+      "loss": 0.5353,
+      "step": 1477
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977350588680676,
+      "loss": 0.4031,
+      "step": 1478
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029773177154135844,
+      "loss": 1.3333,
+      "step": 1479
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029772848184894674,
+      "loss": 0.4079,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977251897908851,
+      "loss": 0.0746,
+      "step": 1481
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029772189536722625,
+      "loss": 0.2525,
+      "step": 1482
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977185985780229,
+      "loss": 0.2004,
+      "step": 1483
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977152994233279,
+      "loss": 0.7164,
+      "step": 1484
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977119979031941,
+      "loss": 0.1899,
+      "step": 1485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977086940176743,
+      "loss": 2.0803,
+      "step": 1486
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002977053877668215,
+      "loss": 0.61,
+      "step": 1487
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029770207915068856,
+      "loss": 0.3833,
+      "step": 1488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029769876816932854,
+      "loss": 0.9496,
+      "step": 1489
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029769545482279434,
+      "loss": 0.5668,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976921391111392,
+      "loss": 0.4094,
+      "step": 1491
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029768882103441596,
+      "loss": 0.6348,
+      "step": 1492
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297685500592678,
+      "loss": 0.9134,
+      "step": 1493
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029768217778597834,
+      "loss": 0.4273,
+      "step": 1494
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976788526143703,
+      "loss": 0.3059,
+      "step": 1495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297675525077907,
+      "loss": 0.356,
+      "step": 1496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029767219517664185,
+      "loss": 1.6428,
+      "step": 1497
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976688629106281,
+      "loss": 0.6173,
+      "step": 1498
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976655282799191,
+      "loss": 0.4395,
+      "step": 1499
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976621912845683,
+      "loss": 0.6387,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976588519246291,
+      "loss": 0.6686,
+      "step": 1501
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029765551020015504,
+      "loss": 0.8186,
+      "step": 1502
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976521661111996,
+      "loss": 0.6635,
+      "step": 1503
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029764881965781625,
+      "loss": 0.5171,
+      "step": 1504
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976454708400587,
+      "loss": 1.035,
+      "step": 1505
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029764211965798044,
+      "loss": 0.4761,
+      "step": 1506
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976387661116353,
+      "loss": 0.4571,
+      "step": 1507
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029763541020107686,
+      "loss": 0.3725,
+      "step": 1508
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029763205192635887,
+      "loss": 0.9942,
+      "step": 1509
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029762869128753517,
+      "loss": 0.8251,
+      "step": 1510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029762532828465953,
+      "loss": 0.2106,
+      "step": 1511
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976219629177858,
+      "loss": 0.5783,
+      "step": 1512
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976185951869679,
+      "loss": 1.5884,
+      "step": 1513
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976152250922597,
+      "loss": 0.5329,
+      "step": 1514
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976118526337153,
+      "loss": 2.1877,
+      "step": 1515
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029760847781138853,
+      "loss": 2.5776,
+      "step": 1516
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976051006253335,
+      "loss": 0.4097,
+      "step": 1517
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002976017210756044,
+      "loss": 0.6592,
+      "step": 1518
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975983391622552,
+      "loss": 0.401,
+      "step": 1519
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975949548853401,
+      "loss": 0.611,
+      "step": 1520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975915682449133,
+      "loss": 0.2612,
+      "step": 1521
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029758817924102913,
+      "loss": 0.9021,
+      "step": 1522
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029758478787374166,
+      "loss": 0.7503,
+      "step": 1523
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029758139414310536,
+      "loss": 0.5166,
+      "step": 1524
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975779980491745,
+      "loss": 0.9672,
+      "step": 1525
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029757459959200354,
+      "loss": 0.7019,
+      "step": 1526
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029757119877164677,
+      "loss": 0.3943,
+      "step": 1527
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029756779558815877,
+      "loss": 0.4275,
+      "step": 1528
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029756439004159397,
+      "loss": 1.1522,
+      "step": 1529
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975609821320069,
+      "loss": 0.5988,
+      "step": 1530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975575718594522,
+      "loss": 0.3106,
+      "step": 1531
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975541592239845,
+      "loss": 0.2668,
+      "step": 1532
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029755074422565825,
+      "loss": 0.842,
+      "step": 1533
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029754732686452834,
+      "loss": 0.5417,
+      "step": 1534
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975439071406495,
+      "loss": 0.7922,
+      "step": 1535
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975404850540763,
+      "loss": 0.84,
+      "step": 1536
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029753706060486375,
+      "loss": 0.2458,
+      "step": 1537
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029753363379306653,
+      "loss": 1.76,
+      "step": 1538
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975302046187396,
+      "loss": 0.355,
+      "step": 1539
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029752677308193787,
+      "loss": 0.9301,
+      "step": 1540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029752333918271626,
+      "loss": 0.6381,
+      "step": 1541
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975199029211298,
+      "loss": 0.6721,
+      "step": 1542
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002975164642972335,
+      "loss": 0.4546,
+      "step": 1543
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029751302331108235,
+      "loss": 0.3785,
+      "step": 1544
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029750957996273154,
+      "loss": 0.3738,
+      "step": 1545
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029750613425223623,
+      "loss": 1.3231,
+      "step": 1546
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029750268617965156,
+      "loss": 0.7197,
+      "step": 1547
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029749923574503266,
+      "loss": 0.6787,
+      "step": 1548
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297495782948435,
+      "loss": 0.3895,
+      "step": 1549
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974923277899136,
+      "loss": 0.5227,
+      "step": 1550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297488870269524,
+      "loss": 0.1752,
+      "step": 1551
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029748541038732146,
+      "loss": 1.0524,
+      "step": 1552
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974819481433614,
+      "loss": 0.7697,
+      "step": 1553
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974784835376993,
+      "loss": 0.8141,
+      "step": 1554
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974750165703907,
+      "loss": 0.473,
+      "step": 1555
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029747154724149094,
+      "loss": 0.8152,
+      "step": 1556
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974680755510557,
+      "loss": 0.1941,
+      "step": 1557
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974646014991406,
+      "loss": 0.6677,
+      "step": 1558
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974611250858011,
+      "loss": 0.5981,
+      "step": 1559
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029745764631109307,
+      "loss": 0.5608,
+      "step": 1560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974541651750722,
+      "loss": 1.0788,
+      "step": 1561
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029745068167779407,
+      "loss": 0.8509,
+      "step": 1562
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974471958193146,
+      "loss": 0.6351,
+      "step": 1563
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029744370759968955,
+      "loss": 0.2036,
+      "step": 1564
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029744021701897484,
+      "loss": 0.6395,
+      "step": 1565
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029743672407722634,
+      "loss": 1.5311,
+      "step": 1566
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029743322877449994,
+      "loss": 0.4036,
+      "step": 1567
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029742973111085164,
+      "loss": 0.7911,
+      "step": 1568
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974262310863375,
+      "loss": 0.0858,
+      "step": 1569
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974227287010135,
+      "loss": 0.2924,
+      "step": 1570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974192239549358,
+      "loss": 1.1187,
+      "step": 1571
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974157168481604,
+      "loss": 0.4094,
+      "step": 1572
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029741220738074354,
+      "loss": 0.627,
+      "step": 1573
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029740869555274147,
+      "loss": 0.7465,
+      "step": 1574
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974051813642103,
+      "loss": 0.6259,
+      "step": 1575
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002974016648152064,
+      "loss": 0.6523,
+      "step": 1576
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973981459057861,
+      "loss": 0.8543,
+      "step": 1577
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029739462463600566,
+      "loss": 1.2156,
+      "step": 1578
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029739110100592155,
+      "loss": 3.5162,
+      "step": 1579
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029738757501559013,
+      "loss": 0.7307,
+      "step": 1580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973840466650679,
+      "loss": 0.6489,
+      "step": 1581
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973805159544114,
+      "loss": 1.438,
+      "step": 1582
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973769828836771,
+      "loss": 0.6054,
+      "step": 1583
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029737344745292153,
+      "loss": 1.0054,
+      "step": 1584
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029736990966220143,
+      "loss": 0.689,
+      "step": 1585
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973663695115734,
+      "loss": 0.1234,
+      "step": 1586
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029736282700109414,
+      "loss": 0.7667,
+      "step": 1587
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029735928213082036,
+      "loss": 0.6197,
+      "step": 1588
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029735573490080884,
+      "loss": 0.8113,
+      "step": 1589
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029735218531111635,
+      "loss": 0.5565,
+      "step": 1590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029734863336179976,
+      "loss": 0.5788,
+      "step": 1591
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297345079052916,
+      "loss": 0.5037,
+      "step": 1592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029734152238452183,
+      "loss": 0.2083,
+      "step": 1593
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973379633566744,
+      "loss": 3.3983,
+      "step": 1594
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029733440196943057,
+      "loss": 0.8043,
+      "step": 1595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973308382228474,
+      "loss": 0.6331,
+      "step": 1596
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297327272116982,
+      "loss": 1.8203,
+      "step": 1597
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973237036518914,
+      "loss": 1.9956,
+      "step": 1598
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029732013282763285,
+      "loss": 0.7871,
+      "step": 1599
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973165596442634,
+      "loss": 0.1454,
+      "step": 1600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029731298410184037,
+      "loss": 0.3272,
+      "step": 1601
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000297309406200421,
+      "loss": 0.3979,
+      "step": 1602
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973058259400626,
+      "loss": 0.3691,
+      "step": 1603
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002973022433208224,
+      "loss": 0.1926,
+      "step": 1604
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972986583427579,
+      "loss": 0.9778,
+      "step": 1605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972950710059264,
+      "loss": 0.6068,
+      "step": 1606
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972914813103854,
+      "loss": 0.3815,
+      "step": 1607
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972878892561924,
+      "loss": 0.7853,
+      "step": 1608
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029728429484340494,
+      "loss": 0.7832,
+      "step": 1609
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972806980720805,
+      "loss": 0.3517,
+      "step": 1610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972770989422767,
+      "loss": 0.7354,
+      "step": 1611
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029727349745405123,
+      "loss": 0.2823,
+      "step": 1612
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029726989360746175,
+      "loss": 0.911,
+      "step": 1613
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029726628740256585,
+      "loss": 0.629,
+      "step": 1614
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972626788394215,
+      "loss": 0.288,
+      "step": 1615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972590679180862,
+      "loss": 0.3261,
+      "step": 1616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00029725545463861805,
+      "loss": 1.3691,
+      "step": 1617
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972518390010748,
+      "loss": 0.5014,
+      "step": 1618
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972482210055143,
+      "loss": 0.3119,
+      "step": 1619
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002972446006519945,
+      "loss": 0.4627,
+      "step": 1620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029724097794057346,
+      "loss": 1.1782,
+      "step": 1621
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002972373528713091,
+      "loss": 0.7433,
+      "step": 1622
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002972337254442595,
+      "loss": 0.7079,
+      "step": 1623
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029723009565948283,
+      "loss": 0.4262,
+      "step": 1624
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002972264635170371,
+      "loss": 0.3033,
+      "step": 1625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002972228290169805,
+      "loss": 0.4089,
+      "step": 1626
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029721919215937125,
+      "loss": 0.3573,
+      "step": 1627
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002972155529442675,
+      "loss": 0.739,
+      "step": 1628
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029721191137172773,
+      "loss": 0.5236,
+      "step": 1629
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029720826744181005,
+      "loss": 0.6916,
+      "step": 1630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029720462115457295,
+      "loss": 1.0784,
+      "step": 1631
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029720097251007476,
+      "loss": 1.0125,
+      "step": 1632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971973215083739,
+      "loss": 0.5124,
+      "step": 1633
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971936681495288,
+      "loss": 0.9443,
+      "step": 1634
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971900124335981,
+      "loss": 0.6616,
+      "step": 1635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029718635436064023,
+      "loss": 0.3269,
+      "step": 1636
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971826939307138,
+      "loss": 0.8858,
+      "step": 1637
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029717903114387736,
+      "loss": 0.964,
+      "step": 1638
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029717536600018963,
+      "loss": 0.3771,
+      "step": 1639
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971716984997093,
+      "loss": 1.756,
+      "step": 1640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971680286424951,
+      "loss": 4.594,
+      "step": 1641
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971643564286058,
+      "loss": 0.1376,
+      "step": 1642
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971606818581002,
+      "loss": 0.2599,
+      "step": 1643
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971570049310371,
+      "loss": 1.1518,
+      "step": 1644
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971533256474755,
+      "loss": 0.8439,
+      "step": 1645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029714964400747416,
+      "loss": 0.3741,
+      "step": 1646
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971459600110921,
+      "loss": 0.4822,
+      "step": 1647
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971422736583884,
+      "loss": 1.0724,
+      "step": 1648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000297138584949422,
+      "loss": 0.4972,
+      "step": 1649
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029713489388425194,
+      "loss": 0.713,
+      "step": 1650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971312004629374,
+      "loss": 2.7292,
+      "step": 1651
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971275046855375,
+      "loss": 0.6436,
+      "step": 1652
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971238065521114,
+      "loss": 0.2636,
+      "step": 1653
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029712010606271834,
+      "loss": 0.7403,
+      "step": 1654
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029711640321741756,
+      "loss": 0.7713,
+      "step": 1655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971126980162684,
+      "loss": 0.6116,
+      "step": 1656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002971089904593302,
+      "loss": 0.94,
+      "step": 1657
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029710528054666227,
+      "loss": 0.919,
+      "step": 1658
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029710156827832406,
+      "loss": 1.0245,
+      "step": 1659
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029709785365437496,
+      "loss": 0.6195,
+      "step": 1660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970941366748745,
+      "loss": 0.2564,
+      "step": 1661
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029709041733988223,
+      "loss": 0.5806,
+      "step": 1662
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970866956494577,
+      "loss": 0.3976,
+      "step": 1663
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029708297160366043,
+      "loss": 0.1976,
+      "step": 1664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029707924520255017,
+      "loss": 0.9875,
+      "step": 1665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970755164461865,
+      "loss": 0.3887,
+      "step": 1666
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970717853346292,
+      "loss": 0.5115,
+      "step": 1667
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970680518679379,
+      "loss": 0.4304,
+      "step": 1668
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970643160461725,
+      "loss": 0.4799,
+      "step": 1669
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970605778693928,
+      "loss": 0.4836,
+      "step": 1670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029705683733765865,
+      "loss": 0.3456,
+      "step": 1671
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029705309445103,
+      "loss": 0.329,
+      "step": 1672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029704934920956667,
+      "loss": 0.4008,
+      "step": 1673
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970456016133287,
+      "loss": 0.7288,
+      "step": 1674
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029704185166237615,
+      "loss": 0.4835,
+      "step": 1675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000297038099356769,
+      "loss": 0.6388,
+      "step": 1676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029703434469656735,
+      "loss": 0.6555,
+      "step": 1677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029703058768183136,
+      "loss": 0.8142,
+      "step": 1678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029702682831262117,
+      "loss": 0.2589,
+      "step": 1679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029702306658899695,
+      "loss": 0.3458,
+      "step": 1680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029701930251101904,
+      "loss": 0.8724,
+      "step": 1681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970155360787476,
+      "loss": 0.6752,
+      "step": 1682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000297011767292243,
+      "loss": 0.158,
+      "step": 1683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029700799615156554,
+      "loss": 0.6774,
+      "step": 1684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970042226567757,
+      "loss": 0.1577,
+      "step": 1685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002970004468079338,
+      "loss": 1.2669,
+      "step": 1686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029699666860510043,
+      "loss": 0.4724,
+      "step": 1687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296992888048336,
+      "loss": 0.5318,
+      "step": 1688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029698910513770106,
+      "loss": 0.5996,
+      "step": 1689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029698531987325624,
+      "loss": 0.5217,
+      "step": 1690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029698153225506205,
+      "loss": 0.3173,
+      "step": 1691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029697774228317923,
+      "loss": 0.3473,
+      "step": 1692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969739499576685,
+      "loss": 2.8165,
+      "step": 1693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969701552785905,
+      "loss": 0.8546,
+      "step": 1694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296966358246006,
+      "loss": 0.6257,
+      "step": 1695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969625588599759,
+      "loss": 0.7742,
+      "step": 1696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029695875712056096,
+      "loss": 0.2028,
+      "step": 1697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029695495302782207,
+      "loss": 0.0211,
+      "step": 1698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029695114658182014,
+      "loss": 1.6398,
+      "step": 1699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029694733778261616,
+      "loss": 0.8282,
+      "step": 1700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969435266302711,
+      "loss": 0.7786,
+      "step": 1701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296939713124846,
+      "loss": 0.8961,
+      "step": 1702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029693589726640193,
+      "loss": 0.1067,
+      "step": 1703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029693207905499993,
+      "loss": 0.6326,
+      "step": 1704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029692825849070124,
+      "loss": 0.5676,
+      "step": 1705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296924435573567,
+      "loss": 0.9614,
+      "step": 1706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969206103036584,
+      "loss": 0.3641,
+      "step": 1707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969167826810368,
+      "loss": 0.7374,
+      "step": 1708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969129527057633,
+      "loss": 0.2661,
+      "step": 1709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969091203778994,
+      "loss": 0.4871,
+      "step": 1710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969052856975065,
+      "loss": 0.3672,
+      "step": 1711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002969014486646458,
+      "loss": 0.6187,
+      "step": 1712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968976092793789,
+      "loss": 0.2748,
+      "step": 1713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968937675417673,
+      "loss": 2.0011,
+      "step": 1714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968899234518724,
+      "loss": 0.7257,
+      "step": 1715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968860770097559,
+      "loss": 0.0846,
+      "step": 1716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968822282154793,
+      "loss": 1.7246,
+      "step": 1717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968783770691042,
+      "loss": 0.0947,
+      "step": 1718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029687452357069237,
+      "loss": 0.216,
+      "step": 1719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968706677203055,
+      "loss": 0.3807,
+      "step": 1720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029686680951800527,
+      "loss": 0.3842,
+      "step": 1721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968629489638535,
+      "loss": 0.1579,
+      "step": 1722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029685908605791206,
+      "loss": 0.9494,
+      "step": 1723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968552208002427,
+      "loss": 0.444,
+      "step": 1724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968513531909074,
+      "loss": 0.7181,
+      "step": 1725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029684748322996807,
+      "loss": 0.7961,
+      "step": 1726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029684361091748666,
+      "loss": 3.5834,
+      "step": 1727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029683973625352524,
+      "loss": 0.4605,
+      "step": 1728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968358592381458,
+      "loss": 0.4479,
+      "step": 1729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029683197987141047,
+      "loss": 1.0883,
+      "step": 1730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968280981533813,
+      "loss": 0.3379,
+      "step": 1731
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968242140841205,
+      "loss": 1.1214,
+      "step": 1732
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968203276636903,
+      "loss": 3.132,
+      "step": 1733
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002968164388921529,
+      "loss": 0.639,
+      "step": 1734
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029681254776957056,
+      "loss": 2.1979,
+      "step": 1735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029680865429600564,
+      "loss": 0.1169,
+      "step": 1736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029680475847152036,
+      "loss": 0.5116,
+      "step": 1737
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029680086029617724,
+      "loss": 0.627,
+      "step": 1738
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029679695977003867,
+      "loss": 0.4306,
+      "step": 1739
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967930568931671,
+      "loss": 0.5763,
+      "step": 1740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029678915166562504,
+      "loss": 0.2895,
+      "step": 1741
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296785244087475,
+      "loss": 0.4716,
+      "step": 1742
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029678133415877956,
+      "loss": 0.4339,
+      "step": 1743
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029677742187960137,
+      "loss": 0.4464,
+      "step": 1744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029677350725000303,
+      "loss": 0.7924,
+      "step": 1745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029676959027004725,
+      "loss": 0.2881,
+      "step": 1746
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029676567093979677,
+      "loss": 0.6795,
+      "step": 1747
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967617492593143,
+      "loss": 1.5929,
+      "step": 1748
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967578252286627,
+      "loss": 0.8956,
+      "step": 1749
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029675389884790477,
+      "loss": 0.02,
+      "step": 1750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029674997011710336,
+      "loss": 0.5965,
+      "step": 1751
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967460390363215,
+      "loss": 1.0091,
+      "step": 1752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029674210560562197,
+      "loss": 0.596,
+      "step": 1753
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967381698250679,
+      "loss": 0.4292,
+      "step": 1754
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967342316947223,
+      "loss": 0.574,
+      "step": 1755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029673029121464814,
+      "loss": 0.3737,
+      "step": 1756
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029672634838490863,
+      "loss": 0.1366,
+      "step": 1757
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029672240320556683,
+      "loss": 0.7354,
+      "step": 1758
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967184556766859,
+      "loss": 0.7025,
+      "step": 1759
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967145057983291,
+      "loss": 0.998,
+      "step": 1760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967105535705597,
+      "loss": 0.6016,
+      "step": 1761
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029670659899344104,
+      "loss": 0.2132,
+      "step": 1762
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002967026420670363,
+      "loss": 0.5612,
+      "step": 1763
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966986827914089,
+      "loss": 0.2157,
+      "step": 1764
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029669472116662226,
+      "loss": 1.6954,
+      "step": 1765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029669075719273983,
+      "loss": 0.5412,
+      "step": 1766
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029668679086982514,
+      "loss": 1.0481,
+      "step": 1767
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029668282219794156,
+      "loss": 0.1535,
+      "step": 1768
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966788511771528,
+      "loss": 0.297,
+      "step": 1769
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966748778075223,
+      "loss": 0.495,
+      "step": 1770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029667090208911377,
+      "loss": 0.4673,
+      "step": 1771
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966669240219909,
+      "loss": 0.3394,
+      "step": 1772
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966629436062173,
+      "loss": 0.2136,
+      "step": 1773
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966589608418568,
+      "loss": 0.8448,
+      "step": 1774
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029665497572897313,
+      "loss": 0.4301,
+      "step": 1775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029665098826763014,
+      "loss": 1.6576,
+      "step": 1776
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966469984578917,
+      "loss": 0.4021,
+      "step": 1777
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029664300629982164,
+      "loss": 0.3263,
+      "step": 1778
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966390117934839,
+      "loss": 0.2508,
+      "step": 1779
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029663501493894245,
+      "loss": 0.8907,
+      "step": 1780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966310157362613,
+      "loss": 0.0378,
+      "step": 1781
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029662701418550454,
+      "loss": 0.5724,
+      "step": 1782
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002966230102867362,
+      "loss": 0.6945,
+      "step": 1783
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029661900404002035,
+      "loss": 2.6513,
+      "step": 1784
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029661499544542125,
+      "loss": 0.2792,
+      "step": 1785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296610984503003,
+      "loss": 0.3918,
+      "step": 1786
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029660697121282987,
+      "loss": 1.9923,
+      "step": 1787
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029660295557496614,
+      "loss": 0.0208,
+      "step": 1788
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965989375894761,
+      "loss": 0.0119,
+      "step": 1789
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029659491725642404,
+      "loss": 0.3145,
+      "step": 1790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965908945758745,
+      "loss": 0.0689,
+      "step": 1791
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029658686954789164,
+      "loss": 0.5381,
+      "step": 1792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029658284217254013,
+      "loss": 0.4148,
+      "step": 1793
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965788124498844,
+      "loss": 0.6205,
+      "step": 1794
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029657478037998895,
+      "loss": 0.2755,
+      "step": 1795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965707459629184,
+      "loss": 0.9325,
+      "step": 1796
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029656670919873734,
+      "loss": 0.3682,
+      "step": 1797
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965626700875103,
+      "loss": 2.6935,
+      "step": 1798
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029655862862930214,
+      "loss": 0.7072,
+      "step": 1799
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965545848241775,
+      "loss": 1.4559,
+      "step": 1800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965505386722011,
+      "loss": 0.6126,
+      "step": 1801
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965464901734378,
+      "loss": 0.2907,
+      "step": 1802
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965424393279524,
+      "loss": 0.637,
+      "step": 1803
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029653838613580974,
+      "loss": 0.0733,
+      "step": 1804
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965343305970747,
+      "loss": 0.4931,
+      "step": 1805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029653027271181237,
+      "loss": 0.2105,
+      "step": 1806
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029652621248008753,
+      "loss": 0.5039,
+      "step": 1807
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029652214990196537,
+      "loss": 0.6053,
+      "step": 1808
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965180849775109,
+      "loss": 1.0178,
+      "step": 1809
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029651401770678916,
+      "loss": 0.745,
+      "step": 1810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965099480898653,
+      "loss": 0.0141,
+      "step": 1811
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002965058761268045,
+      "loss": 1.5108,
+      "step": 1812
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029650180181767195,
+      "loss": 0.7554,
+      "step": 1813
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029649772516253294,
+      "loss": 0.2272,
+      "step": 1814
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029649364616145274,
+      "loss": 0.9214,
+      "step": 1815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029648956481449665,
+      "loss": 0.7316,
+      "step": 1816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029648548112173,
+      "loss": 0.469,
+      "step": 1817
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964813950832182,
+      "loss": 0.1973,
+      "step": 1818
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029647730669902676,
+      "loss": 0.5623,
+      "step": 1819
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964732159692211,
+      "loss": 0.7657,
+      "step": 1820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029646912289386663,
+      "loss": 0.1591,
+      "step": 1821
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000296465027473029,
+      "loss": 0.6613,
+      "step": 1822
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964609297067738,
+      "loss": 0.5821,
+      "step": 1823
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964568295951666,
+      "loss": 0.0904,
+      "step": 1824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964527271382731,
+      "loss": 0.4066,
+      "step": 1825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964486223361589,
+      "loss": 0.2005,
+      "step": 1826
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029644451518888985,
+      "loss": 0.9592,
+      "step": 1827
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964404056965317,
+      "loss": 0.3146,
+      "step": 1828
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029643629385915014,
+      "loss": 0.7746,
+      "step": 1829
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964321796768112,
+      "loss": 1.378,
+      "step": 1830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029642806314958056,
+      "loss": 0.2775,
+      "step": 1831
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029642394427752437,
+      "loss": 1.0065,
+      "step": 1832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029641982306070837,
+      "loss": 0.3598,
+      "step": 1833
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964156994991987,
+      "loss": 1.6378,
+      "step": 1834
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029641157359306135,
+      "loss": 0.1618,
+      "step": 1835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964074453423623,
+      "loss": 0.0622,
+      "step": 1836
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002964033147471678,
+      "loss": 0.5402,
+      "step": 1837
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029639918180754393,
+      "loss": 0.5977,
+      "step": 1838
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963950465235569,
+      "loss": 0.2563,
+      "step": 1839
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963909088952729,
+      "loss": 0.3424,
+      "step": 1840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963867689227582,
+      "loss": 0.9908,
+      "step": 1841
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029638262660607907,
+      "loss": 0.784,
+      "step": 1842
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029637848194530184,
+      "loss": 0.2409,
+      "step": 1843
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029637433494049295,
+      "loss": 0.6625,
+      "step": 1844
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029637018559171876,
+      "loss": 0.2621,
+      "step": 1845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963660338990457,
+      "loss": 0.7249,
+      "step": 1846
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963618798625403,
+      "loss": 0.6126,
+      "step": 1847
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029635772348226896,
+      "loss": 0.4385,
+      "step": 1848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029635356475829846,
+      "loss": 0.0089,
+      "step": 1849
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963494036906952,
+      "loss": 0.6862,
+      "step": 1850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029634524027952587,
+      "loss": 0.3245,
+      "step": 1851
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963410745248572,
+      "loss": 0.8761,
+      "step": 1852
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029633690642675583,
+      "loss": 0.3714,
+      "step": 1853
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029633273598528846,
+      "loss": 0.5311,
+      "step": 1854
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029632856320052197,
+      "loss": 0.3685,
+      "step": 1855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029632438807252316,
+      "loss": 0.6596,
+      "step": 1856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963202106013589,
+      "loss": 0.3706,
+      "step": 1857
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029631603078709604,
+      "loss": 0.2849,
+      "step": 1858
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002963118486298015,
+      "loss": 0.6756,
+      "step": 1859
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029630766412954237,
+      "loss": 2.4017,
+      "step": 1860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029630347728638553,
+      "loss": 0.682,
+      "step": 1861
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962992881003981,
+      "loss": 0.7157,
+      "step": 1862
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029629509657164713,
+      "loss": 0.108,
+      "step": 1863
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029629090270019974,
+      "loss": 0.8233,
+      "step": 1864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962867064861231,
+      "loss": 0.8316,
+      "step": 1865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029628250792948437,
+      "loss": 0.3882,
+      "step": 1866
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962783070303508,
+      "loss": 3.0997,
+      "step": 1867
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00029627410378878975,
+      "loss": 0.495,
+      "step": 1868
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002962698982048684,
+      "loss": 0.7761,
+      "step": 1869
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002962656902786542,
+      "loss": 0.1809,
+      "step": 1870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029626148001021444,
+      "loss": 0.5638,
+      "step": 1871
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002962572673996166,
+      "loss": 0.3015,
+      "step": 1872
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029625305244692805,
+      "loss": 0.564,
+      "step": 1873
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029624883515221647,
+      "loss": 0.6404,
+      "step": 1874
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029624461551554926,
+      "loss": 0.6773,
+      "step": 1875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000296240393536994,
+      "loss": 0.7129,
+      "step": 1876
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029623616921661827,
+      "loss": 0.4507,
+      "step": 1877
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002962319425544898,
+      "loss": 0.6422,
+      "step": 1878
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029622771355067626,
+      "loss": 0.9085,
+      "step": 1879
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002962234822052453,
+      "loss": 0.4543,
+      "step": 1880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029621924851826474,
+      "loss": 0.0061,
+      "step": 1881
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029621501248980233,
+      "loss": 0.1393,
+      "step": 1882
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000296210774119926,
+      "loss": 0.8903,
+      "step": 1883
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002962065334087035,
+      "loss": 0.7297,
+      "step": 1884
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002962022903562028,
+      "loss": 0.4434,
+      "step": 1885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029619804496249186,
+      "loss": 1.4093,
+      "step": 1886
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029619379722763866,
+      "loss": 0.8824,
+      "step": 1887
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002961895471517112,
+      "loss": 0.0887,
+      "step": 1888
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029618529473477746,
+      "loss": 1.1378,
+      "step": 1889
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002961810399769057,
+      "loss": 1.5392,
+      "step": 1890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002961767828781639,
+      "loss": 1.2366,
+      "step": 1891
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002961725234386204,
+      "loss": 0.1585,
+      "step": 1892
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029616826165834324,
+      "loss": 0.267,
+      "step": 1893
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002961639975374008,
+      "loss": 0.6495,
+      "step": 1894
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029615973107586125,
+      "loss": 0.6995,
+      "step": 1895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029615546227379296,
+      "loss": 0.629,
+      "step": 1896
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002961511911312643,
+      "loss": 0.3995,
+      "step": 1897
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029614691764834366,
+      "loss": 0.5558,
+      "step": 1898
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029614264182509947,
+      "loss": 0.1971,
+      "step": 1899
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029613836366160017,
+      "loss": 0.8693,
+      "step": 1900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029613408315791435,
+      "loss": 0.2653,
+      "step": 1901
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029612980031411046,
+      "loss": 0.797,
+      "step": 1902
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029612551513025714,
+      "loss": 0.6717,
+      "step": 1903
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029612122760642307,
+      "loss": 0.7522,
+      "step": 1904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029611693774267677,
+      "loss": 0.6832,
+      "step": 1905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000296112645539087,
+      "loss": 0.4061,
+      "step": 1906
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029610835099572245,
+      "loss": 0.2256,
+      "step": 1907
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029610405411265207,
+      "loss": 0.6422,
+      "step": 1908
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960997548899444,
+      "loss": 0.4659,
+      "step": 1909
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960954533276685,
+      "loss": 0.4552,
+      "step": 1910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029609114942589316,
+      "loss": 1.0883,
+      "step": 1911
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960868431846873,
+      "loss": 0.3115,
+      "step": 1912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960825346041199,
+      "loss": 0.4214,
+      "step": 1913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029607822368425994,
+      "loss": 1.0527,
+      "step": 1914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960739104251765,
+      "loss": 0.7575,
+      "step": 1915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029606959482693856,
+      "loss": 0.5666,
+      "step": 1916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029606527688961527,
+      "loss": 0.1651,
+      "step": 1917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960609566132758,
+      "loss": 0.3545,
+      "step": 1918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960566339979893,
+      "loss": 3.0564,
+      "step": 1919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960523090438251,
+      "loss": 0.3039,
+      "step": 1920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960479817508523,
+      "loss": 0.6919,
+      "step": 1921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960436521191402,
+      "loss": 0.5549,
+      "step": 1922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029603932014875826,
+      "loss": 0.7224,
+      "step": 1923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960349858397758,
+      "loss": 0.1662,
+      "step": 1924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029603064919226215,
+      "loss": 0.5402,
+      "step": 1925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029602631020628683,
+      "loss": 0.7327,
+      "step": 1926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960219688819193,
+      "loss": 1.1573,
+      "step": 1927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002960176252192291,
+      "loss": 1.5393,
+      "step": 1928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029601327921828583,
+      "loss": 0.839,
+      "step": 1929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029600893087915897,
+      "loss": 3.5222,
+      "step": 1930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029600458020191824,
+      "loss": 0.1391,
+      "step": 1931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029600022718663326,
+      "loss": 0.6844,
+      "step": 1932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029599587183337374,
+      "loss": 0.7908,
+      "step": 1933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029599151414220944,
+      "loss": 0.4733,
+      "step": 1934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959871541132102,
+      "loss": 0.1384,
+      "step": 1935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959827917464457,
+      "loss": 0.4375,
+      "step": 1936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295978427041986,
+      "loss": 0.4951,
+      "step": 1937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959740599999008,
+      "loss": 0.5396,
+      "step": 1938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959696906202601,
+      "loss": 1.1492,
+      "step": 1939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029596531890313386,
+      "loss": 0.6476,
+      "step": 1940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959609448485921,
+      "loss": 0.5582,
+      "step": 1941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959565684567049,
+      "loss": 0.5573,
+      "step": 1942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959521897275423,
+      "loss": 0.2748,
+      "step": 1943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029594780866117444,
+      "loss": 0.4142,
+      "step": 1944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029594342525767137,
+      "loss": 0.3269,
+      "step": 1945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029593903951710346,
+      "loss": 1.1907,
+      "step": 1946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029593465143954085,
+      "loss": 0.1428,
+      "step": 1947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959302610250538,
+      "loss": 0.5581,
+      "step": 1948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029592586827371264,
+      "loss": 0.3081,
+      "step": 1949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959214731855877,
+      "loss": 1.2754,
+      "step": 1950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959170757607493,
+      "loss": 0.6728,
+      "step": 1951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959126759992681,
+      "loss": 0.5899,
+      "step": 1952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002959082739012142,
+      "loss": 0.6189,
+      "step": 1953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029590386946665837,
+      "loss": 0.6574,
+      "step": 1954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295899462695671,
+      "loss": 0.3284,
+      "step": 1955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958950535883227,
+      "loss": 1.4612,
+      "step": 1956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029589064214468404,
+      "loss": 0.5912,
+      "step": 1957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958862283648258,
+      "loss": 0.2064,
+      "step": 1958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029588181224881845,
+      "loss": 0.7779,
+      "step": 1959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029587739379673285,
+      "loss": 0.5497,
+      "step": 1960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958729730086398,
+      "loss": 0.5168,
+      "step": 1961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029586854988460987,
+      "loss": 0.8094,
+      "step": 1962
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029586412442471414,
+      "loss": 0.8764,
+      "step": 1963
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958596966290233,
+      "loss": 0.3521,
+      "step": 1964
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029585526649760836,
+      "loss": 0.6901,
+      "step": 1965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958508340305402,
+      "loss": 0.3839,
+      "step": 1966
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958463992278899,
+      "loss": 1.385,
+      "step": 1967
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029584196208972833,
+      "loss": 0.2363,
+      "step": 1968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958375226161266,
+      "loss": 0.5098,
+      "step": 1969
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029583308080715585,
+      "loss": 1.4623,
+      "step": 1970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029582863666288717,
+      "loss": 0.2541,
+      "step": 1971
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029582419018339175,
+      "loss": 0.5022,
+      "step": 1972
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029581974136874077,
+      "loss": 0.4048,
+      "step": 1973
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958152902190054,
+      "loss": 0.6743,
+      "step": 1974
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002958108367342571,
+      "loss": 0.4714,
+      "step": 1975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029580638091456705,
+      "loss": 1.0729,
+      "step": 1976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029580192276000664,
+      "loss": 0.3908,
+      "step": 1977
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029579746227064726,
+      "loss": 0.5313,
+      "step": 1978
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957929994465603,
+      "loss": 0.029,
+      "step": 1979
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957885342878173,
+      "loss": 0.823,
+      "step": 1980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957840667944897,
+      "loss": 1.6146,
+      "step": 1981
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295779596966649,
+      "loss": 0.3431,
+      "step": 1982
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957751248043669,
+      "loss": 0.821,
+      "step": 1983
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295770650307715,
+      "loss": 0.5753,
+      "step": 1984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029576617347676484,
+      "loss": 0.0487,
+      "step": 1985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957616943115882,
+      "loss": 0.36,
+      "step": 1986
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029575721281225676,
+      "loss": 1.5451,
+      "step": 1987
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029575272897884233,
+      "loss": 0.6688,
+      "step": 1988
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029574824281141667,
+      "loss": 0.827,
+      "step": 1989
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957437543100517,
+      "loss": 0.2245,
+      "step": 1990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957392634748191,
+      "loss": 0.4005,
+      "step": 1991
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957347703057911,
+      "loss": 0.623,
+      "step": 1992
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029573027480303934,
+      "loss": 0.3453,
+      "step": 1993
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295725776966636,
+      "loss": 0.554,
+      "step": 1994
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029572127679665296,
+      "loss": 0.4085,
+      "step": 1995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957167742931625,
+      "loss": 1.4576,
+      "step": 1996
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029571226945623646,
+      "loss": 2.4577,
+      "step": 1997
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957077622859472,
+      "loss": 0.3314,
+      "step": 1998
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002957032527823668,
+      "loss": 1.243,
+      "step": 1999
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029569874094556743,
+      "loss": 0.6017,
+      "step": 2000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029569422677562144,
+      "loss": 0.2466,
+      "step": 2001
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029568971027260096,
+      "loss": 0.3344,
+      "step": 2002
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956851914365785,
+      "loss": 0.4456,
+      "step": 2003
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956806702676264,
+      "loss": 2.3467,
+      "step": 2004
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956761467658169,
+      "loss": 0.4629,
+      "step": 2005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029567162093122257,
+      "loss": 1.2373,
+      "step": 2006
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029566709276391585,
+      "loss": 0.495,
+      "step": 2007
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956625622639693,
+      "loss": 0.7967,
+      "step": 2008
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956580294314554,
+      "loss": 0.3093,
+      "step": 2009
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956534942664467,
+      "loss": 0.8315,
+      "step": 2010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029564895676901593,
+      "loss": 0.7545,
+      "step": 2011
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956444169392357,
+      "loss": 1.1071,
+      "step": 2012
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029563987477717875,
+      "loss": 0.1938,
+      "step": 2013
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029563533028291775,
+      "loss": 0.7323,
+      "step": 2014
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956307834565255,
+      "loss": 0.7768,
+      "step": 2015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956262342980748,
+      "loss": 1.0363,
+      "step": 2016
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029562168280763854,
+      "loss": 0.2956,
+      "step": 2017
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956171289852896,
+      "loss": 0.4789,
+      "step": 2018
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956125728311008,
+      "loss": 0.3713,
+      "step": 2019
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956080143451453,
+      "loss": 0.2789,
+      "step": 2020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002956034535274959,
+      "loss": 0.4783,
+      "step": 2021
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029559889037822574,
+      "loss": 0.4352,
+      "step": 2022
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955943248974078,
+      "loss": 0.3339,
+      "step": 2023
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955897570851153,
+      "loss": 0.4524,
+      "step": 2024
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029558518694142134,
+      "loss": 0.0254,
+      "step": 2025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955806144663991,
+      "loss": 0.5248,
+      "step": 2026
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029557603966012184,
+      "loss": 0.8193,
+      "step": 2027
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955714625226627,
+      "loss": 0.4201,
+      "step": 2028
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029556688305409516,
+      "loss": 0.9026,
+      "step": 2029
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029556230125449235,
+      "loss": 0.566,
+      "step": 2030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029555771712392784,
+      "loss": 0.4094,
+      "step": 2031
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029555313066247487,
+      "loss": 0.7143,
+      "step": 2032
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295548541870207,
+      "loss": 0.5933,
+      "step": 2033
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029554395074719766,
+      "loss": 0.5781,
+      "step": 2034
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029553935729352034,
+      "loss": 0.5477,
+      "step": 2035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955347615092487,
+      "loss": 3.0378,
+      "step": 2036
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029553016339445623,
+      "loss": 1.1007,
+      "step": 2037
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029552556294921663,
+      "loss": 0.0105,
+      "step": 2038
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955209601736035,
+      "loss": 1.8098,
+      "step": 2039
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029551635506769064,
+      "loss": 0.1363,
+      "step": 2040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029551174763155176,
+      "loss": 0.0137,
+      "step": 2041
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002955071378652606,
+      "loss": 0.0076,
+      "step": 2042
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029550252576889094,
+      "loss": 0.2702,
+      "step": 2043
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954979113425168,
+      "loss": 1.4455,
+      "step": 2044
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029549329458621184,
+      "loss": 0.9412,
+      "step": 2045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954886755000502,
+      "loss": 0.3426,
+      "step": 2046
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954840540841058,
+      "loss": 0.957,
+      "step": 2047
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954794303384526,
+      "loss": 0.3193,
+      "step": 2048
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029547480426316464,
+      "loss": 1.2109,
+      "step": 2049
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029547017585831596,
+      "loss": 0.7753,
+      "step": 2050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029546554512398086,
+      "loss": 0.4025,
+      "step": 2051
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029546091206023327,
+      "loss": 0.7328,
+      "step": 2052
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954562766671475,
+      "loss": 1.8361,
+      "step": 2053
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954516389447978,
+      "loss": 0.1825,
+      "step": 2054
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029544699889325835,
+      "loss": 0.9426,
+      "step": 2055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029544235651260355,
+      "loss": 1.3362,
+      "step": 2056
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954377118029076,
+      "loss": 0.7484,
+      "step": 2057
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295433064764245,
+      "loss": 1.1604,
+      "step": 2058
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029542841539669017,
+      "loss": 2.5499,
+      "step": 2059
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954237637003175,
+      "loss": 0.9694,
+      "step": 2060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029541910967520154,
+      "loss": 0.864,
+      "step": 2061
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029541445332141673,
+      "loss": 0.4775,
+      "step": 2062
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029540979463903773,
+      "loss": 1.1585,
+      "step": 2063
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954051336281391,
+      "loss": 0.4839,
+      "step": 2064
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002954004702887954,
+      "loss": 0.4499,
+      "step": 2065
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029539580462108146,
+      "loss": 0.2593,
+      "step": 2066
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953911366250719,
+      "loss": 0.7209,
+      "step": 2067
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953864663008415,
+      "loss": 0.4966,
+      "step": 2068
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295381793648465,
+      "loss": 0.4503,
+      "step": 2069
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953771186680173,
+      "loss": 0.5065,
+      "step": 2070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953724413595732,
+      "loss": 0.8913,
+      "step": 2071
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953677617232076,
+      "loss": 0.7994,
+      "step": 2072
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953630797589955,
+      "loss": 0.1868,
+      "step": 2073
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029535839546701185,
+      "loss": 0.7764,
+      "step": 2074
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953537088473316,
+      "loss": 2.459,
+      "step": 2075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953490199000299,
+      "loss": 0.6697,
+      "step": 2076
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029534432862518176,
+      "loss": 0.5883,
+      "step": 2077
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029533963502286235,
+      "loss": 0.0612,
+      "step": 2078
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029533493909314685,
+      "loss": 0.4673,
+      "step": 2079
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029533024083611034,
+      "loss": 0.4739,
+      "step": 2080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953255402518281,
+      "loss": 0.5781,
+      "step": 2081
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029532083734037554,
+      "loss": 0.1359,
+      "step": 2082
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953161321018278,
+      "loss": 0.3603,
+      "step": 2083
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953114245362603,
+      "loss": 0.1357,
+      "step": 2084
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029530671464374843,
+      "loss": 0.1872,
+      "step": 2085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002953020024243676,
+      "loss": 0.7844,
+      "step": 2086
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029529728787819326,
+      "loss": 0.5784,
+      "step": 2087
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029529257100530094,
+      "loss": 0.0629,
+      "step": 2088
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029528785180576615,
+      "loss": 2.4616,
+      "step": 2089
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952831302796644,
+      "loss": 2.2758,
+      "step": 2090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952784064270714,
+      "loss": 0.3345,
+      "step": 2091
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029527368024806277,
+      "loss": 0.4713,
+      "step": 2092
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952689517427141,
+      "loss": 0.3064,
+      "step": 2093
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029526422091110126,
+      "loss": 0.4638,
+      "step": 2094
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029525948775329996,
+      "loss": 0.0402,
+      "step": 2095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952547522693859,
+      "loss": 0.3396,
+      "step": 2096
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295250014459435,
+      "loss": 0.0052,
+      "step": 2097
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952452743235231,
+      "loss": 0.5895,
+      "step": 2098
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029524053186172615,
+      "loss": 1.0444,
+      "step": 2099
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029523578707412,
+      "loss": 3.7695,
+      "step": 2100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029523103996078074,
+      "loss": 0.0255,
+      "step": 2101
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029522629052178436,
+      "loss": 0.5484,
+      "step": 2102
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029522153875720684,
+      "loss": 0.9042,
+      "step": 2103
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952167846671243,
+      "loss": 0.2755,
+      "step": 2104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029521202825161295,
+      "loss": 0.007,
+      "step": 2105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029520726951074886,
+      "loss": 1.1676,
+      "step": 2106
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002952025084446083,
+      "loss": 0.6597,
+      "step": 2107
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951977450532675,
+      "loss": 0.6851,
+      "step": 2108
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951929793368027,
+      "loss": 0.3638,
+      "step": 2109
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029518821129529023,
+      "loss": 0.8805,
+      "step": 2110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00029518344092880646,
+      "loss": 1.0173,
+      "step": 2111
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951786682374278,
+      "loss": 2.3068,
+      "step": 2112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951738932212306,
+      "loss": 0.384,
+      "step": 2113
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951691158802914,
+      "loss": 0.4121,
+      "step": 2114
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951643362146867,
+      "loss": 0.4901,
+      "step": 2115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295159554224493,
+      "loss": 0.737,
+      "step": 2116
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002951547699097869,
+      "loss": 0.4147,
+      "step": 2117
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000295149983270645,
+      "loss": 0.3156,
+      "step": 2118
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029514519430714396,
+      "loss": 0.3517,
+      "step": 2119
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029514040301936045,
+      "loss": 1.1974,
+      "step": 2120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002951356094073712,
+      "loss": 0.4051,
+      "step": 2121
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000295130813471253,
+      "loss": 0.7974,
+      "step": 2122
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029512601521108266,
+      "loss": 0.2159,
+      "step": 2123
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029512121462693696,
+      "loss": 0.2538,
+      "step": 2124
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002951164117188928,
+      "loss": 0.1834,
+      "step": 2125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029511160648702713,
+      "loss": 0.2288,
+      "step": 2126
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029510679893141685,
+      "loss": 1.6398,
+      "step": 2127
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029510198905213894,
+      "loss": 0.2693,
+      "step": 2128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029509717684927044,
+      "loss": 0.3189,
+      "step": 2129
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029509236232288843,
+      "loss": 0.3498,
+      "step": 2130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029508754547307,
+      "loss": 0.5925,
+      "step": 2131
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950827262998923,
+      "loss": 0.6304,
+      "step": 2132
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950779048034324,
+      "loss": 0.6246,
+      "step": 2133
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950730809837676,
+      "loss": 0.9472,
+      "step": 2134
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950682548409752,
+      "loss": 0.4369,
+      "step": 2135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950634263751324,
+      "loss": 0.3263,
+      "step": 2136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950585955863165,
+      "loss": 1.8192,
+      "step": 2137
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950537624746049,
+      "loss": 0.168,
+      "step": 2138
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029504892704007503,
+      "loss": 1.1371,
+      "step": 2139
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029504408928280424,
+      "loss": 0.6745,
+      "step": 2140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029503924920287,
+      "loss": 0.7927,
+      "step": 2141
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029503440680035,
+      "loss": 0.4361,
+      "step": 2142
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950295620753215,
+      "loss": 0.3524,
+      "step": 2143
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950247150278623,
+      "loss": 0.3842,
+      "step": 2144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029501986565804984,
+      "loss": 0.4382,
+      "step": 2145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029501501396596194,
+      "loss": 3.7521,
+      "step": 2146
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029501015995167626,
+      "loss": 0.4647,
+      "step": 2147
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950053036152705,
+      "loss": 1.1955,
+      "step": 2148
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002950004449568224,
+      "loss": 0.4406,
+      "step": 2149
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029499558397640975,
+      "loss": 0.7835,
+      "step": 2150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029499072067411047,
+      "loss": 0.4951,
+      "step": 2151
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029498585505000237,
+      "loss": 0.4499,
+      "step": 2152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029498098710416346,
+      "loss": 0.3909,
+      "step": 2153
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949761168366716,
+      "loss": 0.3976,
+      "step": 2154
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949712442476047,
+      "loss": 0.2121,
+      "step": 2155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029496636933704106,
+      "loss": 0.2361,
+      "step": 2156
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949614921050585,
+      "loss": 0.3713,
+      "step": 2157
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949566125517352,
+      "loss": 1.9898,
+      "step": 2158
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949517306771493,
+      "loss": 0.5532,
+      "step": 2159
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294946846481379,
+      "loss": 0.4467,
+      "step": 2160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949419599645025,
+      "loss": 0.7631,
+      "step": 2161
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294937071126598,
+      "loss": 0.4326,
+      "step": 2162
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949321799677438,
+      "loss": 0.0363,
+      "step": 2163
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029492728648801826,
+      "loss": 0.6426,
+      "step": 2164
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029492239068749985,
+      "loss": 0.2398,
+      "step": 2165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949174925662667,
+      "loss": 0.0255,
+      "step": 2166
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029491259212439746,
+      "loss": 0.5485,
+      "step": 2167
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029490768936197057,
+      "loss": 0.785,
+      "step": 2168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002949027842790645,
+      "loss": 0.0428,
+      "step": 2169
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029489787687575773,
+      "loss": 0.5548,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294892967152129,
+      "loss": 0.4269,
+      "step": 2171
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029488805510825687,
+      "loss": 0.6864,
+      "step": 2172
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948831407442199,
+      "loss": 0.462,
+      "step": 2173
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948782240600969,
+      "loss": 0.1321,
+      "step": 2174
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948733050559667,
+      "loss": 0.0037,
+      "step": 2175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948683837319078,
+      "loss": 0.8147,
+      "step": 2176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029486346008799914,
+      "loss": 0.7243,
+      "step": 2177
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029485853412431964,
+      "loss": 0.7363,
+      "step": 2178
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948536058409481,
+      "loss": 1.7016,
+      "step": 2179
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948486752379635,
+      "loss": 0.5049,
+      "step": 2180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029484374231544466,
+      "loss": 0.9925,
+      "step": 2181
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029483880707347073,
+      "loss": 0.3502,
+      "step": 2182
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029483386951212066,
+      "loss": 0.3021,
+      "step": 2183
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948289296314736,
+      "loss": 0.8339,
+      "step": 2184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029482398743160853,
+      "loss": 1.1447,
+      "step": 2185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029481904291260466,
+      "loss": 0.3248,
+      "step": 2186
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029481409607454113,
+      "loss": 1.3021,
+      "step": 2187
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948091469174972,
+      "loss": 0.3473,
+      "step": 2188
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002948041954415521,
+      "loss": 1.0422,
+      "step": 2189
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947992416467852,
+      "loss": 0.0634,
+      "step": 2190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947942855332757,
+      "loss": 0.1102,
+      "step": 2191
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029478932710110304,
+      "loss": 0.8585,
+      "step": 2192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947843663503466,
+      "loss": 0.9264,
+      "step": 2193
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947794032810858,
+      "loss": 0.3471,
+      "step": 2194
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947744378934002,
+      "loss": 0.4879,
+      "step": 2195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947694701873692,
+      "loss": 0.3766,
+      "step": 2196
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947645001630724,
+      "loss": 0.5377,
+      "step": 2197
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029475952782058944,
+      "loss": 0.5415,
+      "step": 2198
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029475455315999984,
+      "loss": 0.3992,
+      "step": 2199
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029474957618138333,
+      "loss": 0.5332,
+      "step": 2200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947445968848196,
+      "loss": 0.3479,
+      "step": 2201
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029473961527038843,
+      "loss": 2.4514,
+      "step": 2202
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947346313381695,
+      "loss": 0.2617,
+      "step": 2203
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029472964508824273,
+      "loss": 0.3353,
+      "step": 2204
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947246565206878,
+      "loss": 0.3675,
+      "step": 2205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947196656355847,
+      "loss": 0.4993,
+      "step": 2206
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029471467243301347,
+      "loss": 0.2476,
+      "step": 2207
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947096769130538,
+      "loss": 0.5226,
+      "step": 2208
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002947046790757859,
+      "loss": 1.8381,
+      "step": 2209
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946996789212898,
+      "loss": 0.3639,
+      "step": 2210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946946764496454,
+      "loss": 0.526,
+      "step": 2211
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294689671660933,
+      "loss": 0.3114,
+      "step": 2212
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029468466455523254,
+      "loss": 2.0346,
+      "step": 2213
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946796551326244,
+      "loss": 0.3638,
+      "step": 2214
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029467464339318875,
+      "loss": 0.4863,
+      "step": 2215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029466962933700575,
+      "loss": 1.3555,
+      "step": 2216
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946646129641558,
+      "loss": 0.8401,
+      "step": 2217
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946595942747192,
+      "loss": 0.7897,
+      "step": 2218
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946545732687762,
+      "loss": 0.5361,
+      "step": 2219
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946495499464074,
+      "loss": 0.0032,
+      "step": 2220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029464452430769307,
+      "loss": 0.4871,
+      "step": 2221
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029463949635271384,
+      "loss": 1.6663,
+      "step": 2222
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946344660815501,
+      "loss": 0.4013,
+      "step": 2223
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946294334942825,
+      "loss": 0.6075,
+      "step": 2224
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029462439859099154,
+      "loss": 0.4179,
+      "step": 2225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029461936137175785,
+      "loss": 0.3479,
+      "step": 2226
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946143218366622,
+      "loss": 0.2047,
+      "step": 2227
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946092799857852,
+      "loss": 0.5578,
+      "step": 2228
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002946042358192076,
+      "loss": 0.6157,
+      "step": 2229
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029459918933701025,
+      "loss": 0.3897,
+      "step": 2230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945941405392738,
+      "loss": 0.879,
+      "step": 2231
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945890894260792,
+      "loss": 0.8428,
+      "step": 2232
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029458403599750735,
+      "loss": 0.429,
+      "step": 2233
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945789802536392,
+      "loss": 0.3471,
+      "step": 2234
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029457392219455565,
+      "loss": 0.5923,
+      "step": 2235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945688618203377,
+      "loss": 0.3055,
+      "step": 2236
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945637991310664,
+      "loss": 0.4414,
+      "step": 2237
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029455873412682275,
+      "loss": 0.8274,
+      "step": 2238
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294553666807688,
+      "loss": 0.6636,
+      "step": 2239
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029454859717374323,
+      "loss": 0.9697,
+      "step": 2240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945435252250696,
+      "loss": 0.8747,
+      "step": 2241
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945384509617483,
+      "loss": 0.2815,
+      "step": 2242
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029453337438386064,
+      "loss": 0.8887,
+      "step": 2243
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945282954914879,
+      "loss": 1.6843,
+      "step": 2244
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029452321428471143,
+      "loss": 0.1885,
+      "step": 2245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945181307636126,
+      "loss": 0.8009,
+      "step": 2246
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002945130449282728,
+      "loss": 0.631,
+      "step": 2247
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029450795677877346,
+      "loss": 0.5507,
+      "step": 2248
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029450286631519605,
+      "loss": 1.9204,
+      "step": 2249
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944977735376221,
+      "loss": 0.3173,
+      "step": 2250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944926784461332,
+      "loss": 0.3022,
+      "step": 2251
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944875810408109,
+      "loss": 0.4985,
+      "step": 2252
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944824813217368,
+      "loss": 0.1715,
+      "step": 2253
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029447737928899265,
+      "loss": 0.1513,
+      "step": 2254
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029447227494266005,
+      "loss": 0.6207,
+      "step": 2255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029446716828282086,
+      "loss": 0.468,
+      "step": 2256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944620593095567,
+      "loss": 0.4318,
+      "step": 2257
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944569480229495,
+      "loss": 0.3637,
+      "step": 2258
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944518344230811,
+      "loss": 0.7435,
+      "step": 2259
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029444671851003334,
+      "loss": 0.2783,
+      "step": 2260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944416002838882,
+      "loss": 2.6742,
+      "step": 2261
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944364797447276,
+      "loss": 0.4801,
+      "step": 2262
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944313568926335,
+      "loss": 0.4221,
+      "step": 2263
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294426231727688,
+      "loss": 0.3593,
+      "step": 2264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944211042499732,
+      "loss": 0.3617,
+      "step": 2265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002944159744595711,
+      "loss": 0.4354,
+      "step": 2266
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029441084235656395,
+      "loss": 0.4088,
+      "step": 2267
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029440570794103384,
+      "loss": 1.0703,
+      "step": 2268
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029440057121306307,
+      "loss": 0.7056,
+      "step": 2269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943954321727339,
+      "loss": 0.6515,
+      "step": 2270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029439029082012856,
+      "loss": 2.5967,
+      "step": 2271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943851471553294,
+      "loss": 0.5173,
+      "step": 2272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943800011784188,
+      "loss": 0.0723,
+      "step": 2273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029437485288947917,
+      "loss": 0.8957,
+      "step": 2274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029436970228859295,
+      "loss": 0.0129,
+      "step": 2275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029436454937584264,
+      "loss": 0.4418,
+      "step": 2276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943593941513107,
+      "loss": 0.5476,
+      "step": 2277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943542366150797,
+      "loss": 0.6769,
+      "step": 2278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943490767672323,
+      "loss": 0.0081,
+      "step": 2279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294343914607851,
+      "loss": 0.9168,
+      "step": 2280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029433875013701864,
+      "loss": 1.3694,
+      "step": 2281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943335833548177,
+      "loss": 2.4886,
+      "step": 2282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029432841426133114,
+      "loss": 0.7434,
+      "step": 2283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943232428566416,
+      "loss": 0.6971,
+      "step": 2284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943180691408319,
+      "loss": 0.4448,
+      "step": 2285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029431289311398497,
+      "loss": 0.668,
+      "step": 2286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029430771477618355,
+      "loss": 0.5011,
+      "step": 2287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002943025341275107,
+      "loss": 0.4292,
+      "step": 2288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029429735116804934,
+      "loss": 0.905,
+      "step": 2289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942921658978825,
+      "loss": 0.5756,
+      "step": 2290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029428697831709313,
+      "loss": 0.0213,
+      "step": 2291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029428178842576437,
+      "loss": 0.9554,
+      "step": 2292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029427659622397926,
+      "loss": 0.5424,
+      "step": 2293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294271401711821,
+      "loss": 0.0483,
+      "step": 2294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942662048893728,
+      "loss": 0.8414,
+      "step": 2295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942610057567178,
+      "loss": 0.0238,
+      "step": 2296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029425580431393934,
+      "loss": 0.523,
+      "step": 2297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029425060056112066,
+      "loss": 1.3754,
+      "step": 2298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029424539449834507,
+      "loss": 0.7274,
+      "step": 2299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294240186125696,
+      "loss": 0.6601,
+      "step": 2300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942349754432568,
+      "loss": 0.3217,
+      "step": 2301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029422976245111095,
+      "loss": 0.1551,
+      "step": 2302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942245471493419,
+      "loss": 0.1491,
+      "step": 2303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942193295380332,
+      "loss": 0.6983,
+      "step": 2304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002942141096172684,
+      "loss": 0.3163,
+      "step": 2305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294208887387131,
+      "loss": 0.5731,
+      "step": 2306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029420366284770477,
+      "loss": 0.6275,
+      "step": 2307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029419843599907323,
+      "loss": 1.1048,
+      "step": 2308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941932068413202,
+      "loss": 0.5871,
+      "step": 2309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029418797537452936,
+      "loss": 0.5181,
+      "step": 2310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941827415987845,
+      "loss": 0.2913,
+      "step": 2311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941775055141694,
+      "loss": 0.5501,
+      "step": 2312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941722671207679,
+      "loss": 0.9195,
+      "step": 2313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029416702641866393,
+      "loss": 2.8206,
+      "step": 2314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941617834079414,
+      "loss": 0.5744,
+      "step": 2315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941565380886843,
+      "loss": 0.1525,
+      "step": 2316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029415129046097654,
+      "loss": 0.9401,
+      "step": 2317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029414604052490224,
+      "loss": 0.0307,
+      "step": 2318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029414078828054547,
+      "loss": 1.6692,
+      "step": 2319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029413553372799025,
+      "loss": 1.3795,
+      "step": 2320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029413027686732077,
+      "loss": 1.32,
+      "step": 2321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941250176986212,
+      "loss": 0.0076,
+      "step": 2322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029411975622197585,
+      "loss": 0.7726,
+      "step": 2323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029411449243746884,
+      "loss": 0.5174,
+      "step": 2324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029410922634518454,
+      "loss": 0.6221,
+      "step": 2325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002941039579452072,
+      "loss": 0.0207,
+      "step": 2326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029409868723762126,
+      "loss": 1.7489,
+      "step": 2327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029409341422251115,
+      "loss": 2.5337,
+      "step": 2328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029408813889996127,
+      "loss": 3.6233,
+      "step": 2329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029408286127005607,
+      "loss": 2.7544,
+      "step": 2330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029407758133288006,
+      "loss": 0.4265,
+      "step": 2331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940722990885179,
+      "loss": 0.483,
+      "step": 2332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000294067014537054,
+      "loss": 0.6575,
+      "step": 2333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940617276785731,
+      "loss": 0.5222,
+      "step": 2334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029405643851315987,
+      "loss": 0.0063,
+      "step": 2335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940511470408989,
+      "loss": 1.5338,
+      "step": 2336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940458532618751,
+      "loss": 0.3509,
+      "step": 2337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029404055717617306,
+      "loss": 0.498,
+      "step": 2338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029403525878387774,
+      "loss": 3.65,
+      "step": 2339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029402995808507386,
+      "loss": 0.6986,
+      "step": 2340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029402465507984637,
+      "loss": 0.8264,
+      "step": 2341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940193497682802,
+      "loss": 0.6693,
+      "step": 2342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002940140421504603,
+      "loss": 0.1882,
+      "step": 2343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029400873222647163,
+      "loss": 0.5159,
+      "step": 2344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029400341999639924,
+      "loss": 0.5223,
+      "step": 2345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029399810546032815,
+      "loss": 0.3921,
+      "step": 2346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939927886183436,
+      "loss": 1.2905,
+      "step": 2347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939874694705306,
+      "loss": 0.4953,
+      "step": 2348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029398214801697436,
+      "loss": 0.2998,
+      "step": 2349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029397682425776013,
+      "loss": 0.4107,
+      "step": 2350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939714981929731,
+      "loss": 1.5027,
+      "step": 2351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029396616982269865,
+      "loss": 0.3948,
+      "step": 2352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029396083914702206,
+      "loss": 0.3528,
+      "step": 2353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939555061660286,
+      "loss": 0.2366,
+      "step": 2354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939501708798038,
+      "loss": 0.4354,
+      "step": 2355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939448332884331,
+      "loss": 0.1319,
+      "step": 2356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939394933920019,
+      "loss": 0.4216,
+      "step": 2357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939341511905957,
+      "loss": 0.5259,
+      "step": 2358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029392880668430014,
+      "loss": 0.1102,
+      "step": 2359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029392345987320074,
+      "loss": 1.4425,
+      "step": 2360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002939181107573831,
+      "loss": 0.6926,
+      "step": 2361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029391275933693286,
+      "loss": 0.0173,
+      "step": 2362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029390740561193585,
+      "loss": 0.0686,
+      "step": 2363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029390204958247763,
+      "loss": 0.9381,
+      "step": 2364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029389669124864413,
+      "loss": 0.2562,
+      "step": 2365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029389133061052107,
+      "loss": 0.006,
+      "step": 2366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00029388596766819424,
+      "loss": 0.0811,
+      "step": 2367
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029388060242174964,
+      "loss": 0.5287,
+      "step": 2368
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002938752348712731,
+      "loss": 0.5125,
+      "step": 2369
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002938698650168506,
+      "loss": 1.9581,
+      "step": 2370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029386449285856816,
+      "loss": 0.6891,
+      "step": 2371
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002938591183965117,
+      "loss": 0.7247,
+      "step": 2372
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002938537416307675,
+      "loss": 3.7547,
+      "step": 2373
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002938483625614214,
+      "loss": 0.2627,
+      "step": 2374
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002938429811885597,
+      "loss": 0.2221,
+      "step": 2375
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029383759751226853,
+      "loss": 2.346,
+      "step": 2376
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029383221153263413,
+      "loss": 0.8554,
+      "step": 2377
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002938268232497427,
+      "loss": 0.5567,
+      "step": 2378
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029382143266368054,
+      "loss": 0.8884,
+      "step": 2379
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029381603977453395,
+      "loss": 0.0239,
+      "step": 2380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002938106445823894,
+      "loss": 1.6237,
+      "step": 2381
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029380524708733313,
+      "loss": 2.2329,
+      "step": 2382
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029379984728945173,
+      "loss": 0.4629,
+      "step": 2383
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029379444518883155,
+      "loss": 0.2774,
+      "step": 2384
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937890407855591,
+      "loss": 2.6117,
+      "step": 2385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293783634079721,
+      "loss": 0.7404,
+      "step": 2386
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937782250714038,
+      "loss": 0.547,
+      "step": 2387
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937728137606941,
+      "loss": 1.207,
+      "step": 2388
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029376740014767856,
+      "loss": 1.0524,
+      "step": 2389
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029376198423244383,
+      "loss": 0.32,
+      "step": 2390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937565660150767,
+      "loss": 0.6251,
+      "step": 2391
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029375114549566394,
+      "loss": 0.6223,
+      "step": 2392
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937457226742923,
+      "loss": 0.0636,
+      "step": 2393
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029374029755104865,
+      "loss": 0.2682,
+      "step": 2394
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937348701260199,
+      "loss": 0.0394,
+      "step": 2395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937294403992929,
+      "loss": 0.1277,
+      "step": 2396
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937240083709546,
+      "loss": 0.3899,
+      "step": 2397
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293718574041092,
+      "loss": 0.7485,
+      "step": 2398
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002937131374097921,
+      "loss": 0.5985,
+      "step": 2399
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029370769847714205,
+      "loss": 0.4614,
+      "step": 2400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029370225724322885,
+      "loss": 0.8594,
+      "step": 2401
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029369681370813963,
+      "loss": 0.7012,
+      "step": 2402
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936913678719617,
+      "loss": 0.5537,
+      "step": 2403
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029368591973478206,
+      "loss": 0.7513,
+      "step": 2404
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936804692966881,
+      "loss": 0.681,
+      "step": 2405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029367501655776706,
+      "loss": 0.484,
+      "step": 2406
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936695615181063,
+      "loss": 0.1413,
+      "step": 2407
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936641041777931,
+      "loss": 1.6241,
+      "step": 2408
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936586445369148,
+      "loss": 0.1929,
+      "step": 2409
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293653182595559,
+      "loss": 0.4747,
+      "step": 2410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936477183538131,
+      "loss": 0.2735,
+      "step": 2411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936422518117645,
+      "loss": 0.1057,
+      "step": 2412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936367829695008,
+      "loss": 0.5853,
+      "step": 2413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936313118271097,
+      "loss": 0.3527,
+      "step": 2414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029362583838467857,
+      "loss": 0.7464,
+      "step": 2415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936203626422953,
+      "loss": 0.0236,
+      "step": 2416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936148846000474,
+      "loss": 0.0429,
+      "step": 2417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002936094042580227,
+      "loss": 0.8147,
+      "step": 2418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029360392161630885,
+      "loss": 2.0524,
+      "step": 2419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029359843667499375,
+      "loss": 0.5595,
+      "step": 2420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029359294943416526,
+      "loss": 0.0111,
+      "step": 2421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029358745989391114,
+      "loss": 0.2972,
+      "step": 2422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029358196805431933,
+      "loss": 0.019,
+      "step": 2423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935764739154778,
+      "loss": 2.4233,
+      "step": 2424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935709774774745,
+      "loss": 0.1753,
+      "step": 2425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935654787403975,
+      "loss": 0.3079,
+      "step": 2426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935599777043348,
+      "loss": 0.6071,
+      "step": 2427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029355447436937447,
+      "loss": 0.027,
+      "step": 2428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029354896873560473,
+      "loss": 0.5005,
+      "step": 2429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935434608031137,
+      "loss": 0.0087,
+      "step": 2430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935379505719896,
+      "loss": 1.4845,
+      "step": 2431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029353243804232056,
+      "loss": 0.6282,
+      "step": 2432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029352692321419493,
+      "loss": 1.3454,
+      "step": 2433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935214060877011,
+      "loss": 0.3686,
+      "step": 2434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935158866629273,
+      "loss": 0.4234,
+      "step": 2435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293510364939962,
+      "loss": 0.7052,
+      "step": 2436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002935048409188936,
+      "loss": 2.1507,
+      "step": 2437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029349931459981047,
+      "loss": 0.596,
+      "step": 2438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934937859828012,
+      "loss": 0.2778,
+      "step": 2439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934882550679544,
+      "loss": 0.3016,
+      "step": 2440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029348272185535844,
+      "loss": 0.4963,
+      "step": 2441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293477186345102,
+      "loss": 2.7574,
+      "step": 2442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029347164853727386,
+      "loss": 0.0079,
+      "step": 2443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934661084319625,
+      "loss": 0.6956,
+      "step": 2444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934605660292568,
+      "loss": 0.4086,
+      "step": 2445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029345502132924536,
+      "loss": 0.3194,
+      "step": 2446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029344947433201707,
+      "loss": 0.6537,
+      "step": 2447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029344392503766073,
+      "loss": 1.6508,
+      "step": 2448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029343837344626526,
+      "loss": 0.4832,
+      "step": 2449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029343281955791946,
+      "loss": 0.0017,
+      "step": 2450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029342726337271233,
+      "loss": 0.0233,
+      "step": 2451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029342170489073284,
+      "loss": 0.4756,
+      "step": 2452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029341614411207,
+      "loss": 0.7188,
+      "step": 2453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934105810368128,
+      "loss": 0.3446,
+      "step": 2454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002934050156650505,
+      "loss": 0.707,
+      "step": 2455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293399447996872,
+      "loss": 0.8833,
+      "step": 2456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933938780323665,
+      "loss": 0.9731,
+      "step": 2457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933883057716233,
+      "loss": 1.1877,
+      "step": 2458
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029338273121473163,
+      "loss": 0.514,
+      "step": 2459
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933771543617806,
+      "loss": 0.4467,
+      "step": 2460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933715752128597,
+      "loss": 0.5766,
+      "step": 2461
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933659937680582,
+      "loss": 0.694,
+      "step": 2462
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933604100274655,
+      "loss": 0.3897,
+      "step": 2463
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933548239911709,
+      "loss": 0.4439,
+      "step": 2464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029334923565926395,
+      "loss": 3.49,
+      "step": 2465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933436450318342,
+      "loss": 0.7767,
+      "step": 2466
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029333805210897103,
+      "loss": 1.42,
+      "step": 2467
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933324568907641,
+      "loss": 1.0078,
+      "step": 2468
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293326859377303,
+      "loss": 0.4147,
+      "step": 2469
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029332125956867735,
+      "loss": 0.3582,
+      "step": 2470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002933156574649768,
+      "loss": 0.2734,
+      "step": 2471
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029331005306629113,
+      "loss": 0.354,
+      "step": 2472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029330444637271,
+      "loss": 0.0819,
+      "step": 2473
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029329883738432323,
+      "loss": 0.674,
+      "step": 2474
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932932261012206,
+      "loss": 0.2126,
+      "step": 2475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932876125234921,
+      "loss": 0.814,
+      "step": 2476
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932819966512275,
+      "loss": 1.0446,
+      "step": 2477
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932763784845167,
+      "loss": 0.775,
+      "step": 2478
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029327075802344977,
+      "loss": 0.6052,
+      "step": 2479
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029326513526811666,
+      "loss": 0.6949,
+      "step": 2480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029325951021860747,
+      "loss": 0.3504,
+      "step": 2481
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029325951021860747,
+      "loss": 1.5343,
+      "step": 2482
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932538828750122,
+      "loss": 0.3815,
+      "step": 2483
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029324825323742096,
+      "loss": 0.7963,
+      "step": 2484
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932426213059239,
+      "loss": 0.7522,
+      "step": 2485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029323698708061125,
+      "loss": 0.5938,
+      "step": 2486
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932313505615733,
+      "loss": 0.4582,
+      "step": 2487
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029322571174890015,
+      "loss": 0.8109,
+      "step": 2488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932200706426822,
+      "loss": 0.7002,
+      "step": 2489
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029321442724300977,
+      "loss": 0.3347,
+      "step": 2490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002932087815499733,
+      "loss": 0.7794,
+      "step": 2491
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293203133563663,
+      "loss": 0.3165,
+      "step": 2492
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931974832841695,
+      "loss": 0.2842,
+      "step": 2493
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029319183071158317,
+      "loss": 0.3621,
+      "step": 2494
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931861758459946,
+      "loss": 0.4394,
+      "step": 2495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931805186874943,
+      "loss": 2.907,
+      "step": 2496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931748592361729,
+      "loss": 2.5243,
+      "step": 2497
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293169197492121,
+      "loss": 0.3047,
+      "step": 2498
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029316353345542923,
+      "loss": 1.1268,
+      "step": 2499
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931578671261884,
+      "loss": 0.7788,
+      "step": 2500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029315219850448913,
+      "loss": 0.2955,
+      "step": 2501
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029314652759042224,
+      "loss": 1.03,
+      "step": 2502
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029314085438407857,
+      "loss": 0.3138,
+      "step": 2503
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029313517888554893,
+      "loss": 0.7221,
+      "step": 2504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931295010949242,
+      "loss": 0.5437,
+      "step": 2505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931238210122953,
+      "loss": 0.8696,
+      "step": 2506
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931181386377533,
+      "loss": 0.3989,
+      "step": 2507
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000293112453971389,
+      "loss": 0.3588,
+      "step": 2508
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002931067670132936,
+      "loss": 0.5692,
+      "step": 2509
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029310107776355806,
+      "loss": 0.3786,
+      "step": 2510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930953862222735,
+      "loss": 1.2952,
+      "step": 2511
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930896923895311,
+      "loss": 0.8107,
+      "step": 2512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930839962654221,
+      "loss": 0.4119,
+      "step": 2513
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029307829785003755,
+      "loss": 0.6514,
+      "step": 2514
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930725971434688,
+      "loss": 0.9037,
+      "step": 2515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930668941458072,
+      "loss": 0.3681,
+      "step": 2516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930611888571439,
+      "loss": 0.321,
+      "step": 2517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029305548127757044,
+      "loss": 0.9423,
+      "step": 2518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930497714071781,
+      "loss": 0.0062,
+      "step": 2519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029304405924605837,
+      "loss": 0.7918,
+      "step": 2520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930383447943027,
+      "loss": 0.858,
+      "step": 2521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029303262805200264,
+      "loss": 0.4526,
+      "step": 2522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029302690901924973,
+      "loss": 0.4959,
+      "step": 2523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029302118769613546,
+      "loss": 0.6132,
+      "step": 2524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029301546408275154,
+      "loss": 0.6183,
+      "step": 2525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002930097381791896,
+      "loss": 0.6585,
+      "step": 2526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029300400998554134,
+      "loss": 1.7988,
+      "step": 2527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929982795018985,
+      "loss": 0.5585,
+      "step": 2528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029299254672835275,
+      "loss": 0.0661,
+      "step": 2529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029298681166499603,
+      "loss": 0.7228,
+      "step": 2530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929810743119201,
+      "loss": 0.1031,
+      "step": 2531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929753346692169,
+      "loss": 1.8304,
+      "step": 2532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029296959273697817,
+      "loss": 0.2459,
+      "step": 2533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029296384851529604,
+      "loss": 0.4619,
+      "step": 2534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929581020042624,
+      "loss": 0.7704,
+      "step": 2535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029295235320396935,
+      "loss": 0.3514,
+      "step": 2536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929466021145089,
+      "loss": 0.6979,
+      "step": 2537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929408487359732,
+      "loss": 0.3962,
+      "step": 2538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929350930684543,
+      "loss": 0.0136,
+      "step": 2539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029292933511204433,
+      "loss": 1.1373,
+      "step": 2540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029292357486683567,
+      "loss": 0.2949,
+      "step": 2541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029291781233292035,
+      "loss": 1.2874,
+      "step": 2542
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029291204751039085,
+      "loss": 0.4398,
+      "step": 2543
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029290628039933933,
+      "loss": 0.0291,
+      "step": 2544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002929005109998582,
+      "loss": 1.0888,
+      "step": 2545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928947393120399,
+      "loss": 0.439,
+      "step": 2546
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928889653359768,
+      "loss": 0.7187,
+      "step": 2547
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928831890717613,
+      "loss": 0.6237,
+      "step": 2548
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292877410519486,
+      "loss": 0.8481,
+      "step": 2549
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928716296792434,
+      "loss": 0.5366,
+      "step": 2550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928658465511261,
+      "loss": 0.023,
+      "step": 2551
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029286006113522664,
+      "loss": 0.9813,
+      "step": 2552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029285427343163774,
+      "loss": 0.9033,
+      "step": 2553
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292848483440452,
+      "loss": 0.5136,
+      "step": 2554
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928426911617622,
+      "loss": 0.4399,
+      "step": 2555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029283689659566115,
+      "loss": 0.0583,
+      "step": 2556
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029283109974224145,
+      "loss": 0.6765,
+      "step": 2557
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928253006015961,
+      "loss": 0.0725,
+      "step": 2558
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928194991738179,
+      "loss": 0.8215,
+      "step": 2559
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002928136954589998,
+      "loss": 0.4982,
+      "step": 2560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029280788945723466,
+      "loss": 0.4659,
+      "step": 2561
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029280208116861555,
+      "loss": 0.0044,
+      "step": 2562
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927962705932354,
+      "loss": 0.0135,
+      "step": 2563
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029279045773118723,
+      "loss": 0.4199,
+      "step": 2564
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029278464258256423,
+      "loss": 0.0103,
+      "step": 2565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029277882514745947,
+      "loss": 0.4628,
+      "step": 2566
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029277300542596607,
+      "loss": 0.0025,
+      "step": 2567
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927671834181773,
+      "loss": 0.5404,
+      "step": 2568
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029276135912418633,
+      "loss": 0.4926,
+      "step": 2569
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927555325440864,
+      "loss": 0.2918,
+      "step": 2570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292749703677971,
+      "loss": 0.7254,
+      "step": 2571
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927438725259332,
+      "loss": 0.2525,
+      "step": 2572
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029273803908806657,
+      "loss": 0.9592,
+      "step": 2573
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029273220336446446,
+      "loss": 0.367,
+      "step": 2574
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029272636535522035,
+      "loss": 1.0793,
+      "step": 2575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029272052506042765,
+      "loss": 0.424,
+      "step": 2576
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927146824801799,
+      "loss": 2.5424,
+      "step": 2577
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927088376145708,
+      "loss": 0.1345,
+      "step": 2578
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002927029904636938,
+      "loss": 0.6754,
+      "step": 2579
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029269714102764255,
+      "loss": 0.5494,
+      "step": 2580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029269128930651075,
+      "loss": 0.3204,
+      "step": 2581
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029268543530039213,
+      "loss": 0.4086,
+      "step": 2582
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926795790093804,
+      "loss": 0.4667,
+      "step": 2583
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926737204335693,
+      "loss": 0.5598,
+      "step": 2584
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029266785957305275,
+      "loss": 0.5391,
+      "step": 2585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926619964279245,
+      "loss": 2.0431,
+      "step": 2586
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926561309982785,
+      "loss": 0.4481,
+      "step": 2587
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926502632842086,
+      "loss": 0.5743,
+      "step": 2588
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926443932858089,
+      "loss": 0.1951,
+      "step": 2589
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926385210031733,
+      "loss": 0.3339,
+      "step": 2590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029263264643639587,
+      "loss": 0.5305,
+      "step": 2591
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029262676958557066,
+      "loss": 0.5555,
+      "step": 2592
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029262089045079177,
+      "loss": 0.912,
+      "step": 2593
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029261500903215336,
+      "loss": 0.7105,
+      "step": 2594
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029260912532974966,
+      "loss": 0.0008,
+      "step": 2595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002926032393436748,
+      "loss": 0.0005,
+      "step": 2596
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925973510740231,
+      "loss": 0.0014,
+      "step": 2597
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925914605208888,
+      "loss": 2.6329,
+      "step": 2598
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925855676843663,
+      "loss": 0.001,
+      "step": 2599
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925796725645499,
+      "loss": 0.079,
+      "step": 2600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029257377516153405,
+      "loss": 0.7009,
+      "step": 2601
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925678754754131,
+      "loss": 0.6578,
+      "step": 2602
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925619735062817,
+      "loss": 0.0046,
+      "step": 2603
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925560692542342,
+      "loss": 0.6126,
+      "step": 2604
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925501627193652,
+      "loss": 0.0085,
+      "step": 2605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029254425390176923,
+      "loss": 0.3498,
+      "step": 2606
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292538342801541,
+      "loss": 1.5423,
+      "step": 2607
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029253242941877514,
+      "loss": 0.3087,
+      "step": 2608
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925265137535663,
+      "loss": 0.6949,
+      "step": 2609
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925205958060093,
+      "loss": 0.4507,
+      "step": 2610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029251467557619884,
+      "loss": 0.597,
+      "step": 2611
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002925087530642297,
+      "loss": 0.0088,
+      "step": 2612
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00029250282827019677,
+      "loss": 0.4552,
+      "step": 2613
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002924969011941949,
+      "loss": 0.0164,
+      "step": 2614
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000292490971836319,
+      "loss": 0.4816,
+      "step": 2615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002924850401966641,
+      "loss": 0.6562,
+      "step": 2616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002924791062753251,
+      "loss": 1.0109,
+      "step": 2617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000292473170072397,
+      "loss": 0.6601,
+      "step": 2618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002924672315879749,
+      "loss": 0.0311,
+      "step": 2619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029246129082215395,
+      "loss": 0.8509,
+      "step": 2620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002924553477750292,
+      "loss": 0.2096,
+      "step": 2621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002924494024466959,
+      "loss": 0.3817,
+      "step": 2622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029244345483724917,
+      "loss": 0.455,
+      "step": 2623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002924375049467843,
+      "loss": 0.9321,
+      "step": 2624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029243155277539653,
+      "loss": 0.3475,
+      "step": 2625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002924255983231812,
+      "loss": 0.5622,
+      "step": 2626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002924196415902337,
+      "loss": 0.4503,
+      "step": 2627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002924136825766494,
+      "loss": 1.2062,
+      "step": 2628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002924077212825236,
+      "loss": 0.6234,
+      "step": 2629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000292401757707952,
+      "loss": 0.4917,
+      "step": 2630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002923957918530299,
+      "loss": 0.7625,
+      "step": 2631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002923898237178529,
+      "loss": 0.303,
+      "step": 2632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002923838533025166,
+      "loss": 2.6312,
+      "step": 2633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029237788060711655,
+      "loss": 0.3118,
+      "step": 2634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002923719056317484,
+      "loss": 0.5686,
+      "step": 2635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029236592837650783,
+      "loss": 0.2564,
+      "step": 2636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029235994884149063,
+      "loss": 0.994,
+      "step": 2637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029235396702679254,
+      "loss": 0.3185,
+      "step": 2638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029234798293250924,
+      "loss": 0.9648,
+      "step": 2639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002923419965587367,
+      "loss": 0.4886,
+      "step": 2640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029233600790557065,
+      "loss": 0.3284,
+      "step": 2641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002923300169731071,
+      "loss": 0.2507,
+      "step": 2642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029232402376144187,
+      "loss": 1.321,
+      "step": 2643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029231802827067103,
+      "loss": 0.5593,
+      "step": 2644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002923120305008905,
+      "loss": 1.5628,
+      "step": 2645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029230603045219647,
+      "loss": 0.2047,
+      "step": 2646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029230002812468493,
+      "loss": 0.446,
+      "step": 2647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029229402351845196,
+      "loss": 0.2008,
+      "step": 2648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029228801663359373,
+      "loss": 1.16,
+      "step": 2649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002922820074702065,
+      "loss": 0.3847,
+      "step": 2650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002922759960283865,
+      "loss": 0.7801,
+      "step": 2651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002922699823082299,
+      "loss": 5.334,
+      "step": 2652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029226396630983306,
+      "loss": 0.6861,
+      "step": 2653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029225794803329225,
+      "loss": 0.3213,
+      "step": 2654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029225192747870396,
+      "loss": 2.008,
+      "step": 2655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029224590464616457,
+      "loss": 0.4802,
+      "step": 2656
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029223987953577046,
+      "loss": 0.3488,
+      "step": 2657
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002922338521476181,
+      "loss": 0.6615,
+      "step": 2658
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002922278224818041,
+      "loss": 0.2752,
+      "step": 2659
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029222179053842504,
+      "loss": 0.2712,
+      "step": 2660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029221575631757743,
+      "loss": 0.0553,
+      "step": 2661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029220971981935785,
+      "loss": 0.8702,
+      "step": 2662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002922036810438631,
+      "loss": 0.9495,
+      "step": 2663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029219763999118977,
+      "loss": 1.0555,
+      "step": 2664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921915966614347,
+      "loss": 0.0646,
+      "step": 2665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029218555105469453,
+      "loss": 0.8341,
+      "step": 2666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921795031710662,
+      "loss": 0.0125,
+      "step": 2667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029217345301064653,
+      "loss": 0.6714,
+      "step": 2668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921674005735324,
+      "loss": 0.6335,
+      "step": 2669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029216134585982064,
+      "loss": 0.8972,
+      "step": 2670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921552888696083,
+      "loss": 0.2506,
+      "step": 2671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029214922960299234,
+      "loss": 1.3595,
+      "step": 2672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029214316806006985,
+      "loss": 0.396,
+      "step": 2673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029213710424093786,
+      "loss": 1.505,
+      "step": 2674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029213103814569336,
+      "loss": 0.4928,
+      "step": 2675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921249697744337,
+      "loss": 0.6055,
+      "step": 2676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921188991272559,
+      "loss": 0.9607,
+      "step": 2677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921128262042572,
+      "loss": 0.0079,
+      "step": 2678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921067510055349,
+      "loss": 0.006,
+      "step": 2679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002921006735311863,
+      "loss": 1.0296,
+      "step": 2680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029209459378130864,
+      "loss": 0.2374,
+      "step": 2681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029208851175599926,
+      "loss": 1.7443,
+      "step": 2682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920824274553557,
+      "loss": 0.198,
+      "step": 2683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029207634087947525,
+      "loss": 0.578,
+      "step": 2684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029207025202845547,
+      "loss": 0.8216,
+      "step": 2685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029206416090239376,
+      "loss": 0.008,
+      "step": 2686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029205806750138777,
+      "loss": 1.1571,
+      "step": 2687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029205197182553496,
+      "loss": 0.8046,
+      "step": 2688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920458738749331,
+      "loss": 1.0991,
+      "step": 2689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920397736496797,
+      "loss": 0.0614,
+      "step": 2690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920336711498725,
+      "loss": 0.0156,
+      "step": 2691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029202756637560927,
+      "loss": 0.0434,
+      "step": 2692
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029202145932698763,
+      "loss": 0.0111,
+      "step": 2693
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920153500041055,
+      "loss": 0.0893,
+      "step": 2694
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002920092384070607,
+      "loss": 0.0237,
+      "step": 2695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029200312453595104,
+      "loss": 0.0285,
+      "step": 2696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002919970083908745,
+      "loss": 0.9549,
+      "step": 2697
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029199088997192894,
+      "loss": 0.2691,
+      "step": 2698
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002919847692792124,
+      "loss": 0.5661,
+      "step": 2699
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029197864631282284,
+      "loss": 0.8415,
+      "step": 2700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029197252107285835,
+      "loss": 0.4173,
+      "step": 2701
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029196639355941705,
+      "loss": 0.5952,
+      "step": 2702
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029196026377259693,
+      "loss": 0.3228,
+      "step": 2703
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029195413171249635,
+      "loss": 0.3265,
+      "step": 2704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002919479973792133,
+      "loss": 1.1885,
+      "step": 2705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002919418607728461,
+      "loss": 1.156,
+      "step": 2706
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029193572189349307,
+      "loss": 0.9065,
+      "step": 2707
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029192958074125247,
+      "loss": 0.0123,
+      "step": 2708
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029192343731622265,
+      "loss": 0.9571,
+      "step": 2709
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029191729161850196,
+      "loss": 0.2603,
+      "step": 2710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029191114364818883,
+      "loss": 0.9116,
+      "step": 2711
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002919049934053817,
+      "loss": 0.3535,
+      "step": 2712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918988408901791,
+      "loss": 0.1799,
+      "step": 2713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029189268610267953,
+      "loss": 0.0012,
+      "step": 2714
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918865290429815,
+      "loss": 0.0018,
+      "step": 2715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918803697111837,
+      "loss": 0.7749,
+      "step": 2716
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918742081073847,
+      "loss": 0.6578,
+      "step": 2717
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918680442316832,
+      "loss": 0.5879,
+      "step": 2718
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918618780841778,
+      "loss": 0.6032,
+      "step": 2719
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029185570966496735,
+      "loss": 0.4864,
+      "step": 2720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029184953897415063,
+      "loss": 0.5627,
+      "step": 2721
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918433660118265,
+      "loss": 0.4295,
+      "step": 2722
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029183719077809366,
+      "loss": 0.3666,
+      "step": 2723
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029183101327305105,
+      "loss": 0.08,
+      "step": 2724
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029182483349679767,
+      "loss": 1.0876,
+      "step": 2725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918186514494325,
+      "loss": 0.0015,
+      "step": 2726
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029181246713105434,
+      "loss": 0.4418,
+      "step": 2727
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918062805417624,
+      "loss": 1.1223,
+      "step": 2728
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002918000916816557,
+      "loss": 0.5687,
+      "step": 2729
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029179390055083333,
+      "loss": 1.6339,
+      "step": 2730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002917877071493945,
+      "loss": 0.0204,
+      "step": 2731
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002917815114774383,
+      "loss": 0.5665,
+      "step": 2732
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029177531353506394,
+      "loss": 0.5404,
+      "step": 2733
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029176911332237074,
+      "loss": 0.4842,
+      "step": 2734
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029176291083945794,
+      "loss": 0.3993,
+      "step": 2735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002917567060864249,
+      "loss": 0.3167,
+      "step": 2736
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291750499063371,
+      "loss": 5.222,
+      "step": 2737
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029174428977039554,
+      "loss": 0.7891,
+      "step": 2738
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291738078207598,
+      "loss": 0.8394,
+      "step": 2739
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029173186437507785,
+      "loss": 0.9859,
+      "step": 2740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029172564827293463,
+      "loss": 1.1328,
+      "step": 2741
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029171942990126787,
+      "loss": 0.8112,
+      "step": 2742
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002917132092601771,
+      "loss": 0.4057,
+      "step": 2743
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291706986349762,
+      "loss": 0.3766,
+      "step": 2744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029170076117012215,
+      "loss": 0.0013,
+      "step": 2745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029169453372135726,
+      "loss": 0.3128,
+      "step": 2746
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916883040035671,
+      "loss": 2.0938,
+      "step": 2747
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916820720168514,
+      "loss": 0.0011,
+      "step": 2748
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916758377613099,
+      "loss": 0.5517,
+      "step": 2749
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916696012370425,
+      "loss": 0.0005,
+      "step": 2750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029166336244414913,
+      "loss": 1.2712,
+      "step": 2751
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029165712138272957,
+      "loss": 0.2364,
+      "step": 2752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916508780528838,
+      "loss": 0.004,
+      "step": 2753
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916446324547118,
+      "loss": 0.9027,
+      "step": 2754
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029163838458831355,
+      "loss": 0.4729,
+      "step": 2755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916321344537892,
+      "loss": 1.2828,
+      "step": 2756
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916258820512388,
+      "loss": 0.816,
+      "step": 2757
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916196273807624,
+      "loss": 0.6651,
+      "step": 2758
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002916133704424602,
+      "loss": 1.2438,
+      "step": 2759
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029160711123643243,
+      "loss": 0.4163,
+      "step": 2760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029160084976277933,
+      "loss": 3.6647,
+      "step": 2761
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029159458602160104,
+      "loss": 0.4261,
+      "step": 2762
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029158832001299805,
+      "loss": 0.4375,
+      "step": 2763
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029158205173707055,
+      "loss": 2.5335,
+      "step": 2764
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291575781193919,
+      "loss": 0.6596,
+      "step": 2765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029156950838364376,
+      "loss": 0.3776,
+      "step": 2766
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915632333063454,
+      "loss": 0.3683,
+      "step": 2767
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915569559621242,
+      "loss": 0.5658,
+      "step": 2768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029155067635108084,
+      "loss": 0.2712,
+      "step": 2769
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029154439447331585,
+      "loss": 0.0091,
+      "step": 2770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029153811032892984,
+      "loss": 1.256,
+      "step": 2771
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029153182391802336,
+      "loss": 0.6606,
+      "step": 2772
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915255352406972,
+      "loss": 0.1948,
+      "step": 2773
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915192442970519,
+      "loss": 0.5809,
+      "step": 2774
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029151295108718835,
+      "loss": 4.0452,
+      "step": 2775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002915066556112073,
+      "loss": 0.4954,
+      "step": 2776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029150035786920946,
+      "loss": 0.2425,
+      "step": 2777
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914940578612958,
+      "loss": 1.2172,
+      "step": 2778
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914877555875671,
+      "loss": 0.4861,
+      "step": 2779
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029148145104812437,
+      "loss": 2.7737,
+      "step": 2780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914751442430685,
+      "loss": 0.8931,
+      "step": 2781
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029146883517250057,
+      "loss": 0.256,
+      "step": 2782
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029146252383652156,
+      "loss": 1.1459,
+      "step": 2783
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029145621023523254,
+      "loss": 0.6312,
+      "step": 2784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029144989436873456,
+      "loss": 0.233,
+      "step": 2785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914435762371289,
+      "loss": 0.1031,
+      "step": 2786
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029143725584051655,
+      "loss": 0.2966,
+      "step": 2787
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914309331789989,
+      "loss": 0.555,
+      "step": 2788
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029142460825267705,
+      "loss": 0.9459,
+      "step": 2789
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914182810616523,
+      "loss": 0.5671,
+      "step": 2790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914119516060261,
+      "loss": 0.2625,
+      "step": 2791
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002914056198858997,
+      "loss": 0.3098,
+      "step": 2792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029139928590137453,
+      "loss": 0.6231,
+      "step": 2793
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291392949652552,
+      "loss": 0.408,
+      "step": 2794
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913866111395336,
+      "loss": 0.0048,
+      "step": 2795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029138027036242076,
+      "loss": 0.3608,
+      "step": 2796
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029137392732131516,
+      "loss": 0.2772,
+      "step": 2797
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913675820163182,
+      "loss": 0.3357,
+      "step": 2798
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913612344475316,
+      "loss": 0.4085,
+      "step": 2799
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291354884615057,
+      "loss": 0.1726,
+      "step": 2800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913485325189961,
+      "loss": 0.4893,
+      "step": 2801
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029134217815945056,
+      "loss": 0.5994,
+      "step": 2802
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913358215365221,
+      "loss": 1.0739,
+      "step": 2803
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029132946265031265,
+      "loss": 0.4155,
+      "step": 2804
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913231015009239,
+      "loss": 0.9725,
+      "step": 2805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002913167380884578,
+      "loss": 1.2451,
+      "step": 2806
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029131037241301625,
+      "loss": 5.6612,
+      "step": 2807
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029130400447470114,
+      "loss": 0.5568,
+      "step": 2808
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029129763427361447,
+      "loss": 0.1197,
+      "step": 2809
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029129126180985824,
+      "loss": 0.0079,
+      "step": 2810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002912848870835345,
+      "loss": 0.6988,
+      "step": 2811
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029127851009474537,
+      "loss": 0.525,
+      "step": 2812
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029127213084359293,
+      "loss": 0.7679,
+      "step": 2813
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029126574933017926,
+      "loss": 1.0434,
+      "step": 2814
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002912593655546067,
+      "loss": 0.583,
+      "step": 2815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002912529795169774,
+      "loss": 0.746,
+      "step": 2816
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029124659121739356,
+      "loss": 0.3334,
+      "step": 2817
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029124020065595757,
+      "loss": 0.7069,
+      "step": 2818
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029123380783277177,
+      "loss": 0.1506,
+      "step": 2819
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002912274127479385,
+      "loss": 0.3845,
+      "step": 2820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002912210154015601,
+      "loss": 0.3683,
+      "step": 2821
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029121461579373917,
+      "loss": 0.297,
+      "step": 2822
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291208213924578,
+      "loss": 0.3974,
+      "step": 2823
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029120180979417935,
+      "loss": 0.0225,
+      "step": 2824
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029119540340264555,
+      "loss": 0.7232,
+      "step": 2825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029118899475007924,
+      "loss": 0.322,
+      "step": 2826
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002911825838365831,
+      "loss": 0.1299,
+      "step": 2827
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002911761706622598,
+      "loss": 0.1465,
+      "step": 2828
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002911697552272119,
+      "loss": 0.8805,
+      "step": 2829
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029116333753154237,
+      "loss": 0.6294,
+      "step": 2830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029115691757535374,
+      "loss": 0.3931,
+      "step": 2831
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291150495358749,
+      "loss": 1.1673,
+      "step": 2832
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029114407088183086,
+      "loss": 0.3674,
+      "step": 2833
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002911376441447022,
+      "loss": 0.56,
+      "step": 2834
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002911312151474661,
+      "loss": 0.3035,
+      "step": 2835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002911247838902254,
+      "loss": 0.3793,
+      "step": 2836
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291118350373083,
+      "loss": 0.4079,
+      "step": 2837
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000291111914596142,
+      "loss": 0.5314,
+      "step": 2838
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029110547655950556,
+      "loss": 0.7743,
+      "step": 2839
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029109903626327663,
+      "loss": 1.8775,
+      "step": 2840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029109259370755844,
+      "loss": 0.0161,
+      "step": 2841
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029108614889245405,
+      "loss": 0.7485,
+      "step": 2842
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029107970181806674,
+      "loss": 0.5147,
+      "step": 2843
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029107325248449976,
+      "loss": 0.3187,
+      "step": 2844
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910668008918563,
+      "loss": 0.2234,
+      "step": 2845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910603470402398,
+      "loss": 0.0054,
+      "step": 2846
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910538909297536,
+      "loss": 1.0557,
+      "step": 2847
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910474325605009,
+      "loss": 1.633,
+      "step": 2848
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910409719325853,
+      "loss": 0.4538,
+      "step": 2849
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910345090461102,
+      "loss": 0.4249,
+      "step": 2850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910280439011791,
+      "loss": 0.2575,
+      "step": 2851
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029102157649789557,
+      "loss": 0.7226,
+      "step": 2852
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910151068363631,
+      "loss": 0.9339,
+      "step": 2853
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029100863491668535,
+      "loss": 0.2573,
+      "step": 2854
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002910021607389659,
+      "loss": 0.4022,
+      "step": 2855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029099568430330846,
+      "loss": 0.528,
+      "step": 2856
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909892056098168,
+      "loss": 0.581,
+      "step": 2857
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909827246585946,
+      "loss": 0.0095,
+      "step": 2858
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909762414497456,
+      "loss": 1.0283,
+      "step": 2859
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909697559833737,
+      "loss": 0.9226,
+      "step": 2860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909632682595827,
+      "loss": 0.3038,
+      "step": 2861
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029095677827847653,
+      "loss": 0.7273,
+      "step": 2862
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909502860401591,
+      "loss": 0.7953,
+      "step": 2863
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029094379154473443,
+      "loss": 0.4912,
+      "step": 2864
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002909372947923064,
+      "loss": 0.6562,
+      "step": 2865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00029093079578297907,
+      "loss": 0.7922,
+      "step": 2866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002909242945168566,
+      "loss": 1.1633,
+      "step": 2867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029091779099404307,
+      "loss": 0.7039,
+      "step": 2868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029091128521464263,
+      "loss": 0.7093,
+      "step": 2869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002909047771787594,
+      "loss": 0.9817,
+      "step": 2870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029089826688649765,
+      "loss": 0.7708,
+      "step": 2871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002908917543379616,
+      "loss": 0.692,
+      "step": 2872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029088523953325554,
+      "loss": 0.0064,
+      "step": 2873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002908787224724838,
+      "loss": 0.6329,
+      "step": 2874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029087220315575076,
+      "loss": 0.5718,
+      "step": 2875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029086568158316086,
+      "loss": 0.8213,
+      "step": 2876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002908591577548184,
+      "loss": 0.0086,
+      "step": 2877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000290852631670828,
+      "loss": 0.7785,
+      "step": 2878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029084610333129403,
+      "loss": 0.3129,
+      "step": 2879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029083957273632114,
+      "loss": 0.5478,
+      "step": 2880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002908330398860138,
+      "loss": 0.5136,
+      "step": 2881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029082650478047675,
+      "loss": 0.514,
+      "step": 2882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029081996741981455,
+      "loss": 0.1883,
+      "step": 2883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002908134278041319,
+      "loss": 0.8436,
+      "step": 2884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029080688593353357,
+      "loss": 0.0077,
+      "step": 2885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029080034180812425,
+      "loss": 0.2907,
+      "step": 2886
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002907937954280088,
+      "loss": 0.3741,
+      "step": 2887
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029078724679329205,
+      "loss": 0.4401,
+      "step": 2888
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029078069590407875,
+      "loss": 0.8538,
+      "step": 2889
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029077414276047395,
+      "loss": 0.4022,
+      "step": 2890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002907675873625825,
+      "loss": 0.9429,
+      "step": 2891
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002907610297105094,
+      "loss": 0.2936,
+      "step": 2892
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029075446980435964,
+      "loss": 0.823,
+      "step": 2893
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029074790764423836,
+      "loss": 0.598,
+      "step": 2894
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002907413432302505,
+      "loss": 1.1347,
+      "step": 2895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029073477656250127,
+      "loss": 1.4007,
+      "step": 2896
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029072820764109586,
+      "loss": 0.0052,
+      "step": 2897
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002907216364661393,
+      "loss": 0.5646,
+      "step": 2898
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000290715063037737,
+      "loss": 0.1755,
+      "step": 2899
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029070848735599414,
+      "loss": 0.1118,
+      "step": 2900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000290701909421016,
+      "loss": 0.5721,
+      "step": 2901
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000290695329232908,
+      "loss": 1.3955,
+      "step": 2902
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029068874679177544,
+      "loss": 0.898,
+      "step": 2903
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002906821620977237,
+      "loss": 0.3702,
+      "step": 2904
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029067557515085837,
+      "loss": 0.2938,
+      "step": 2905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029066898595128476,
+      "loss": 0.0051,
+      "step": 2906
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002906623944991084,
+      "loss": 0.5489,
+      "step": 2907
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029065580079443496,
+      "loss": 0.3766,
+      "step": 2908
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029064920483736997,
+      "loss": 0.0956,
+      "step": 2909
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029064260662801906,
+      "loss": 1.4928,
+      "step": 2910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002906360061664878,
+      "loss": 0.2605,
+      "step": 2911
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029062940345288204,
+      "loss": 0.5212,
+      "step": 2912
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029062279848730745,
+      "loss": 0.5279,
+      "step": 2913
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029061619126986973,
+      "loss": 0.7536,
+      "step": 2914
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002906095818006748,
+      "loss": 0.0069,
+      "step": 2915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029060297007982836,
+      "loss": 0.5163,
+      "step": 2916
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029059635610743646,
+      "loss": 0.6662,
+      "step": 2917
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029058973988360486,
+      "loss": 1.7453,
+      "step": 2918
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002905831214084396,
+      "loss": 0.0071,
+      "step": 2919
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002905765006820466,
+      "loss": 0.3121,
+      "step": 2920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000290569877704532,
+      "loss": 0.6735,
+      "step": 2921
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002905632524760017,
+      "loss": 0.2238,
+      "step": 2922
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002905566249965619,
+      "loss": 0.5351,
+      "step": 2923
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002905499952663187,
+      "loss": 0.3849,
+      "step": 2924
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002905433632853783,
+      "loss": 0.7099,
+      "step": 2925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029053672905384674,
+      "loss": 1.0698,
+      "step": 2926
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029053009257183046,
+      "loss": 0.5514,
+      "step": 2927
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029052345383943565,
+      "loss": 0.3332,
+      "step": 2928
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002905168128567686,
+      "loss": 0.6902,
+      "step": 2929
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029051016962393573,
+      "loss": 0.6623,
+      "step": 2930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029050352414104335,
+      "loss": 0.3268,
+      "step": 2931
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029049687640819786,
+      "loss": 0.5317,
+      "step": 2932
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029049022642550575,
+      "loss": 0.1906,
+      "step": 2933
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904835741930736,
+      "loss": 0.2501,
+      "step": 2934
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904769197110077,
+      "loss": 0.5151,
+      "step": 2935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029047026297941486,
+      "loss": 0.0083,
+      "step": 2936
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904636039984016,
+      "loss": 1.0653,
+      "step": 2937
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029045694276807446,
+      "loss": 0.1498,
+      "step": 2938
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904502792885402,
+      "loss": 0.6291,
+      "step": 2939
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029044361355990556,
+      "loss": 0.8857,
+      "step": 2940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904369455822771,
+      "loss": 0.673,
+      "step": 2941
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029043027535576184,
+      "loss": 0.0244,
+      "step": 2942
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904236028804664,
+      "loss": 0.9737,
+      "step": 2943
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904169281564977,
+      "loss": 0.5094,
+      "step": 2944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002904102511839627,
+      "loss": 0.015,
+      "step": 2945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029040357196296826,
+      "loss": 0.2281,
+      "step": 2946
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029039689049362123,
+      "loss": 0.968,
+      "step": 2947
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903902067760288,
+      "loss": 0.0025,
+      "step": 2948
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903835208102979,
+      "loss": 0.2425,
+      "step": 2949
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903768325965355,
+      "loss": 0.0125,
+      "step": 2950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903701421348489,
+      "loss": 0.3842,
+      "step": 2951
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903634494253451,
+      "loss": 0.4831,
+      "step": 2952
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029035675446813133,
+      "loss": 3.0789,
+      "step": 2953
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903500572633148,
+      "loss": 0.2437,
+      "step": 2954
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903433578110027,
+      "loss": 0.5762,
+      "step": 2955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029033665611130237,
+      "loss": 0.4102,
+      "step": 2956
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903299521643211,
+      "loss": 0.7511,
+      "step": 2957
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029032324597016626,
+      "loss": 2.2648,
+      "step": 2958
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903165375289452,
+      "loss": 0.3693,
+      "step": 2959
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029030982684076544,
+      "loss": 0.2735,
+      "step": 2960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002903031139057344,
+      "loss": 1.9174,
+      "step": 2961
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029029639872395945,
+      "loss": 0.3768,
+      "step": 2962
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902896812955483,
+      "loss": 0.4183,
+      "step": 2963
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902829616206084,
+      "loss": 0.4495,
+      "step": 2964
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902762396992475,
+      "loss": 1.867,
+      "step": 2965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029026951553157306,
+      "loss": 0.0752,
+      "step": 2966
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029026278911769285,
+      "loss": 1.1424,
+      "step": 2967
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902560604577146,
+      "loss": 0.3862,
+      "step": 2968
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000290249329551746,
+      "loss": 0.3463,
+      "step": 2969
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029024259639989493,
+      "loss": 0.2845,
+      "step": 2970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902358610022691,
+      "loss": 0.5284,
+      "step": 2971
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902291233589765,
+      "loss": 0.2335,
+      "step": 2972
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902223834701249,
+      "loss": 0.3516,
+      "step": 2973
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902156413358222,
+      "loss": 2.339,
+      "step": 2974
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002902088969561765,
+      "loss": 0.3963,
+      "step": 2975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029020215033129574,
+      "loss": 0.4147,
+      "step": 2976
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029019540146128797,
+      "loss": 0.7152,
+      "step": 2977
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901886503462612,
+      "loss": 1.9053,
+      "step": 2978
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901818969863237,
+      "loss": 0.389,
+      "step": 2979
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901751413815834,
+      "loss": 0.6216,
+      "step": 2980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029016838353214863,
+      "loss": 0.147,
+      "step": 2981
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901616234381275,
+      "loss": 1.0092,
+      "step": 2982
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901548610996284,
+      "loss": 0.7278,
+      "step": 2983
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901480965167595,
+      "loss": 0.711,
+      "step": 2984
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901413296896292,
+      "loss": 0.4619,
+      "step": 2985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029013456061834574,
+      "loss": 0.4105,
+      "step": 2986
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029012778930301774,
+      "loss": 0.3418,
+      "step": 2987
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901210157437534,
+      "loss": 0.5282,
+      "step": 2988
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901142399406613,
+      "loss": 0.8139,
+      "step": 2989
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029010746189384994,
+      "loss": 0.4609,
+      "step": 2990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002901006816034279,
+      "loss": 0.6258,
+      "step": 2991
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900938990695036,
+      "loss": 0.1288,
+      "step": 2992
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900871142921858,
+      "loss": 0.3169,
+      "step": 2993
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029008032727158317,
+      "loss": 0.1637,
+      "step": 2994
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900735380078043,
+      "loss": 0.4503,
+      "step": 2995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900667465009579,
+      "loss": 1.3429,
+      "step": 2996
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900599527511528,
+      "loss": 0.3106,
+      "step": 2997
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900531567584977,
+      "loss": 0.0013,
+      "step": 2998
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029004635852310153,
+      "loss": 0.6926,
+      "step": 2999
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029003955804507306,
+      "loss": 0.2336,
+      "step": 3000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029003275532452137,
+      "loss": 1.184,
+      "step": 3001
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029002595036155515,
+      "loss": 0.335,
+      "step": 3002
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900191431562835,
+      "loss": 0.3167,
+      "step": 3003
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002900123337088154,
+      "loss": 0.4289,
+      "step": 3004
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00029000552201925993,
+      "loss": 2.2371,
+      "step": 3005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899987080877262,
+      "loss": 1.559,
+      "step": 3006
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028999189191432313,
+      "loss": 0.4815,
+      "step": 3007
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028998507349916014,
+      "loss": 0.1133,
+      "step": 3008
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899782528423462,
+      "loss": 0.5336,
+      "step": 3009
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899714299439907,
+      "loss": 0.884,
+      "step": 3010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028996460480420274,
+      "loss": 0.0013,
+      "step": 3011
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028995777742309175,
+      "loss": 0.2799,
+      "step": 3012
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028995094780076693,
+      "loss": 0.742,
+      "step": 3013
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899441159373378,
+      "loss": 0.3783,
+      "step": 3014
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899372818329136,
+      "loss": 0.8425,
+      "step": 3015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899304454876039,
+      "loss": 2.6509,
+      "step": 3016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028992360690151817,
+      "loss": 0.6559,
+      "step": 3017
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899167660747657,
+      "loss": 1.2047,
+      "step": 3018
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899099230074564,
+      "loss": 0.2862,
+      "step": 3019
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002899030776996996,
+      "loss": 1.0497,
+      "step": 3020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898962301516049,
+      "loss": 2.3888,
+      "step": 3021
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898893803632821,
+      "loss": 0.5187,
+      "step": 3022
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898825283348408,
+      "loss": 0.0009,
+      "step": 3023
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028987567406639074,
+      "loss": 1.049,
+      "step": 3024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898688175580417,
+      "loss": 0.0268,
+      "step": 3025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898619588099034,
+      "loss": 0.9113,
+      "step": 3026
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898550978220858,
+      "loss": 0.5525,
+      "step": 3027
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898482345946987,
+      "loss": 0.9018,
+      "step": 3028
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000289841369127852,
+      "loss": 0.4968,
+      "step": 3029
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028983450142165565,
+      "loss": 0.1195,
+      "step": 3030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898276314762196,
+      "loss": 0.5656,
+      "step": 3031
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898207592916539,
+      "loss": 0.4641,
+      "step": 3032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028981388486806856,
+      "loss": 0.3192,
+      "step": 3033
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898070082055737,
+      "loss": 0.0203,
+      "step": 3034
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002898001293042794,
+      "loss": 1.1226,
+      "step": 3035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028979324816429586,
+      "loss": 0.5605,
+      "step": 3036
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028978636478573324,
+      "loss": 0.3611,
+      "step": 3037
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028977947916870183,
+      "loss": 0.9629,
+      "step": 3038
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002897725913133118,
+      "loss": 0.6379,
+      "step": 3039
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028976570121967346,
+      "loss": 1.027,
+      "step": 3040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028975880888789724,
+      "loss": 0.3077,
+      "step": 3041
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002897519143180934,
+      "loss": 0.5043,
+      "step": 3042
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002897450175103724,
+      "loss": 1.3095,
+      "step": 3043
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028973811846484465,
+      "loss": 1.034,
+      "step": 3044
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028973121718162066,
+      "loss": 0.8137,
+      "step": 3045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028972431366081097,
+      "loss": 0.6275,
+      "step": 3046
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028971740790252604,
+      "loss": 0.1889,
+      "step": 3047
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028971049990687657,
+      "loss": 1.1624,
+      "step": 3048
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002897035896739731,
+      "loss": 0.1554,
+      "step": 3049
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002896966772039263,
+      "loss": 0.2187,
+      "step": 3050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028968976249684684,
+      "loss": 6.6514,
+      "step": 3051
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028968284555284555,
+      "loss": 0.3444,
+      "step": 3052
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028967592637203305,
+      "loss": 0.2976,
+      "step": 3053
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028966900495452024,
+      "loss": 0.6929,
+      "step": 3054
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028966208130041794,
+      "loss": 0.4317,
+      "step": 3055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028965515540983695,
+      "loss": 0.2994,
+      "step": 3056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002896482272828883,
+      "loss": 2.25,
+      "step": 3057
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002896412969196829,
+      "loss": 0.3307,
+      "step": 3058
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028963436432033165,
+      "loss": 0.5437,
+      "step": 3059
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028962742948494564,
+      "loss": 0.0202,
+      "step": 3060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002896204924136359,
+      "loss": 0.001,
+      "step": 3061
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002896135531065135,
+      "loss": 0.4305,
+      "step": 3062
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002896066115636896,
+      "loss": 1.1666,
+      "step": 3063
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895996677852753,
+      "loss": 0.4174,
+      "step": 3064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028959272177138184,
+      "loss": 0.3312,
+      "step": 3065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028958577352212046,
+      "loss": 0.7167,
+      "step": 3066
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895788230376024,
+      "loss": 1.0622,
+      "step": 3067
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028957187031793894,
+      "loss": 3.5701,
+      "step": 3068
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895649153632415,
+      "loss": 0.8698,
+      "step": 3069
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028955795817362136,
+      "loss": 0.3839,
+      "step": 3070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895509987491899,
+      "loss": 0.0012,
+      "step": 3071
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028954403709005876,
+      "loss": 0.4878,
+      "step": 3072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895370731963392,
+      "loss": 1.0188,
+      "step": 3073
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895301070681429,
+      "loss": 0.5564,
+      "step": 3074
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895231387055813,
+      "loss": 0.5411,
+      "step": 3075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028951616810876603,
+      "loss": 0.4348,
+      "step": 3076
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002895091952778087,
+      "loss": 0.408,
+      "step": 3077
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028950222021282104,
+      "loss": 0.6566,
+      "step": 3078
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028949524291391464,
+      "loss": 0.9367,
+      "step": 3079
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894882633812013,
+      "loss": 5.1926,
+      "step": 3080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028948128161479275,
+      "loss": 0.2961,
+      "step": 3081
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028947429761480076,
+      "loss": 0.531,
+      "step": 3082
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894673113813373,
+      "loss": 0.6398,
+      "step": 3083
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894603229145141,
+      "loss": 0.0075,
+      "step": 3084
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028945333221444314,
+      "loss": 0.4314,
+      "step": 3085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028944633928123636,
+      "loss": 2.7501,
+      "step": 3086
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028943934411500576,
+      "loss": 1.1642,
+      "step": 3087
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894323467158633,
+      "loss": 1.1582,
+      "step": 3088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028942534708392107,
+      "loss": 2.1948,
+      "step": 3089
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894183452192912,
+      "loss": 0.8927,
+      "step": 3090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894113411220857,
+      "loss": 0.3329,
+      "step": 3091
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002894043347924168,
+      "loss": 1.1188,
+      "step": 3092
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028939732623039675,
+      "loss": 0.7112,
+      "step": 3093
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893903154361377,
+      "loss": 0.761,
+      "step": 3094
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028938330240975196,
+      "loss": 0.1155,
+      "step": 3095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028937628715135184,
+      "loss": 0.1082,
+      "step": 3096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893692696610496,
+      "loss": 0.0623,
+      "step": 3097
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893622499389577,
+      "loss": 0.0157,
+      "step": 3098
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893552279851885,
+      "loss": 0.6298,
+      "step": 3099
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028934820379985447,
+      "loss": 0.6629,
+      "step": 3100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028934117738306807,
+      "loss": 2.0375,
+      "step": 3101
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893341487349419,
+      "loss": 0.7951,
+      "step": 3102
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028932711785558844,
+      "loss": 0.4666,
+      "step": 3103
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893200847451202,
+      "loss": 0.0292,
+      "step": 3104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893130494036499,
+      "loss": 2.7496,
+      "step": 3105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002893060118312903,
+      "loss": 0.3631,
+      "step": 3106
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028929897202815394,
+      "loss": 0.8596,
+      "step": 3107
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002892919299943536,
+      "loss": 0.2715,
+      "step": 3108
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028928488573000204,
+      "loss": 0.5081,
+      "step": 3109
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002892778392352121,
+      "loss": 0.6111,
+      "step": 3110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028927079051009656,
+      "loss": 0.0088,
+      "step": 3111
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028926373955476834,
+      "loss": 0.5952,
+      "step": 3112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002892566863693403,
+      "loss": 0.1215,
+      "step": 3113
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028924963095392544,
+      "loss": 0.7766,
+      "step": 3114
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00028924257330863683,
+      "loss": 0.9469,
+      "step": 3115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028923551343358725,
+      "loss": 0.061,
+      "step": 3116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028922845132888994,
+      "loss": 0.4277,
+      "step": 3117
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002892213869946579,
+      "loss": 0.5685,
+      "step": 3118
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002892143204310043,
+      "loss": 0.8966,
+      "step": 3119
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002892072516380423,
+      "loss": 0.0575,
+      "step": 3120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028920018061588513,
+      "loss": 0.0124,
+      "step": 3121
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891931073646459,
+      "loss": 0.1106,
+      "step": 3122
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000289186031884438,
+      "loss": 0.6484,
+      "step": 3123
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891789541753747,
+      "loss": 0.7255,
+      "step": 3124
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891718742375693,
+      "loss": 0.4977,
+      "step": 3125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891647920711352,
+      "loss": 0.2894,
+      "step": 3126
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891577076761858,
+      "loss": 2.1231,
+      "step": 3127
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891506210528346,
+      "loss": 0.0032,
+      "step": 3128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000289143532201195,
+      "loss": 0.1613,
+      "step": 3129
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891364411213806,
+      "loss": 1.4024,
+      "step": 3130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891293478135049,
+      "loss": 0.2116,
+      "step": 3131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028912225227768144,
+      "loss": 0.4568,
+      "step": 3132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028911515451402397,
+      "loss": 0.7067,
+      "step": 3133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028910805452264605,
+      "loss": 0.4694,
+      "step": 3134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002891009523036614,
+      "loss": 0.7889,
+      "step": 3135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002890938478571838,
+      "loss": 0.1517,
+      "step": 3136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028908674118332693,
+      "loss": 0.5504,
+      "step": 3137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002890796322822047,
+      "loss": 0.5454,
+      "step": 3138
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028907252115393086,
+      "loss": 0.1451,
+      "step": 3139
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002890654077986193,
+      "loss": 0.3295,
+      "step": 3140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002890582922163839,
+      "loss": 0.4959,
+      "step": 3141
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028905117440733874,
+      "loss": 0.3678,
+      "step": 3142
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028904405437159763,
+      "loss": 0.8208,
+      "step": 3143
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002890369321092747,
+      "loss": 0.6408,
+      "step": 3144
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000289029807620484,
+      "loss": 0.4105,
+      "step": 3145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002890226809053395,
+      "loss": 0.2085,
+      "step": 3146
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002890155519639555,
+      "loss": 0.7684,
+      "step": 3147
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000289008420796446,
+      "loss": 0.9252,
+      "step": 3148
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028900128740292534,
+      "loss": 0.8446,
+      "step": 3149
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002889941517835076,
+      "loss": 0.7054,
+      "step": 3150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028898701393830715,
+      "loss": 0.4193,
+      "step": 3151
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028897987386743824,
+      "loss": 0.5485,
+      "step": 3152
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028897273157101527,
+      "loss": 0.0086,
+      "step": 3153
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028896558704915253,
+      "loss": 0.7933,
+      "step": 3154
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002889584403019645,
+      "loss": 0.5166,
+      "step": 3155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002889512913295656,
+      "loss": 0.1461,
+      "step": 3156
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002889441401320703,
+      "loss": 0.7665,
+      "step": 3157
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002889369867095931,
+      "loss": 0.02,
+      "step": 3158
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002889298310622486,
+      "loss": 1.2706,
+      "step": 3159
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028892267319015135,
+      "loss": 1.4492,
+      "step": 3160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288915513093416,
+      "loss": 0.3956,
+      "step": 3161
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002889083507721572,
+      "loss": 0.4451,
+      "step": 3162
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028890118622648957,
+      "loss": 0.4761,
+      "step": 3163
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028889401945652796,
+      "loss": 0.4134,
+      "step": 3164
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288886850462387,
+      "loss": 0.3646,
+      "step": 3165
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028887967924418165,
+      "loss": 0.2121,
+      "step": 3166
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002888725058020267,
+      "loss": 0.4993,
+      "step": 3167
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028886533013603685,
+      "loss": 0.5327,
+      "step": 3168
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028885815224632727,
+      "loss": 0.9227,
+      "step": 3169
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028885097213301276,
+      "loss": 0.2107,
+      "step": 3170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028884378979620825,
+      "loss": 0.3218,
+      "step": 3171
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002888366052360289,
+      "loss": 0.01,
+      "step": 3172
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028882941845258963,
+      "loss": 0.3596,
+      "step": 3173
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002888222294460056,
+      "loss": 0.1847,
+      "step": 3174
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002888150382163919,
+      "loss": 0.4068,
+      "step": 3175
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002888078447638637,
+      "loss": 2.1745,
+      "step": 3176
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002888006490885362,
+      "loss": 0.3407,
+      "step": 3177
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028879345119052464,
+      "loss": 0.4212,
+      "step": 3178
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887862510699442,
+      "loss": 0.429,
+      "step": 3179
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887790487269102,
+      "loss": 0.4952,
+      "step": 3180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028877184416153815,
+      "loss": 0.7496,
+      "step": 3181
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887646373739432,
+      "loss": 0.5347,
+      "step": 3182
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028875742836424083,
+      "loss": 0.4059,
+      "step": 3183
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887502171325465,
+      "loss": 0.1724,
+      "step": 3184
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887430036789757,
+      "loss": 0.4237,
+      "step": 3185
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887357880036439,
+      "loss": 0.5653,
+      "step": 3186
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028872857010666666,
+      "loss": 0.4832,
+      "step": 3187
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028872134998815963,
+      "loss": 1.4477,
+      "step": 3188
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887141276482383,
+      "loss": 0.6316,
+      "step": 3189
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002887069030870184,
+      "loss": 0.2408,
+      "step": 3190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028869967630461567,
+      "loss": 0.5575,
+      "step": 3191
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886924473011458,
+      "loss": 0.3371,
+      "step": 3192
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028868521607672444,
+      "loss": 0.1188,
+      "step": 3193
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028867798263146756,
+      "loss": 0.9227,
+      "step": 3194
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886707469654908,
+      "loss": 0.5838,
+      "step": 3195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028866350907891025,
+      "loss": 0.1321,
+      "step": 3196
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886562689718417,
+      "loss": 0.4928,
+      "step": 3197
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028864902664440113,
+      "loss": 0.5588,
+      "step": 3198
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028864178209670436,
+      "loss": 0.002,
+      "step": 3199
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886345353288676,
+      "loss": 3.4513,
+      "step": 3200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028862728634100685,
+      "loss": 0.5147,
+      "step": 3201
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886200351332381,
+      "loss": 0.2316,
+      "step": 3202
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028861278170567756,
+      "loss": 0.0005,
+      "step": 3203
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002886055260584413,
+      "loss": 0.2747,
+      "step": 3204
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028859826819164564,
+      "loss": 0.0126,
+      "step": 3205
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028859100810540663,
+      "loss": 3.129,
+      "step": 3206
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885837457998407,
+      "loss": 0.7262,
+      "step": 3207
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288576481275064,
+      "loss": 2.1168,
+      "step": 3208
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028856921453119296,
+      "loss": 2.2462,
+      "step": 3209
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028856194556834395,
+      "loss": 0.2795,
+      "step": 3210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885546743866333,
+      "loss": 0.3482,
+      "step": 3211
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885474009861775,
+      "loss": 1.6671,
+      "step": 3212
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028854012536709295,
+      "loss": 1.7521,
+      "step": 3213
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885328475294962,
+      "loss": 0.2476,
+      "step": 3214
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885255674735039,
+      "loss": 0.2298,
+      "step": 3215
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885182851992325,
+      "loss": 0.502,
+      "step": 3216
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002885110007067986,
+      "loss": 0.8081,
+      "step": 3217
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028850371399631895,
+      "loss": 0.0252,
+      "step": 3218
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028849642506791013,
+      "loss": 0.481,
+      "step": 3219
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028848913392168896,
+      "loss": 0.9902,
+      "step": 3220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884818405577722,
+      "loss": 1.0947,
+      "step": 3221
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884745449762765,
+      "loss": 0.3267,
+      "step": 3222
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028846724717731886,
+      "loss": 0.8801,
+      "step": 3223
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028845994716101604,
+      "loss": 0.2168,
+      "step": 3224
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884526449274849,
+      "loss": 0.7358,
+      "step": 3225
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028844534047684254,
+      "loss": 1.3145,
+      "step": 3226
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884380338092058,
+      "loss": 0.9678,
+      "step": 3227
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028843072492469173,
+      "loss": 0.4736,
+      "step": 3228
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884234138234173,
+      "loss": 0.2361,
+      "step": 3229
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884161005054997,
+      "loss": 0.4887,
+      "step": 3230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884087849710559,
+      "loss": 0.2936,
+      "step": 3231
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002884014672202032,
+      "loss": 0.767,
+      "step": 3232
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028839414725305874,
+      "loss": 0.6175,
+      "step": 3233
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028838682506973965,
+      "loss": 0.3445,
+      "step": 3234
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002883795006703633,
+      "loss": 0.1338,
+      "step": 3235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028837217405504686,
+      "loss": 0.5931,
+      "step": 3236
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028836484522390773,
+      "loss": 0.6628,
+      "step": 3237
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002883575141770633,
+      "loss": 0.4431,
+      "step": 3238
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002883501809146309,
+      "loss": 0.9977,
+      "step": 3239
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028834284543672803,
+      "loss": 1.4095,
+      "step": 3240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002883355077434721,
+      "loss": 0.1119,
+      "step": 3241
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002883281678349806,
+      "loss": 1.3921,
+      "step": 3242
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002883208257113711,
+      "loss": 2.2476,
+      "step": 3243
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028831348137276115,
+      "loss": 0.5313,
+      "step": 3244
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002883061348192684,
+      "loss": 0.8864,
+      "step": 3245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028829878605101045,
+      "loss": 0.1949,
+      "step": 3246
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000288291435068105,
+      "loss": 0.3381,
+      "step": 3247
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882840818706698,
+      "loss": 0.7211,
+      "step": 3248
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028827672645882253,
+      "loss": 0.2927,
+      "step": 3249
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028826936883268105,
+      "loss": 0.5008,
+      "step": 3250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028826200899236313,
+      "loss": 0.3796,
+      "step": 3251
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028825464693798663,
+      "loss": 0.5459,
+      "step": 3252
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882472826696694,
+      "loss": 0.3055,
+      "step": 3253
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028823991618752946,
+      "loss": 0.5042,
+      "step": 3254
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882325474916847,
+      "loss": 0.9706,
+      "step": 3255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002882251765822532,
+      "loss": 0.6708,
+      "step": 3256
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028821780345935293,
+      "loss": 0.564,
+      "step": 3257
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028821042812310197,
+      "loss": 0.2427,
+      "step": 3258
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028820305057361843,
+      "loss": 0.3675,
+      "step": 3259
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028819567081102043,
+      "loss": 0.3778,
+      "step": 3260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028818828883542616,
+      "loss": 0.4025,
+      "step": 3261
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881809046469539,
+      "loss": 0.5742,
+      "step": 3262
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028817351824572175,
+      "loss": 0.519,
+      "step": 3263
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028816612963184814,
+      "loss": 0.3622,
+      "step": 3264
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881587388054513,
+      "loss": 0.1245,
+      "step": 3265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881513457666496,
+      "loss": 0.4583,
+      "step": 3266
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881439505155614,
+      "loss": 0.6157,
+      "step": 3267
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881365530523052,
+      "loss": 1.5013,
+      "step": 3268
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028812915337699945,
+      "loss": 0.4184,
+      "step": 3269
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881217514897625,
+      "loss": 0.064,
+      "step": 3270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881143473907131,
+      "loss": 0.166,
+      "step": 3271
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002881069410799697,
+      "loss": 0.4742,
+      "step": 3272
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028809953255765086,
+      "loss": 0.4448,
+      "step": 3273
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028809212182387525,
+      "loss": 1.1866,
+      "step": 3274
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880847088787616,
+      "loss": 0.55,
+      "step": 3275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028807729372242857,
+      "loss": 1.667,
+      "step": 3276
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880698763549949,
+      "loss": 0.669,
+      "step": 3277
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028806245677657935,
+      "loss": 0.6921,
+      "step": 3278
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028805503498730077,
+      "loss": 0.4814,
+      "step": 3279
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028804761098727807,
+      "loss": 0.067,
+      "step": 3280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028804018477662996,
+      "loss": 0.2714,
+      "step": 3281
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880327563554755,
+      "loss": 0.0265,
+      "step": 3282
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880253257239336,
+      "loss": 0.3783,
+      "step": 3283
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880178928821232,
+      "loss": 0.0872,
+      "step": 3284
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002880104578301635,
+      "loss": 0.8803,
+      "step": 3285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028800302056817336,
+      "loss": 0.561,
+      "step": 3286
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028799558109627194,
+      "loss": 0.5218,
+      "step": 3287
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028798813941457844,
+      "loss": 0.2249,
+      "step": 3288
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000287980695523212,
+      "loss": 0.0717,
+      "step": 3289
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028797324942229176,
+      "loss": 2.4117,
+      "step": 3290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000287965801111937,
+      "loss": 0.3477,
+      "step": 3291
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000287958350592267,
+      "loss": 1.4316,
+      "step": 3292
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028795089786340107,
+      "loss": 0.2983,
+      "step": 3293
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028794344292545855,
+      "loss": 0.2698,
+      "step": 3294
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879359857785588,
+      "loss": 0.0191,
+      "step": 3295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879285264228212,
+      "loss": 0.0139,
+      "step": 3296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879210648583654,
+      "loss": 0.7425,
+      "step": 3297
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879136010853106,
+      "loss": 0.0274,
+      "step": 3298
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002879061351037765,
+      "loss": 0.23,
+      "step": 3299
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878986669138826,
+      "loss": 0.7008,
+      "step": 3300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878911965157485,
+      "loss": 0.3203,
+      "step": 3301
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878837239094939,
+      "loss": 0.8218,
+      "step": 3302
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878762490952383,
+      "loss": 0.477,
+      "step": 3303
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028786877207310157,
+      "loss": 1.3941,
+      "step": 3304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028786129284320334,
+      "loss": 0.6598,
+      "step": 3305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028785381140566344,
+      "loss": 0.2398,
+      "step": 3306
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878463277606016,
+      "loss": 0.5775,
+      "step": 3307
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878388419081377,
+      "loss": 2.4318,
+      "step": 3308
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878313538483916,
+      "loss": 0.5031,
+      "step": 3309
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028782386358148327,
+      "loss": 0.7223,
+      "step": 3310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878163711075326,
+      "loss": 0.306,
+      "step": 3311
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028780887642665956,
+      "loss": 0.8527,
+      "step": 3312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002878013795389842,
+      "loss": 0.4233,
+      "step": 3313
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028779388044462655,
+      "loss": 0.6732,
+      "step": 3314
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877863791437067,
+      "loss": 0.2712,
+      "step": 3315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028777887563634477,
+      "loss": 0.5593,
+      "step": 3316
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028777136992266095,
+      "loss": 0.929,
+      "step": 3317
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877638620027754,
+      "loss": 0.8708,
+      "step": 3318
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877563518768083,
+      "loss": 0.2016,
+      "step": 3319
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028774883954488006,
+      "loss": 0.0155,
+      "step": 3320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028774132500711085,
+      "loss": 0.2918,
+      "step": 3321
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000287733808263621,
+      "loss": 0.332,
+      "step": 3322
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028772628931453097,
+      "loss": 0.0269,
+      "step": 3323
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877187681599611,
+      "loss": 0.5706,
+      "step": 3324
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877112448000318,
+      "loss": 0.4927,
+      "step": 3325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002877037192348636,
+      "loss": 0.0095,
+      "step": 3326
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876961914645771,
+      "loss": 0.0144,
+      "step": 3327
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876886614892927,
+      "loss": 1.3695,
+      "step": 3328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876811293091309,
+      "loss": 0.5251,
+      "step": 3329
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876735949242126,
+      "loss": 0.5272,
+      "step": 3330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876660583346583,
+      "loss": 0.8878,
+      "step": 3331
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876585195405886,
+      "loss": 0.4554,
+      "step": 3332
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876509785421243,
+      "loss": 0.3405,
+      "step": 3333
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028764343533938625,
+      "loss": 0.1924,
+      "step": 3334
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028763588993249505,
+      "loss": 0.0155,
+      "step": 3335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028762834232157176,
+      "loss": 0.3429,
+      "step": 3336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028762079250673703,
+      "loss": 0.7408,
+      "step": 3337
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028761324048811183,
+      "loss": 0.0024,
+      "step": 3338
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002876056862658172,
+      "loss": 1.6212,
+      "step": 3339
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028759812983997397,
+      "loss": 0.3311,
+      "step": 3340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875905712107032,
+      "loss": 0.5793,
+      "step": 3341
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028758301037812593,
+      "loss": 0.3985,
+      "step": 3342
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875754473423632,
+      "loss": 0.5095,
+      "step": 3343
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028756788210353623,
+      "loss": 0.578,
+      "step": 3344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000287560314661766,
+      "loss": 2.0177,
+      "step": 3345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875527450171738,
+      "loss": 1.614,
+      "step": 3346
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875451731698809,
+      "loss": 0.2675,
+      "step": 3347
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875375991200084,
+      "loss": 1.0604,
+      "step": 3348
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002875300228676777,
+      "loss": 0.1997,
+      "step": 3349
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028752244441301004,
+      "loss": 0.248,
+      "step": 3350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028751486375612684,
+      "loss": 0.4078,
+      "step": 3351
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028750728089714956,
+      "loss": 0.0068,
+      "step": 3352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028749969583619946,
+      "loss": 0.8588,
+      "step": 3353
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874921085733981,
+      "loss": 0.0444,
+      "step": 3354
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000287484519108867,
+      "loss": 0.1252,
+      "step": 3355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028747692744272765,
+      "loss": 0.6066,
+      "step": 3356
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874693335751016,
+      "loss": 0.0231,
+      "step": 3357
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874617375061105,
+      "loss": 1.9157,
+      "step": 3358
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000287454139235876,
+      "loss": 1.3243,
+      "step": 3359
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874465387645197,
+      "loss": 0.5614,
+      "step": 3360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028743893609216347,
+      "loss": 0.012,
+      "step": 3361
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028743133121892885,
+      "loss": 0.733,
+      "step": 3362
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00028742372414493773,
+      "loss": 2.0274,
+      "step": 3363
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874161148703119,
+      "loss": 4.109,
+      "step": 3364
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002874085033951733,
+      "loss": 0.3453,
+      "step": 3365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028740088971964366,
+      "loss": 0.0025,
+      "step": 3366
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000287393273843845,
+      "loss": 0.5288,
+      "step": 3367
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028738565576789927,
+      "loss": 0.4556,
+      "step": 3368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028737803549192843,
+      "loss": 1.181,
+      "step": 3369
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002873704130160545,
+      "loss": 0.0452,
+      "step": 3370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002873627883403996,
+      "loss": 0.2214,
+      "step": 3371
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002873551614650858,
+      "loss": 0.5592,
+      "step": 3372
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002873475323902353,
+      "loss": 0.0044,
+      "step": 3373
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028733990111597015,
+      "loss": 2.9592,
+      "step": 3374
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002873322676424125,
+      "loss": 0.2757,
+      "step": 3375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028732463196968485,
+      "loss": 0.4214,
+      "step": 3376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002873169940979092,
+      "loss": 0.7574,
+      "step": 3377
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028730935402720807,
+      "loss": 0.4462,
+      "step": 3378
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002873017117577036,
+      "loss": 0.3293,
+      "step": 3379
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028729406728951836,
+      "loss": 0.4389,
+      "step": 3380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872864206227746,
+      "loss": 0.6759,
+      "step": 3381
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872787717575949,
+      "loss": 0.9859,
+      "step": 3382
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872711206941017,
+      "loss": 0.0189,
+      "step": 3383
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872634674324176,
+      "loss": 0.2962,
+      "step": 3384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872558119726649,
+      "loss": 0.6803,
+      "step": 3385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872481543149665,
+      "loss": 0.8611,
+      "step": 3386
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872404944594448,
+      "loss": 0.4107,
+      "step": 3387
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872328324062226,
+      "loss": 0.2605,
+      "step": 3388
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872251681554226,
+      "loss": 0.9027,
+      "step": 3389
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872175017071674,
+      "loss": 0.0042,
+      "step": 3390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002872098330615798,
+      "loss": 0.0042,
+      "step": 3391
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028720216221878275,
+      "loss": 0.3481,
+      "step": 3392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028719448917889894,
+      "loss": 1.3039,
+      "step": 3393
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028718681394205133,
+      "loss": 0.5327,
+      "step": 3394
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002871791365083627,
+      "loss": 0.0213,
+      "step": 3395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002871714568779561,
+      "loss": 0.2536,
+      "step": 3396
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002871637750509545,
+      "loss": 0.7808,
+      "step": 3397
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002871560910274809,
+      "loss": 0.0023,
+      "step": 3398
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028714840480765835,
+      "loss": 0.1971,
+      "step": 3399
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028714071639160995,
+      "loss": 0.6374,
+      "step": 3400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028713302577945877,
+      "loss": 0.3377,
+      "step": 3401
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028712533297132797,
+      "loss": 0.0112,
+      "step": 3402
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002871176379673408,
+      "loss": 0.3465,
+      "step": 3403
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002871099407676204,
+      "loss": 0.0117,
+      "step": 3404
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002871022413722901,
+      "loss": 2.3986,
+      "step": 3405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028709453978147315,
+      "loss": 0.0197,
+      "step": 3406
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002870868359952929,
+      "loss": 1.1365,
+      "step": 3407
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028707913001387275,
+      "loss": 0.254,
+      "step": 3408
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000287071421837336,
+      "loss": 0.009,
+      "step": 3409
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028706371146580614,
+      "loss": 0.6086,
+      "step": 3410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002870559988994067,
+      "loss": 0.3411,
+      "step": 3411
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002870482841382611,
+      "loss": 0.2377,
+      "step": 3412
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002870405671824929,
+      "loss": 3.1681,
+      "step": 3413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002870328480322257,
+      "loss": 0.4563,
+      "step": 3414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028702512668758306,
+      "loss": 0.2908,
+      "step": 3415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028701740314868873,
+      "loss": 0.4455,
+      "step": 3416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002870096774156663,
+      "loss": 0.3911,
+      "step": 3417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028700194948863946,
+      "loss": 0.336,
+      "step": 3418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002869942193677321,
+      "loss": 0.3945,
+      "step": 3419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028698648705306783,
+      "loss": 0.8353,
+      "step": 3420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028697875254477055,
+      "loss": 0.5819,
+      "step": 3421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002869710158429642,
+      "loss": 0.336,
+      "step": 3422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028696327694777255,
+      "loss": 0.8992,
+      "step": 3423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028695553585931957,
+      "loss": 0.5391,
+      "step": 3424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028694779257772927,
+      "loss": 0.549,
+      "step": 3425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028694004710312556,
+      "loss": 0.5265,
+      "step": 3426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028693229943563254,
+      "loss": 0.4261,
+      "step": 3427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028692454957537423,
+      "loss": 0.3091,
+      "step": 3428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002869167975224747,
+      "loss": 0.789,
+      "step": 3429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002869090432770583,
+      "loss": 0.2855,
+      "step": 3430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002869012868392489,
+      "loss": 2.3675,
+      "step": 3431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028689352820917093,
+      "loss": 0.1859,
+      "step": 3432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002868857673869485,
+      "loss": 0.3175,
+      "step": 3433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028687800437270593,
+      "loss": 2.3094,
+      "step": 3434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028687023916656757,
+      "loss": 0.0125,
+      "step": 3435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028686247176865776,
+      "loss": 0.9028,
+      "step": 3436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028685470217910085,
+      "loss": 0.2157,
+      "step": 3437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028684693039802126,
+      "loss": 0.0076,
+      "step": 3438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002868391564255435,
+      "loss": 0.4915,
+      "step": 3439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028683138026179203,
+      "loss": 0.3953,
+      "step": 3440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002868236019068913,
+      "loss": 0.5489,
+      "step": 3441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028681582136096595,
+      "loss": 0.7711,
+      "step": 3442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028680803862414056,
+      "loss": 0.5793,
+      "step": 3443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002868002536965397,
+      "loss": 0.7162,
+      "step": 3444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028679246657828817,
+      "loss": 1.7203,
+      "step": 3445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028678467726951054,
+      "loss": 1.0123,
+      "step": 3446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028677688577033163,
+      "loss": 0.6052,
+      "step": 3447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028676909208087616,
+      "loss": 1.3425,
+      "step": 3448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002867612962012689,
+      "loss": 0.3425,
+      "step": 3449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028675349813163473,
+      "loss": 0.825,
+      "step": 3450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002867456978720985,
+      "loss": 0.7634,
+      "step": 3451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028673789542278525,
+      "loss": 0.3686,
+      "step": 3452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028673009078381977,
+      "loss": 1.0143,
+      "step": 3453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286722283955327,
+      "loss": 0.2612,
+      "step": 3454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002867144749374322,
+      "loss": 0.7032,
+      "step": 3455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028670666373026017,
+      "loss": 0.5906,
+      "step": 3456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002866988503339361,
+      "loss": 0.3826,
+      "step": 3457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002866910347485851,
+      "loss": 0.3805,
+      "step": 3458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002866832169743323,
+      "loss": 0.6455,
+      "step": 3459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028667539701130297,
+      "loss": 3.9404,
+      "step": 3460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028666757485962225,
+      "loss": 1.531,
+      "step": 3461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002866597505194154,
+      "loss": 0.2142,
+      "step": 3462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028665192399080777,
+      "loss": 0.7215,
+      "step": 3463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028664409527392467,
+      "loss": 0.493,
+      "step": 3464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028663626436889147,
+      "loss": 0.3323,
+      "step": 3465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002866284312758335,
+      "loss": 0.0069,
+      "step": 3466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002866205959948763,
+      "loss": 0.5897,
+      "step": 3467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028661275852614527,
+      "loss": 0.4736,
+      "step": 3468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028660491886976596,
+      "loss": 0.4896,
+      "step": 3469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002865970770258639,
+      "loss": 1.2272,
+      "step": 3470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028658923299456463,
+      "loss": 0.3456,
+      "step": 3471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002865813867759938,
+      "loss": 0.7517,
+      "step": 3472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028657353837027705,
+      "loss": 0.6229,
+      "step": 3473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028656568777754,
+      "loss": 0.5297,
+      "step": 3474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028655783499790844,
+      "loss": 0.4587,
+      "step": 3475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028654998003150813,
+      "loss": 1.8559,
+      "step": 3476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002865421228784648,
+      "loss": 0.6183,
+      "step": 3477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002865342635389043,
+      "loss": 0.912,
+      "step": 3478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002865264020129525,
+      "loss": 0.4947,
+      "step": 3479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028651853830073524,
+      "loss": 0.8186,
+      "step": 3480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002865106724023784,
+      "loss": 0.3694,
+      "step": 3481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028650280431800814,
+      "loss": 0.2892,
+      "step": 3482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864949340477503,
+      "loss": 4.106,
+      "step": 3483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028648706159173094,
+      "loss": 0.3842,
+      "step": 3484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864791869500761,
+      "loss": 0.2068,
+      "step": 3485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864713101229119,
+      "loss": 0.4897,
+      "step": 3486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864634311103645,
+      "loss": 3.6938,
+      "step": 3487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028645554991256006,
+      "loss": 0.0163,
+      "step": 3488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864476665296248,
+      "loss": 1.2725,
+      "step": 3489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864397809616849,
+      "loss": 0.6327,
+      "step": 3490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864318932088667,
+      "loss": 1.254,
+      "step": 3491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028642400327129645,
+      "loss": 0.2348,
+      "step": 3492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028641611114910056,
+      "loss": 0.4009,
+      "step": 3493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002864082168424054,
+      "loss": 0.2209,
+      "step": 3494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028640032035133733,
+      "loss": 0.3194,
+      "step": 3495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863924216760229,
+      "loss": 0.8821,
+      "step": 3496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028638452081658847,
+      "loss": 0.3167,
+      "step": 3497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863766177731606,
+      "loss": 1.3509,
+      "step": 3498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028636871254586596,
+      "loss": 1.2844,
+      "step": 3499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286360805134831,
+      "loss": 0.8201,
+      "step": 3500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028635289554018235,
+      "loss": 0.3622,
+      "step": 3501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863449837620468,
+      "loss": 0.2047,
+      "step": 3502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863370698005509,
+      "loss": 1.0408,
+      "step": 3503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028632915365582145,
+      "loss": 0.6501,
+      "step": 3504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028632123532798526,
+      "loss": 0.0761,
+      "step": 3505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286313314817169,
+      "loss": 0.3956,
+      "step": 3506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002863053921234996,
+      "loss": 0.3704,
+      "step": 3507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028629746724710394,
+      "loss": 0.4581,
+      "step": 3508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002862895401881089,
+      "loss": 0.2292,
+      "step": 3509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028628161094664136,
+      "loss": 0.4931,
+      "step": 3510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002862736795228284,
+      "loss": 0.5352,
+      "step": 3511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000286265745916797,
+      "loss": 0.019,
+      "step": 3512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028625781012867415,
+      "loss": 0.9202,
+      "step": 3513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028624987215858697,
+      "loss": 0.6549,
+      "step": 3514
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028624193200666256,
+      "loss": 3.2373,
+      "step": 3515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002862339896730281,
+      "loss": 0.8273,
+      "step": 3516
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002862260451578107,
+      "loss": 0.538,
+      "step": 3517
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002862180984611377,
+      "loss": 0.1792,
+      "step": 3518
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028621014958313625,
+      "loss": 1.946,
+      "step": 3519
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028620219852393367,
+      "loss": 1.4633,
+      "step": 3520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861942452836573,
+      "loss": 0.5146,
+      "step": 3521
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861862898624345,
+      "loss": 1.1917,
+      "step": 3522
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861783322603927,
+      "loss": 0.63,
+      "step": 3523
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861703724776592,
+      "loss": 0.126,
+      "step": 3524
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861624105143616,
+      "loss": 0.7556,
+      "step": 3525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861544463706273,
+      "loss": 0.4045,
+      "step": 3526
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028614648004658395,
+      "loss": 0.2249,
+      "step": 3527
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028613851154235905,
+      "loss": 0.5386,
+      "step": 3528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028613054085808016,
+      "loss": 0.7869,
+      "step": 3529
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028612256799387495,
+      "loss": 0.5024,
+      "step": 3530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028611459294987113,
+      "loss": 0.8106,
+      "step": 3531
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002861066157261964,
+      "loss": 0.6646,
+      "step": 3532
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028609863632297845,
+      "loss": 0.5082,
+      "step": 3533
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028609065474034513,
+      "loss": 0.6754,
+      "step": 3534
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028608267097842414,
+      "loss": 0.5459,
+      "step": 3535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002860746850373435,
+      "loss": 0.2801,
+      "step": 3536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002860666969172309,
+      "loss": 0.3741,
+      "step": 3537
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028605870661821447,
+      "loss": 0.4803,
+      "step": 3538
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028605071414042197,
+      "loss": 0.3651,
+      "step": 3539
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028604271948398153,
+      "loss": 0.8864,
+      "step": 3540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028603472264902103,
+      "loss": 0.342,
+      "step": 3541
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002860267236356687,
+      "loss": 0.5878,
+      "step": 3542
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028601872244405247,
+      "loss": 0.5592,
+      "step": 3543
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028601071907430055,
+      "loss": 0.4243,
+      "step": 3544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028600271352654106,
+      "loss": 4.0689,
+      "step": 3545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859947058009023,
+      "loss": 0.8625,
+      "step": 3546
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028598669589751237,
+      "loss": 0.3842,
+      "step": 3547
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859786838164996,
+      "loss": 0.2096,
+      "step": 3548
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028597066955799225,
+      "loss": 0.4136,
+      "step": 3549
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028596265312211874,
+      "loss": 0.6745,
+      "step": 3550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859546345090074,
+      "loss": 0.1574,
+      "step": 3551
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859466137187866,
+      "loss": 0.6172,
+      "step": 3552
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859385907515848,
+      "loss": 0.5135,
+      "step": 3553
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028593056560753053,
+      "loss": 0.7359,
+      "step": 3554
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859225382867522,
+      "loss": 0.3809,
+      "step": 3555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859145087893784,
+      "loss": 0.4328,
+      "step": 3556
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002859064771155378,
+      "loss": 1.2179,
+      "step": 3557
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028589844326535886,
+      "loss": 0.9037,
+      "step": 3558
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028589040723897037,
+      "loss": 0.3571,
+      "step": 3559
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002858823690365009,
+      "loss": 0.2284,
+      "step": 3560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028587432865807925,
+      "loss": 0.938,
+      "step": 3561
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002858662861038341,
+      "loss": 0.1851,
+      "step": 3562
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002858582413738944,
+      "loss": 2.1646,
+      "step": 3563
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028585019446838875,
+      "loss": 2.0679,
+      "step": 3564
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002858421453874462,
+      "loss": 0.6205,
+      "step": 3565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028583409413119546,
+      "loss": 0.9392,
+      "step": 3566
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002858260406997656,
+      "loss": 0.1014,
+      "step": 3567
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028581798509328554,
+      "loss": 0.4444,
+      "step": 3568
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028580992731188433,
+      "loss": 0.1879,
+      "step": 3569
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028580186735569093,
+      "loss": 0.8931,
+      "step": 3570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857938052248344,
+      "loss": 0.0751,
+      "step": 3571
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857857409194439,
+      "loss": 0.6503,
+      "step": 3572
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857776744396486,
+      "loss": 0.9782,
+      "step": 3573
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857696057855775,
+      "loss": 0.0689,
+      "step": 3574
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028576153495736003,
+      "loss": 0.2574,
+      "step": 3575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857534619551253,
+      "loss": 0.8925,
+      "step": 3576
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857453867790026,
+      "loss": 0.3454,
+      "step": 3577
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028573730942912126,
+      "loss": 0.3887,
+      "step": 3578
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028572922990561064,
+      "loss": 0.4318,
+      "step": 3579
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002857211482086001,
+      "loss": 0.9379,
+      "step": 3580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000285713064338219,
+      "loss": 0.9601,
+      "step": 3581
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028570497829459693,
+      "loss": 0.5525,
+      "step": 3582
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856968900778633,
+      "loss": 0.7546,
+      "step": 3583
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028568879968814755,
+      "loss": 2.0454,
+      "step": 3584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028568070712557935,
+      "loss": 0.4037,
+      "step": 3585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028567261239028827,
+      "loss": 0.5389,
+      "step": 3586
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856645154824039,
+      "loss": 0.506,
+      "step": 3587
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856564164020559,
+      "loss": 0.4321,
+      "step": 3588
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028564831514937405,
+      "loss": 0.6231,
+      "step": 3589
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028564021172448797,
+      "loss": 0.9701,
+      "step": 3590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028563210612752747,
+      "loss": 0.4565,
+      "step": 3591
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856239983586224,
+      "loss": 0.1245,
+      "step": 3592
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856158884179024,
+      "loss": 0.377,
+      "step": 3593
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002856077763054976,
+      "loss": 0.9277,
+      "step": 3594
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028559966202153773,
+      "loss": 0.4197,
+      "step": 3595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855915455661528,
+      "loss": 0.6178,
+      "step": 3596
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028558342693947277,
+      "loss": 0.008,
+      "step": 3597
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855753061416276,
+      "loss": 1.759,
+      "step": 3598
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855671831727474,
+      "loss": 0.4555,
+      "step": 3599
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028555905803296223,
+      "loss": 1.057,
+      "step": 3600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028555093072240217,
+      "loss": 0.4822,
+      "step": 3601
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855428012411974,
+      "loss": 0.3562,
+      "step": 3602
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028553466958947815,
+      "loss": 0.3969,
+      "step": 3603
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028552653576737445,
+      "loss": 1.1281,
+      "step": 3604
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028551839977501676,
+      "loss": 0.0119,
+      "step": 3605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855102616125353,
+      "loss": 0.4535,
+      "step": 3606
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002855021212800603,
+      "loss": 0.6482,
+      "step": 3607
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002854939787777222,
+      "loss": 0.2514,
+      "step": 3608
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028548583410565145,
+      "loss": 0.4342,
+      "step": 3609
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028547768726397833,
+      "loss": 0.2888,
+      "step": 3610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028546953825283336,
+      "loss": 0.0227,
+      "step": 3611
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002854613870723471,
+      "loss": 0.2697,
+      "step": 3612
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028545323372265,
+      "loss": 0.2405,
+      "step": 3613
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00028544507820387263,
+      "loss": 0.6486,
+      "step": 3614
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028543692051614564,
+      "loss": 0.8433,
+      "step": 3615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028542876065959966,
+      "loss": 0.3777,
+      "step": 3616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028542059863436527,
+      "loss": 0.0132,
+      "step": 3617
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002854124344405732,
+      "loss": 0.522,
+      "step": 3618
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028540426807835434,
+      "loss": 0.7425,
+      "step": 3619
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028539609954783925,
+      "loss": 0.367,
+      "step": 3620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028538792884915884,
+      "loss": 0.6717,
+      "step": 3621
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028537975598244396,
+      "loss": 0.5022,
+      "step": 3622
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028537158094782547,
+      "loss": 1.7339,
+      "step": 3623
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028536340374543423,
+      "loss": 1.2409,
+      "step": 3624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028535522437540123,
+      "loss": 0.6108,
+      "step": 3625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002853470428378575,
+      "loss": 0.3327,
+      "step": 3626
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028533885913293395,
+      "loss": 0.4287,
+      "step": 3627
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028533067326076174,
+      "loss": 0.306,
+      "step": 3628
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028532248522147185,
+      "loss": 0.6526,
+      "step": 3629
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002853142950151955,
+      "loss": 0.9243,
+      "step": 3630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002853061026420638,
+      "loss": 1.0839,
+      "step": 3631
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028529790810220785,
+      "loss": 0.3178,
+      "step": 3632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000285289711395759,
+      "loss": 0.4686,
+      "step": 3633
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002852815125228485,
+      "loss": 0.3977,
+      "step": 3634
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002852733114836076,
+      "loss": 0.6751,
+      "step": 3635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002852651082781676,
+      "loss": 0.2635,
+      "step": 3636
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028525690290665993,
+      "loss": 0.0178,
+      "step": 3637
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002852486953692159,
+      "loss": 0.4116,
+      "step": 3638
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028524048566596705,
+      "loss": 0.0023,
+      "step": 3639
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002852322737970448,
+      "loss": 0.2307,
+      "step": 3640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028522405976258056,
+      "loss": 0.3647,
+      "step": 3641
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000285215843562706,
+      "loss": 0.206,
+      "step": 3642
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028520762519755263,
+      "loss": 2.0139,
+      "step": 3643
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028519940466725205,
+      "loss": 0.513,
+      "step": 3644
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028519118197193596,
+      "loss": 0.6198,
+      "step": 3645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002851829571117359,
+      "loss": 2.4242,
+      "step": 3646
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002851747300867837,
+      "loss": 0.2906,
+      "step": 3647
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002851665008972111,
+      "loss": 0.4993,
+      "step": 3648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028515826954314977,
+      "loss": 0.4166,
+      "step": 3649
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028515003602473167,
+      "loss": 0.0241,
+      "step": 3650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028514180034208854,
+      "loss": 0.3588,
+      "step": 3651
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002851335624953523,
+      "loss": 0.4358,
+      "step": 3652
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028512532248465487,
+      "loss": 0.9162,
+      "step": 3653
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002851170803101282,
+      "loss": 0.7644,
+      "step": 3654
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028510883597190423,
+      "loss": 0.4105,
+      "step": 3655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028510058947011507,
+      "loss": 0.1203,
+      "step": 3656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028509234080489274,
+      "loss": 0.0325,
+      "step": 3657
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028508408997636925,
+      "loss": 0.6934,
+      "step": 3658
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002850758369846769,
+      "loss": 0.6503,
+      "step": 3659
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002850675818299476,
+      "loss": 0.7976,
+      "step": 3660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002850593245123138,
+      "loss": 0.4508,
+      "step": 3661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028505106503190757,
+      "loss": 0.0024,
+      "step": 3662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002850428033888612,
+      "loss": 0.4479,
+      "step": 3663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000285034539583307,
+      "loss": 0.1874,
+      "step": 3664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028502627361537734,
+      "loss": 0.01,
+      "step": 3665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002850180054852046,
+      "loss": 0.2426,
+      "step": 3666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028500973519292105,
+      "loss": 0.004,
+      "step": 3667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028500146273865924,
+      "loss": 0.5028,
+      "step": 3668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028499318812255165,
+      "loss": 0.3234,
+      "step": 3669
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028498491134473065,
+      "loss": 0.2297,
+      "step": 3670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000284976632405329,
+      "loss": 0.5333,
+      "step": 3671
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002849683513044791,
+      "loss": 0.5568,
+      "step": 3672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002849600680423136,
+      "loss": 0.962,
+      "step": 3673
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028495178261896516,
+      "loss": 0.9128,
+      "step": 3674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028494349503456647,
+      "loss": 0.4102,
+      "step": 3675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002849352052892502,
+      "loss": 0.3746,
+      "step": 3676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028492691338314913,
+      "loss": 0.5391,
+      "step": 3677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028491861931639607,
+      "loss": 2.8051,
+      "step": 3678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028491032308912385,
+      "loss": 4.7542,
+      "step": 3679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002849020247014652,
+      "loss": 1.0177,
+      "step": 3680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848937241535531,
+      "loss": 0.4718,
+      "step": 3681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028488542144552043,
+      "loss": 0.7812,
+      "step": 3682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848771165775002,
+      "loss": 1.125,
+      "step": 3683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848688095496254,
+      "loss": 0.322,
+      "step": 3684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000284860500362029,
+      "loss": 0.3227,
+      "step": 3685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848521890148441,
+      "loss": 0.363,
+      "step": 3686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028484387550820376,
+      "loss": 1.6935,
+      "step": 3687
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848355598422411,
+      "loss": 0.9436,
+      "step": 3688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028482724201708934,
+      "loss": 2.2755,
+      "step": 3689
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028481892203288163,
+      "loss": 0.0053,
+      "step": 3690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028481059988975124,
+      "loss": 0.0056,
+      "step": 3691
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002848022755878314,
+      "loss": 1.4552,
+      "step": 3692
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847939491272554,
+      "loss": 0.4262,
+      "step": 3693
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028478562050815665,
+      "loss": 0.488,
+      "step": 3694
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028477728973066844,
+      "loss": 0.1671,
+      "step": 3695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847689567949242,
+      "loss": 0.2465,
+      "step": 3696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028476062170105745,
+      "loss": 0.3162,
+      "step": 3697
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028475228444920154,
+      "loss": 0.3165,
+      "step": 3698
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028474394503949003,
+      "loss": 0.2957,
+      "step": 3699
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028473560347205644,
+      "loss": 0.8295,
+      "step": 3700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847272597470344,
+      "loss": 0.7106,
+      "step": 3701
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847189138645575,
+      "loss": 0.7689,
+      "step": 3702
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847105658247593,
+      "loss": 0.5874,
+      "step": 3703
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002847022156277736,
+      "loss": 0.6538,
+      "step": 3704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028469386327373416,
+      "loss": 0.5386,
+      "step": 3705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028468550876277455,
+      "loss": 0.5167,
+      "step": 3706
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002846771520950287,
+      "loss": 0.7555,
+      "step": 3707
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028466879327063036,
+      "loss": 0.1263,
+      "step": 3708
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002846604322897134,
+      "loss": 0.7866,
+      "step": 3709
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028465206915241175,
+      "loss": 1.6987,
+      "step": 3710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002846437038588593,
+      "loss": 0.4501,
+      "step": 3711
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028463533640918994,
+      "loss": 0.12,
+      "step": 3712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002846269668035378,
+      "loss": 0.9733,
+      "step": 3713
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002846185950420368,
+      "loss": 0.4929,
+      "step": 3714
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000284610221124821,
+      "loss": 0.4276,
+      "step": 3715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028460184505202455,
+      "loss": 0.0078,
+      "step": 3716
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002845934668237816,
+      "loss": 1.5462,
+      "step": 3717
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002845850864402263,
+      "loss": 0.3844,
+      "step": 3718
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028457670390149274,
+      "loss": 0.461,
+      "step": 3719
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028456831920771524,
+      "loss": 0.831,
+      "step": 3720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028455993235902813,
+      "loss": 0.7782,
+      "step": 3721
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028455154335556564,
+      "loss": 0.2744,
+      "step": 3722
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028454315219746205,
+      "loss": 0.7242,
+      "step": 3723
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028453475888485185,
+      "loss": 0.6372,
+      "step": 3724
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002845263634178694,
+      "loss": 0.4257,
+      "step": 3725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028451796579664917,
+      "loss": 0.0071,
+      "step": 3726
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002845095660213256,
+      "loss": 0.2734,
+      "step": 3727
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002845011640920331,
+      "loss": 1.7249,
+      "step": 3728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028449276000890635,
+      "loss": 0.0042,
+      "step": 3729
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028448435377207993,
+      "loss": 0.3845,
+      "step": 3730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028447594538168837,
+      "loss": 0.4642,
+      "step": 3731
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028446753483786643,
+      "loss": 0.5409,
+      "step": 3732
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002844591221407487,
+      "loss": 0.4596,
+      "step": 3733
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002844507072904699,
+      "loss": 0.0179,
+      "step": 3734
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002844422902871648,
+      "loss": 0.427,
+      "step": 3735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028443387113096826,
+      "loss": 0.5443,
+      "step": 3736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028442544982201497,
+      "loss": 0.8151,
+      "step": 3737
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028441702636043984,
+      "loss": 0.0022,
+      "step": 3738
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028440860074637777,
+      "loss": 0.992,
+      "step": 3739
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002844001729799637,
+      "loss": 1.0488,
+      "step": 3740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843917430613326,
+      "loss": 0.4679,
+      "step": 3741
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028438331099061936,
+      "loss": 0.4786,
+      "step": 3742
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843748767679591,
+      "loss": 0.6736,
+      "step": 3743
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843664403934869,
+      "loss": 0.6597,
+      "step": 3744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843580018673378,
+      "loss": 0.432,
+      "step": 3745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843495611896469,
+      "loss": 0.3286,
+      "step": 3746
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843411183605494,
+      "loss": 0.5017,
+      "step": 3747
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028433267338018054,
+      "loss": 2.6801,
+      "step": 3748
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843242262486755,
+      "loss": 0.5735,
+      "step": 3749
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002843157769661696,
+      "loss": 0.2785,
+      "step": 3750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028430732553279807,
+      "loss": 0.0078,
+      "step": 3751
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028429887194869635,
+      "loss": 0.4213,
+      "step": 3752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842904162139997,
+      "loss": 0.2119,
+      "step": 3753
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842819583288436,
+      "loss": 3.2309,
+      "step": 3754
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842734982933634,
+      "loss": 0.1309,
+      "step": 3755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842650361076947,
+      "loss": 0.4571,
+      "step": 3756
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028425657177197296,
+      "loss": 0.9511,
+      "step": 3757
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842481052863337,
+      "loss": 0.2949,
+      "step": 3758
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842396366509126,
+      "loss": 0.4608,
+      "step": 3759
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028423116586584506,
+      "loss": 5.6621,
+      "step": 3760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842226929312669,
+      "loss": 0.7883,
+      "step": 3761
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028421421784731373,
+      "loss": 0.9131,
+      "step": 3762
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002842057406141213,
+      "loss": 0.0833,
+      "step": 3763
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841972612318253,
+      "loss": 0.3698,
+      "step": 3764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841887797005616,
+      "loss": 0.3154,
+      "step": 3765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028418029602046604,
+      "loss": 0.7328,
+      "step": 3766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028417181019167433,
+      "loss": 0.3897,
+      "step": 3767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841633222143225,
+      "loss": 0.8632,
+      "step": 3768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841548320885464,
+      "loss": 0.063,
+      "step": 3769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028414633981448204,
+      "loss": 0.9287,
+      "step": 3770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841378453922653,
+      "loss": 0.5314,
+      "step": 3771
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841293488220323,
+      "loss": 2.0727,
+      "step": 3772
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841208501039191,
+      "loss": 0.2057,
+      "step": 3773
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002841123492380618,
+      "loss": 0.4533,
+      "step": 3774
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028410384622459643,
+      "loss": 0.0874,
+      "step": 3775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840953410636593,
+      "loss": 1.344,
+      "step": 3776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028408683375538647,
+      "loss": 0.3745,
+      "step": 3777
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028407832429991423,
+      "loss": 0.4969,
+      "step": 3778
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840698126973789,
+      "loss": 0.5038,
+      "step": 3779
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840612989479167,
+      "loss": 0.2629,
+      "step": 3780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000284052783051664,
+      "loss": 0.8141,
+      "step": 3781
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840442650087572,
+      "loss": 0.1384,
+      "step": 3782
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840357448193326,
+      "loss": 0.3164,
+      "step": 3783
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028402722248352673,
+      "loss": 0.6467,
+      "step": 3784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840186980014761,
+      "loss": 0.5706,
+      "step": 3785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840101713733171,
+      "loss": 2.5658,
+      "step": 3786
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002840016425991863,
+      "loss": 0.2649,
+      "step": 3787
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002839931116792203,
+      "loss": 2.2538,
+      "step": 3788
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028398457861355576,
+      "loss": 0.6807,
+      "step": 3789
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002839760434023293,
+      "loss": 0.3224,
+      "step": 3790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002839675060456775,
+      "loss": 0.4197,
+      "step": 3791
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002839589665437372,
+      "loss": 0.4672,
+      "step": 3792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002839504248966451,
+      "loss": 0.0993,
+      "step": 3793
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000283941881104538,
+      "loss": 2.5867,
+      "step": 3794
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028393333516755256,
+      "loss": 0.2776,
+      "step": 3795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002839247870858259,
+      "loss": 1.7821,
+      "step": 3796
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002839162368594947,
+      "loss": 1.0648,
+      "step": 3797
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028390768448869594,
+      "loss": 0.5082,
+      "step": 3798
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028389912997356663,
+      "loss": 0.5488,
+      "step": 3799
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838905733142437,
+      "loss": 0.0054,
+      "step": 3800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028388201451086414,
+      "loss": 0.0097,
+      "step": 3801
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838734535635651,
+      "loss": 0.5075,
+      "step": 3802
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028386489047248356,
+      "loss": 0.4991,
+      "step": 3803
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028385632523775673,
+      "loss": 0.0192,
+      "step": 3804
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838477578595217,
+      "loss": 0.9071,
+      "step": 3805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028383918833791577,
+      "loss": 0.2727,
+      "step": 3806
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838306166730761,
+      "loss": 0.5166,
+      "step": 3807
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028382204286513997,
+      "loss": 0.1999,
+      "step": 3808
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838134669142446,
+      "loss": 1.0827,
+      "step": 3809
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002838048888205274,
+      "loss": 2.9902,
+      "step": 3810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002837963085841257,
+      "loss": 0.4825,
+      "step": 3811
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028378772620517696,
+      "loss": 1.1124,
+      "step": 3812
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028377914168381855,
+      "loss": 0.4816,
+      "step": 3813
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000283770555020188,
+      "loss": 0.0068,
+      "step": 3814
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002837619662144227,
+      "loss": 0.3639,
+      "step": 3815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028375337526666034,
+      "loss": 0.5187,
+      "step": 3816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028374478217703836,
+      "loss": 0.3376,
+      "step": 3817
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002837361869456944,
+      "loss": 0.4811,
+      "step": 3818
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028372758957276615,
+      "loss": 0.3213,
+      "step": 3819
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002837189900583913,
+      "loss": 0.6542,
+      "step": 3820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002837103884027074,
+      "loss": 0.4749,
+      "step": 3821
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002837017846058524,
+      "loss": 4.2183,
+      "step": 3822
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002836931786679639,
+      "loss": 0.5738,
+      "step": 3823
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028368457058917983,
+      "loss": 2.4494,
+      "step": 3824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000283675960369638,
+      "loss": 0.4284,
+      "step": 3825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028366734800947625,
+      "loss": 0.0136,
+      "step": 3826
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028365873350883256,
+      "loss": 0.3473,
+      "step": 3827
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028365011686784487,
+      "loss": 0.0087,
+      "step": 3828
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002836414980866511,
+      "loss": 0.348,
+      "step": 3829
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002836328771653893,
+      "loss": 0.5328,
+      "step": 3830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028362425410419754,
+      "loss": 0.2252,
+      "step": 3831
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002836156289032139,
+      "loss": 0.3036,
+      "step": 3832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002836070015625764,
+      "loss": 0.0098,
+      "step": 3833
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835983720824234,
+      "loss": 0.8061,
+      "step": 3834
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028358974046289293,
+      "loss": 0.4004,
+      "step": 3835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835811067041233,
+      "loss": 2.1369,
+      "step": 3836
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835724708062527,
+      "loss": 0.2569,
+      "step": 3837
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835638327694194,
+      "loss": 1.0764,
+      "step": 3838
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028355519259376185,
+      "loss": 0.5194,
+      "step": 3839
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028354655027941825,
+      "loss": 0.5982,
+      "step": 3840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835379058265271,
+      "loss": 0.138,
+      "step": 3841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002835292592352268,
+      "loss": 0.361,
+      "step": 3842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028352061050565585,
+      "loss": 0.308,
+      "step": 3843
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028351195963795267,
+      "loss": 1.3702,
+      "step": 3844
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028350330663225583,
+      "loss": 0.7233,
+      "step": 3845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834946514887039,
+      "loss": 0.2216,
+      "step": 3846
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834859942074355,
+      "loss": 0.0434,
+      "step": 3847
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028347733478858924,
+      "loss": 0.5975,
+      "step": 3848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834686732323037,
+      "loss": 0.6725,
+      "step": 3849
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028346000953871777,
+      "loss": 2.4808,
+      "step": 3850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028345134370797,
+      "loss": 0.0543,
+      "step": 3851
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028344267574019927,
+      "loss": 0.9902,
+      "step": 3852
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834340056355444,
+      "loss": 0.8356,
+      "step": 3853
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002834253333941441,
+      "loss": 0.7317,
+      "step": 3854
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028341665901613744,
+      "loss": 0.3115,
+      "step": 3855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028340798250166315,
+      "loss": 0.3436,
+      "step": 3856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028339930385086026,
+      "loss": 0.7363,
+      "step": 3857
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002833906230638677,
+      "loss": 1.0451,
+      "step": 3858
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028338194014082454,
+      "loss": 0.5825,
+      "step": 3859
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002833732550818697,
+      "loss": 0.3183,
+      "step": 3860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028336456788714245,
+      "loss": 1.8364,
+      "step": 3861
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00028335587855678174,
+      "loss": 0.5371,
+      "step": 3862
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002833471870909268,
+      "loss": 0.205,
+      "step": 3863
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028333849348971676,
+      "loss": 0.0951,
+      "step": 3864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002833297977532909,
+      "loss": 0.4156,
+      "step": 3865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002833210998817884,
+      "loss": 0.5349,
+      "step": 3866
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002833123998753486,
+      "loss": 0.4445,
+      "step": 3867
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002833036977341107,
+      "loss": 0.0353,
+      "step": 3868
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002832949934582142,
+      "loss": 0.5954,
+      "step": 3869
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028328628704779845,
+      "loss": 1.0516,
+      "step": 3870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028327757850300284,
+      "loss": 0.6938,
+      "step": 3871
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002832688678239668,
+      "loss": 0.3589,
+      "step": 3872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002832601550108299,
+      "loss": 0.7331,
+      "step": 3873
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002832514400637316,
+      "loss": 0.39,
+      "step": 3874
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028324272298281143,
+      "loss": 0.5919,
+      "step": 3875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028323400376820904,
+      "loss": 0.5673,
+      "step": 3876
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002832252824200641,
+      "loss": 0.3725,
+      "step": 3877
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002832165589385161,
+      "loss": 0.5343,
+      "step": 3878
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028320783332370485,
+      "loss": 0.4679,
+      "step": 3879
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028319910557577017,
+      "loss": 0.3979,
+      "step": 3880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028319037569485164,
+      "loss": 0.657,
+      "step": 3881
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028318164368108914,
+      "loss": 1.5665,
+      "step": 3882
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028317290953462247,
+      "loss": 0.9722,
+      "step": 3883
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028316417325559154,
+      "loss": 0.6615,
+      "step": 3884
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002831554348441362,
+      "loss": 0.0097,
+      "step": 3885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002831466943003965,
+      "loss": 0.1128,
+      "step": 3886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028313795162451225,
+      "loss": 0.3375,
+      "step": 3887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002831292068166235,
+      "loss": 0.3425,
+      "step": 3888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028312045987687035,
+      "loss": 0.3479,
+      "step": 3889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002831117108053928,
+      "loss": 0.0494,
+      "step": 3890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000283102959602331,
+      "loss": 0.9703,
+      "step": 3891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028309420626782504,
+      "loss": 0.4379,
+      "step": 3892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028308545080201505,
+      "loss": 0.0061,
+      "step": 3893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002830766932050414,
+      "loss": 0.6393,
+      "step": 3894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028306793347704416,
+      "loss": 0.5325,
+      "step": 3895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002830591716181637,
+      "loss": 0.0383,
+      "step": 3896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002830504076285403,
+      "loss": 0.3364,
+      "step": 3897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028304164150831434,
+      "loss": 4.015,
+      "step": 3898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028303287325762605,
+      "loss": 0.2847,
+      "step": 3899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000283024102876616,
+      "loss": 0.9237,
+      "step": 3900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028301533036542465,
+      "loss": 0.2906,
+      "step": 3901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028300655572419234,
+      "loss": 0.4824,
+      "step": 3902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028299777895305966,
+      "loss": 0.362,
+      "step": 3903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028298900005216724,
+      "loss": 2.0821,
+      "step": 3904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028298021902165547,
+      "loss": 0.7065,
+      "step": 3905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002829714358616651,
+      "loss": 0.5206,
+      "step": 3906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002829626505723367,
+      "loss": 0.8235,
+      "step": 3907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002829538631538111,
+      "loss": 1.0255,
+      "step": 3908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028294507360622885,
+      "loss": 0.4224,
+      "step": 3909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028293628192973077,
+      "loss": 0.0421,
+      "step": 3910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028292748812445767,
+      "loss": 0.5057,
+      "step": 3911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002829186921905503,
+      "loss": 0.5649,
+      "step": 3912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028290989412814964,
+      "loss": 0.2942,
+      "step": 3913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028290109393739644,
+      "loss": 0.313,
+      "step": 3914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002828922916184317,
+      "loss": 0.1747,
+      "step": 3915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002828834871713963,
+      "loss": 0.6228,
+      "step": 3916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028287468059643133,
+      "loss": 0.1012,
+      "step": 3917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028286587189367774,
+      "loss": 0.0353,
+      "step": 3918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028285706106327663,
+      "loss": 0.1336,
+      "step": 3919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002828482481053691,
+      "loss": 1.1515,
+      "step": 3920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028283943302009624,
+      "loss": 0.7156,
+      "step": 3921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028283061580759917,
+      "loss": 0.2488,
+      "step": 3922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002828217964680192,
+      "loss": 0.7509,
+      "step": 3923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028281297500149747,
+      "loss": 0.2181,
+      "step": 3924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002828041514081753,
+      "loss": 1.0415,
+      "step": 3925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028279532568819394,
+      "loss": 0.4923,
+      "step": 3926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002827864978416948,
+      "loss": 1.7824,
+      "step": 3927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028277766786881907,
+      "loss": 0.718,
+      "step": 3928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002827688357697083,
+      "loss": 0.7834,
+      "step": 3929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028276000154450385,
+      "loss": 0.3372,
+      "step": 3930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028275116519334726,
+      "loss": 0.4229,
+      "step": 3931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028274232671638,
+      "loss": 0.146,
+      "step": 3932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002827334861137436,
+      "loss": 0.3976,
+      "step": 3933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028272464338557966,
+      "loss": 1.7805,
+      "step": 3934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002827157985320297,
+      "loss": 0.3198,
+      "step": 3935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002827069515532354,
+      "loss": 0.421,
+      "step": 3936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826981024493384,
+      "loss": 0.2223,
+      "step": 3937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826892512204805,
+      "loss": 0.414,
+      "step": 3938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826803978668034,
+      "loss": 0.1488,
+      "step": 3939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826715423884488,
+      "loss": 0.3334,
+      "step": 3940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028266268478555854,
+      "loss": 0.6445,
+      "step": 3941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028265382505827453,
+      "loss": 0.2507,
+      "step": 3942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826449632067386,
+      "loss": 0.5217,
+      "step": 3943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826360992310927,
+      "loss": 0.3384,
+      "step": 3944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826272331314786,
+      "loss": 0.261,
+      "step": 3945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826183649080385,
+      "loss": 1.1914,
+      "step": 3946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002826094945609143,
+      "loss": 0.9391,
+      "step": 3947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028260062209024806,
+      "loss": 0.4602,
+      "step": 3948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825917474961819,
+      "loss": 0.018,
+      "step": 3949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028258287077885785,
+      "loss": 2.0218,
+      "step": 3950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028257399193841806,
+      "loss": 0.2938,
+      "step": 3951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825651109750048,
+      "loss": 0.1649,
+      "step": 3952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825562278887603,
+      "loss": 0.6509,
+      "step": 3953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825473426798267,
+      "loss": 0.5051,
+      "step": 3954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825384553483463,
+      "loss": 0.6537,
+      "step": 3955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825295658944615,
+      "loss": 0.3921,
+      "step": 3956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002825206743183146,
+      "loss": 0.7091,
+      "step": 3957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282511780620048,
+      "loss": 1.6545,
+      "step": 3958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282502884799804,
+      "loss": 0.3621,
+      "step": 3959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028249398685772524,
+      "loss": 0.4944,
+      "step": 3960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028248508679395415,
+      "loss": 0.0146,
+      "step": 3961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002824761846086332,
+      "loss": 0.5072,
+      "step": 3962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028246728030190496,
+      "loss": 0.6659,
+      "step": 3963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028245837387391205,
+      "loss": 0.9645,
+      "step": 3964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002824494653247971,
+      "loss": 1.3156,
+      "step": 3965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028244055465470266,
+      "loss": 0.8181,
+      "step": 3966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028243164186377153,
+      "loss": 0.4426,
+      "step": 3967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002824227269521465,
+      "loss": 0.4215,
+      "step": 3968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028241380991997014,
+      "loss": 1.3217,
+      "step": 3969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028240489076738535,
+      "loss": 0.1371,
+      "step": 3970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028239596949453495,
+      "loss": 0.6227,
+      "step": 3971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002823870461015618,
+      "loss": 0.0106,
+      "step": 3972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002823781205886088,
+      "loss": 1.1805,
+      "step": 3973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002823691929558189,
+      "loss": 0.9174,
+      "step": 3974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282360263203335,
+      "loss": 0.5624,
+      "step": 3975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028235133133130015,
+      "loss": 0.2993,
+      "step": 3976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002823423973398573,
+      "loss": 0.3535,
+      "step": 3977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002823334612291496,
+      "loss": 0.586,
+      "step": 3978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028232452299932015,
+      "loss": 0.2663,
+      "step": 3979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028231558265051204,
+      "loss": 0.5788,
+      "step": 3980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002823066401828685,
+      "loss": 0.6351,
+      "step": 3981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002822976955965326,
+      "loss": 0.2307,
+      "step": 3982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002822887488916477,
+      "loss": 0.2432,
+      "step": 3983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282279800068357,
+      "loss": 0.5875,
+      "step": 3984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028227084912680386,
+      "loss": 0.4632,
+      "step": 3985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002822618960671316,
+      "loss": 0.8313,
+      "step": 3986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028225294088948355,
+      "loss": 0.0039,
+      "step": 3987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002822439835940031,
+      "loss": 0.8477,
+      "step": 3988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028223502418083375,
+      "loss": 0.363,
+      "step": 3989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028222606265011893,
+      "loss": 0.5953,
+      "step": 3990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028221709900200215,
+      "loss": 0.2107,
+      "step": 3991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282208133236627,
+      "loss": 0.2315,
+      "step": 3992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282199165354137,
+      "loss": 0.9206,
+      "step": 3993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821901953546758,
+      "loss": 0.2808,
+      "step": 3994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282181223238387,
+      "loss": 0.3459,
+      "step": 3995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028217224900541424,
+      "loss": 0.312,
+      "step": 3996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028216327265590126,
+      "loss": 0.4644,
+      "step": 3997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821542941899919,
+      "loss": 0.415,
+      "step": 3998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821453136078298,
+      "loss": 0.5242,
+      "step": 3999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821363309095588,
+      "loss": 0.0152,
+      "step": 4000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028212734609532283,
+      "loss": 0.1088,
+      "step": 4001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028211835916526566,
+      "loss": 0.4315,
+      "step": 4002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002821093701195313,
+      "loss": 2.5104,
+      "step": 4003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028210037895826356,
+      "loss": 0.4073,
+      "step": 4004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820913856816066,
+      "loss": 0.5876,
+      "step": 4005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820823902897043,
+      "loss": 0.5914,
+      "step": 4006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028207339278270077,
+      "loss": 0.3061,
+      "step": 4007
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028206439316074,
+      "loss": 0.3871,
+      "step": 4008
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820553914239663,
+      "loss": 2.0868,
+      "step": 4009
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820463875725236,
+      "loss": 0.1274,
+      "step": 4010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820373816065562,
+      "loss": 0.3229,
+      "step": 4011
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028202837352620826,
+      "loss": 0.4397,
+      "step": 4012
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820193633316241,
+      "loss": 0.1604,
+      "step": 4013
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000282010351022948,
+      "loss": 0.4313,
+      "step": 4014
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002820013366003242,
+      "loss": 1.9944,
+      "step": 4015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819923200638971,
+      "loss": 0.4672,
+      "step": 4016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819833014138111,
+      "loss": 0.0356,
+      "step": 4017
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819742806502106,
+      "loss": 0.1579,
+      "step": 4018
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819652577732401,
+      "loss": 0.0066,
+      "step": 4019
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028195623278304396,
+      "loss": 0.4014,
+      "step": 4020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028194720567976684,
+      "loss": 0.9764,
+      "step": 4021
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028193817646355327,
+      "loss": 0.5534,
+      "step": 4022
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819291451345478,
+      "loss": 0.7895,
+      "step": 4023
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028192011169289507,
+      "loss": 0.6791,
+      "step": 4024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028191107613873977,
+      "loss": 0.8852,
+      "step": 4025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002819020384722265,
+      "loss": 0.9089,
+      "step": 4026
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002818929986935,
+      "loss": 0.4116,
+      "step": 4027
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002818839568027052,
+      "loss": 0.8005,
+      "step": 4028
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002818749127999867,
+      "loss": 0.0153,
+      "step": 4029
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028186586668548934,
+      "loss": 0.5332,
+      "step": 4030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028185681845935805,
+      "loss": 0.7026,
+      "step": 4031
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028184776812173777,
+      "loss": 1.1567,
+      "step": 4032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002818387156727733,
+      "loss": 2.2191,
+      "step": 4033
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028182966111260963,
+      "loss": 1.0056,
+      "step": 4034
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028182060444139186,
+      "loss": 0.3547,
+      "step": 4035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028181154565926487,
+      "loss": 0.2662,
+      "step": 4036
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028180248476637385,
+      "loss": 0.563,
+      "step": 4037
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028179342176286385,
+      "loss": 0.346,
+      "step": 4038
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028178435664887996,
+      "loss": 0.4959,
+      "step": 4039
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002817752894245674,
+      "loss": 0.1405,
+      "step": 4040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002817662200900713,
+      "loss": 0.4412,
+      "step": 4041
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028175714864553706,
+      "loss": 0.1607,
+      "step": 4042
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002817480750911097,
+      "loss": 0.8594,
+      "step": 4043
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028173899942693473,
+      "loss": 0.4573,
+      "step": 4044
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002817299216531573,
+      "loss": 2.1642,
+      "step": 4045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000281720841769923,
+      "loss": 0.949,
+      "step": 4046
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000281711759777377,
+      "loss": 0.4012,
+      "step": 4047
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002817026756756649,
+      "loss": 0.3865,
+      "step": 4048
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000281693589464932,
+      "loss": 0.8176,
+      "step": 4049
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000281684501145324,
+      "loss": 0.6412,
+      "step": 4050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028167541071698636,
+      "loss": 0.2613,
+      "step": 4051
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002816663181800646,
+      "loss": 0.4028,
+      "step": 4052
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002816572235347044,
+      "loss": 0.5256,
+      "step": 4053
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002816481267810513,
+      "loss": 0.3726,
+      "step": 4054
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028163902791925104,
+      "loss": 0.6552,
+      "step": 4055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002816299269494493,
+      "loss": 0.9967,
+      "step": 4056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028162082387179186,
+      "loss": 0.1381,
+      "step": 4057
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028161171868642447,
+      "loss": 0.2301,
+      "step": 4058
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002816026113934929,
+      "loss": 0.5813,
+      "step": 4059
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028159350199314303,
+      "loss": 0.2552,
+      "step": 4060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815843904855207,
+      "loss": 0.682,
+      "step": 4061
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815752768707718,
+      "loss": 0.4241,
+      "step": 4062
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815661611490424,
+      "loss": 2.4526,
+      "step": 4063
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815570433204783,
+      "loss": 0.1253,
+      "step": 4064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028154792338522564,
+      "loss": 0.4189,
+      "step": 4065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028153880134343043,
+      "loss": 0.2804,
+      "step": 4066
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028152967719523867,
+      "loss": 0.6276,
+      "step": 4067
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028152055094079657,
+      "loss": 0.4402,
+      "step": 4068
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815114225802502,
+      "loss": 0.8173,
+      "step": 4069
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002815022921137458,
+      "loss": 0.2763,
+      "step": 4070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814931595414295,
+      "loss": 0.7075,
+      "step": 4071
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814840248634477,
+      "loss": 0.1557,
+      "step": 4072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814748880799464,
+      "loss": 0.0104,
+      "step": 4073
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028146574919107225,
+      "loss": 0.0287,
+      "step": 4074
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814566081969713,
+      "loss": 0.2974,
+      "step": 4075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028144746509779013,
+      "loss": 0.4801,
+      "step": 4076
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814383198936751,
+      "loss": 0.6304,
+      "step": 4077
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814291725847726,
+      "loss": 0.695,
+      "step": 4078
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028142002317122915,
+      "loss": 0.032,
+      "step": 4079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814108716531913,
+      "loss": 0.4559,
+      "step": 4080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002814017180308055,
+      "loss": 1.0522,
+      "step": 4081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002813925623042185,
+      "loss": 0.0771,
+      "step": 4082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028138340447357666,
+      "loss": 2.0958,
+      "step": 4083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028137424453902687,
+      "loss": 0.1076,
+      "step": 4084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028136508250071573,
+      "loss": 0.3002,
+      "step": 4085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002813559183587899,
+      "loss": 0.6258,
+      "step": 4086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028134675211339617,
+      "loss": 0.6258,
+      "step": 4087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002813375837646814,
+      "loss": 0.2202,
+      "step": 4088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028132841331279226,
+      "loss": 0.5469,
+      "step": 4089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028131924075787564,
+      "loss": 0.1395,
+      "step": 4090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002813100661000786,
+      "loss": 0.4671,
+      "step": 4091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002813008893395478,
+      "loss": 0.4417,
+      "step": 4092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002812917104764304,
+      "loss": 0.3437,
+      "step": 4093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002812825295108732,
+      "loss": 0.51,
+      "step": 4094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002812733464430234,
+      "loss": 1.2999,
+      "step": 4095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028126416127302795,
+      "loss": 0.6424,
+      "step": 4096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028125497400103395,
+      "loss": 3.9963,
+      "step": 4097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002812457846271885,
+      "loss": 0.0157,
+      "step": 4098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028123659315163884,
+      "loss": 0.8515,
+      "step": 4099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028122739957453204,
+      "loss": 0.6472,
+      "step": 4100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002812182038960154,
+      "loss": 0.2547,
+      "step": 4101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028120900611623614,
+      "loss": 0.3136,
+      "step": 4102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811998062353416,
+      "loss": 2.047,
+      "step": 4103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811906042534791,
+      "loss": 0.5325,
+      "step": 4104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811814001707959,
+      "loss": 0.6103,
+      "step": 4105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028117219398743944,
+      "loss": 0.0539,
+      "step": 4106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811629857035572,
+      "loss": 0.5971,
+      "step": 4107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811537753192966,
+      "loss": 1.6261,
+      "step": 4108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028114456283480507,
+      "loss": 0.0148,
+      "step": 4109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811353482502302,
+      "loss": 0.8907,
+      "step": 4110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00028112613156571956,
+      "loss": 0.1884,
+      "step": 4111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002811169127814207,
+      "loss": 0.6855,
+      "step": 4112
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028110769189748126,
+      "loss": 0.0039,
+      "step": 4113
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002810984689140489,
+      "loss": 0.1323,
+      "step": 4114
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002810892438312713,
+      "loss": 1.2032,
+      "step": 4115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002810800166492962,
+      "loss": 0.0307,
+      "step": 4116
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002810707873682714,
+      "loss": 0.3828,
+      "step": 4117
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028106155598834465,
+      "loss": 0.9292,
+      "step": 4118
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028105232250966373,
+      "loss": 0.4576,
+      "step": 4119
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002810430869323765,
+      "loss": 0.5067,
+      "step": 4120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000281033849256631,
+      "loss": 0.4815,
+      "step": 4121
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028102460948257503,
+      "loss": 0.7411,
+      "step": 4122
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002810153676103566,
+      "loss": 0.6507,
+      "step": 4123
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002810061236401236,
+      "loss": 0.8208,
+      "step": 4124
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002809968775720242,
+      "loss": 0.5499,
+      "step": 4125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002809876294062064,
+      "loss": 0.2598,
+      "step": 4126
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028097837914281836,
+      "loss": 0.1875,
+      "step": 4127
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028096912678200807,
+      "loss": 0.0199,
+      "step": 4128
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002809598723239238,
+      "loss": 0.481,
+      "step": 4129
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002809506157687137,
+      "loss": 0.229,
+      "step": 4130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280941357116526,
+      "loss": 0.1214,
+      "step": 4131
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028093209636750904,
+      "loss": 0.3319,
+      "step": 4132
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028092283352181105,
+      "loss": 0.3965,
+      "step": 4133
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002809135685795803,
+      "loss": 0.0541,
+      "step": 4134
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002809043015409653,
+      "loss": 0.6846,
+      "step": 4135
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002808950324061144,
+      "loss": 0.8124,
+      "step": 4136
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002808857611751759,
+      "loss": 0.3368,
+      "step": 4137
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028087648784829843,
+      "loss": 0.654,
+      "step": 4138
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028086721242563033,
+      "loss": 0.7277,
+      "step": 4139
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028085793490732037,
+      "loss": 0.413,
+      "step": 4140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028084865529351687,
+      "loss": 0.5271,
+      "step": 4141
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002808393735843686,
+      "loss": 0.9577,
+      "step": 4142
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028083008978002405,
+      "loss": 0.2275,
+      "step": 4143
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280820803880632,
+      "loss": 0.619,
+      "step": 4144
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028081151588634114,
+      "loss": 0.5106,
+      "step": 4145
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028080222579730015,
+      "loss": 1.0352,
+      "step": 4146
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002807929336136578,
+      "loss": 0.6676,
+      "step": 4147
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028078363933556295,
+      "loss": 0.2838,
+      "step": 4148
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002807743429631644,
+      "loss": 0.76,
+      "step": 4149
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002807650444966109,
+      "loss": 0.8321,
+      "step": 4150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002807557439360516,
+      "loss": 0.6694,
+      "step": 4151
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002807464412816352,
+      "loss": 0.2467,
+      "step": 4152
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002807371365335108,
+      "loss": 0.8754,
+      "step": 4153
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028072782969182743,
+      "loss": 0.0797,
+      "step": 4154
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280718520756734,
+      "loss": 0.0201,
+      "step": 4155
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002807092097283797,
+      "loss": 0.5242,
+      "step": 4156
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002806998966069135,
+      "loss": 0.6695,
+      "step": 4157
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002806905813924846,
+      "loss": 0.8619,
+      "step": 4158
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028068126408524225,
+      "loss": 0.5803,
+      "step": 4159
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002806719446853355,
+      "loss": 0.2317,
+      "step": 4160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002806626231929137,
+      "loss": 0.7624,
+      "step": 4161
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002806532996081262,
+      "loss": 0.6095,
+      "step": 4162
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280643973931122,
+      "loss": 0.4621,
+      "step": 4163
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002806346461620507,
+      "loss": 0.611,
+      "step": 4164
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002806253163010616,
+      "loss": 0.3934,
+      "step": 4165
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028061598434830416,
+      "loss": 0.0561,
+      "step": 4166
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002806066503039277,
+      "loss": 1.0086,
+      "step": 4167
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805973141680818,
+      "loss": 0.0764,
+      "step": 4168
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028058797594091587,
+      "loss": 0.28,
+      "step": 4169
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805786356225795,
+      "loss": 0.7364,
+      "step": 4170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805692932132222,
+      "loss": 0.5019,
+      "step": 4171
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805599487129937,
+      "loss": 0.3854,
+      "step": 4172
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805506021220435,
+      "loss": 0.3332,
+      "step": 4173
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805412534405214,
+      "loss": 1.1816,
+      "step": 4174
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280531902668577,
+      "loss": 1.8419,
+      "step": 4175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028052254980636006,
+      "loss": 0.3354,
+      "step": 4176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805131948540204,
+      "loss": 0.2983,
+      "step": 4177
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002805038378117078,
+      "loss": 0.7225,
+      "step": 4178
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804944786795721,
+      "loss": 0.2588,
+      "step": 4179
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028048511745776316,
+      "loss": 0.5814,
+      "step": 4180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028047575414643093,
+      "loss": 0.3001,
+      "step": 4181
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804663887457253,
+      "loss": 1.9747,
+      "step": 4182
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028045702125579624,
+      "loss": 0.4426,
+      "step": 4183
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804476516767937,
+      "loss": 0.9451,
+      "step": 4184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804382800088679,
+      "loss": 0.6191,
+      "step": 4185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804289062521688,
+      "loss": 0.4526,
+      "step": 4186
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804195304068465,
+      "loss": 0.5799,
+      "step": 4187
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028041015247305115,
+      "loss": 0.0662,
+      "step": 4188
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002804007724509329,
+      "loss": 3.295,
+      "step": 4189
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280391390340642,
+      "loss": 0.2687,
+      "step": 4190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028038200614232864,
+      "loss": 0.752,
+      "step": 4191
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028037261985614317,
+      "loss": 0.9116,
+      "step": 4192
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028036323148223583,
+      "loss": 0.4969,
+      "step": 4193
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280353841020757,
+      "loss": 0.0517,
+      "step": 4194
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280344448471857,
+      "loss": 4.5277,
+      "step": 4195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028033505383568633,
+      "loss": 0.3933,
+      "step": 4196
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002803256571123954,
+      "loss": 0.5507,
+      "step": 4197
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002803162583021346,
+      "loss": 0.2749,
+      "step": 4198
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028030685740505453,
+      "loss": 0.6076,
+      "step": 4199
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802974544213057,
+      "loss": 0.7901,
+      "step": 4200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028028804935103873,
+      "loss": 0.3741,
+      "step": 4201
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802786421944041,
+      "loss": 0.2838,
+      "step": 4202
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802692329515526,
+      "loss": 0.1035,
+      "step": 4203
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802598216226348,
+      "loss": 0.5401,
+      "step": 4204
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802504082078015,
+      "loss": 0.2434,
+      "step": 4205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028024099270720335,
+      "loss": 0.3238,
+      "step": 4206
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802315751209912,
+      "loss": 0.7792,
+      "step": 4207
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802221554493158,
+      "loss": 0.5513,
+      "step": 4208
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280212733692328,
+      "loss": 0.4654,
+      "step": 4209
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002802033098501787,
+      "loss": 0.9078,
+      "step": 4210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801938839230188,
+      "loss": 0.6671,
+      "step": 4211
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801844559109992,
+      "loss": 0.0216,
+      "step": 4212
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000280175025814271,
+      "loss": 0.6659,
+      "step": 4213
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801655936329851,
+      "loss": 0.6683,
+      "step": 4214
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801561593672926,
+      "loss": 0.3009,
+      "step": 4215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028014672301734445,
+      "loss": 0.9509,
+      "step": 4216
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028013728458329195,
+      "loss": 2.0034,
+      "step": 4217
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801278440652861,
+      "loss": 0.108,
+      "step": 4218
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801184014634782,
+      "loss": 0.5128,
+      "step": 4219
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002801089567780192,
+      "loss": 0.4548,
+      "step": 4220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028009951000906066,
+      "loss": 0.2766,
+      "step": 4221
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800900611567537,
+      "loss": 0.4293,
+      "step": 4222
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028008061022124966,
+      "loss": 0.6536,
+      "step": 4223
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028007115720269987,
+      "loss": 0.3032,
+      "step": 4224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800617021012557,
+      "loss": 0.5988,
+      "step": 4225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028005224491706854,
+      "loss": 0.4227,
+      "step": 4226
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800427856502899,
+      "loss": 0.7656,
+      "step": 4227
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800333243010712,
+      "loss": 0.5999,
+      "step": 4228
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800238608695639,
+      "loss": 0.6409,
+      "step": 4229
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002800143953559197,
+      "loss": 0.3997,
+      "step": 4230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00028000492776029,
+      "loss": 0.3835,
+      "step": 4231
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027999545808282656,
+      "loss": 1.8859,
+      "step": 4232
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027998598632368094,
+      "loss": 0.5593,
+      "step": 4233
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027997651248300485,
+      "loss": 0.2204,
+      "step": 4234
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002799670365609499,
+      "loss": 0.3413,
+      "step": 4235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000279957558557668,
+      "loss": 0.5179,
+      "step": 4236
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027994807847331077,
+      "loss": 0.5735,
+      "step": 4237
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002799385963080301,
+      "loss": 1.3478,
+      "step": 4238
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027992911206197786,
+      "loss": 0.3329,
+      "step": 4239
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002799196257353059,
+      "loss": 0.0861,
+      "step": 4240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002799101373281661,
+      "loss": 1.1619,
+      "step": 4241
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027990064684071036,
+      "loss": 0.0601,
+      "step": 4242
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027989115427309077,
+      "loss": 0.4092,
+      "step": 4243
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027988165962545923,
+      "loss": 0.7797,
+      "step": 4244
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002798721628979679,
+      "loss": 0.8289,
+      "step": 4245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002798626640907688,
+      "loss": 1.6118,
+      "step": 4246
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000279853163204014,
+      "loss": 0.497,
+      "step": 4247
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027984366023785565,
+      "loss": 0.2213,
+      "step": 4248
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000279834155192446,
+      "loss": 0.461,
+      "step": 4249
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002798246480679372,
+      "loss": 0.5236,
+      "step": 4250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027981513886448143,
+      "loss": 0.5033,
+      "step": 4251
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027980562758223113,
+      "loss": 0.264,
+      "step": 4252
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027979611422133846,
+      "loss": 0.0229,
+      "step": 4253
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027978659878195584,
+      "loss": 2.088,
+      "step": 4254
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002797770812642356,
+      "loss": 0.26,
+      "step": 4255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002797675616683302,
+      "loss": 0.0154,
+      "step": 4256
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027975803999439204,
+      "loss": 0.7586,
+      "step": 4257
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002797485162425736,
+      "loss": 0.3637,
+      "step": 4258
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002797389904130274,
+      "loss": 1.2241,
+      "step": 4259
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027972946250590596,
+      "loss": 0.5586,
+      "step": 4260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027971993252136194,
+      "loss": 0.5215,
+      "step": 4261
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002797104004595478,
+      "loss": 0.5738,
+      "step": 4262
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027970086632061634,
+      "loss": 0.5211,
+      "step": 4263
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027969133010472013,
+      "loss": 0.3055,
+      "step": 4264
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002796817918120119,
+      "loss": 0.567,
+      "step": 4265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002796722514426444,
+      "loss": 0.9663,
+      "step": 4266
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027966270899677044,
+      "loss": 0.3425,
+      "step": 4267
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027965316447454275,
+      "loss": 0.4018,
+      "step": 4268
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002796436178761142,
+      "loss": 0.0241,
+      "step": 4269
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027963406920163776,
+      "loss": 0.0209,
+      "step": 4270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027962451845126615,
+      "loss": 0.5274,
+      "step": 4271
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002796149656251525,
+      "loss": 0.2309,
+      "step": 4272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027960541072344965,
+      "loss": 0.7831,
+      "step": 4273
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795958537463106,
+      "loss": 0.0416,
+      "step": 4274
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027958629469388856,
+      "loss": 0.0158,
+      "step": 4275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795767335663365,
+      "loss": 0.7389,
+      "step": 4276
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795671703638074,
+      "loss": 0.0448,
+      "step": 4277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795576050864546,
+      "loss": 0.5461,
+      "step": 4278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795480377344312,
+      "loss": 0.4247,
+      "step": 4279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027953846830789037,
+      "loss": 1.6354,
+      "step": 4280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795288968069854,
+      "loss": 0.7513,
+      "step": 4281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027951932323186954,
+      "loss": 0.6022,
+      "step": 4282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795097475826961,
+      "loss": 0.4227,
+      "step": 4283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002795001698596184,
+      "loss": 1.1749,
+      "step": 4284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002794905900627899,
+      "loss": 0.4575,
+      "step": 4285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002794810081923639,
+      "loss": 0.8736,
+      "step": 4286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027947142424849386,
+      "loss": 0.0147,
+      "step": 4287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027946183823133325,
+      "loss": 0.3215,
+      "step": 4288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002794522501410356,
+      "loss": 0.3259,
+      "step": 4289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002794426599777545,
+      "loss": 0.9876,
+      "step": 4290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002794330677416434,
+      "loss": 0.0068,
+      "step": 4291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000279423473432856,
+      "loss": 0.5751,
+      "step": 4292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002794138770515459,
+      "loss": 0.4189,
+      "step": 4293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002794042785978668,
+      "loss": 0.0084,
+      "step": 4294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027939467807197235,
+      "loss": 1.9854,
+      "step": 4295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793850754740164,
+      "loss": 1.2459,
+      "step": 4296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793754708041526,
+      "loss": 1.2858,
+      "step": 4297
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793658640625348,
+      "loss": 0.2324,
+      "step": 4298
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793562552493168,
+      "loss": 0.9893,
+      "step": 4299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793466443646526,
+      "loss": 0.3138,
+      "step": 4300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027933703140869596,
+      "loss": 0.0068,
+      "step": 4301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793274163816009,
+      "loss": 1.1611,
+      "step": 4302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002793177992835213,
+      "loss": 0.7265,
+      "step": 4303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027930818011461124,
+      "loss": 0.3686,
+      "step": 4304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027929855887502474,
+      "loss": 0.035,
+      "step": 4305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027928893556491594,
+      "loss": 0.2936,
+      "step": 4306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002792793101844388,
+      "loss": 0.2819,
+      "step": 4307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027926968273374753,
+      "loss": 0.3809,
+      "step": 4308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002792600532129963,
+      "loss": 0.003,
+      "step": 4309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027925042162233935,
+      "loss": 0.3725,
+      "step": 4310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002792407879619309,
+      "loss": 1.1052,
+      "step": 4311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027923115223192516,
+      "loss": 0.8955,
+      "step": 4312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002792215144324764,
+      "loss": 0.527,
+      "step": 4313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027921187456373905,
+      "loss": 0.4939,
+      "step": 4314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002792022326258675,
+      "loss": 0.4732,
+      "step": 4315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791925886190161,
+      "loss": 0.371,
+      "step": 4316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791829425433392,
+      "loss": 0.734,
+      "step": 4317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791732943989914,
+      "loss": 0.7179,
+      "step": 4318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791636441861272,
+      "loss": 0.1081,
+      "step": 4319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000279153991904901,
+      "loss": 0.8209,
+      "step": 4320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027914433755546757,
+      "loss": 0.0046,
+      "step": 4321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027913468113798125,
+      "loss": 0.3426,
+      "step": 4322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791250226525969,
+      "loss": 0.7745,
+      "step": 4323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791153620994691,
+      "loss": 0.0237,
+      "step": 4324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002791056994787525,
+      "loss": 0.0078,
+      "step": 4325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027909603479060193,
+      "loss": 0.4542,
+      "step": 4326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002790863680351721,
+      "loss": 1.8731,
+      "step": 4327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027907669921261783,
+      "loss": 0.629,
+      "step": 4328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027906702832309395,
+      "loss": 0.8727,
+      "step": 4329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027905735536675526,
+      "loss": 1.1504,
+      "step": 4330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027904768034375674,
+      "loss": 1.3006,
+      "step": 4331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002790380032542533,
+      "loss": 1.0055,
+      "step": 4332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002790283240983999,
+      "loss": 1.8324,
+      "step": 4333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002790186428763515,
+      "loss": 0.2731,
+      "step": 4334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002790089595882632,
+      "loss": 0.2351,
+      "step": 4335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027899927423429005,
+      "loss": 0.5169,
+      "step": 4336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002789895868145871,
+      "loss": 0.2773,
+      "step": 4337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002789798973293095,
+      "loss": 0.4029,
+      "step": 4338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002789702057786124,
+      "loss": 2.1134,
+      "step": 4339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027896051216265097,
+      "loss": 0.5441,
+      "step": 4340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027895081648158056,
+      "loss": 0.005,
+      "step": 4341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027894111873555627,
+      "loss": 0.3038,
+      "step": 4342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027893141892473356,
+      "loss": 0.3177,
+      "step": 4343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002789217170492677,
+      "loss": 0.0039,
+      "step": 4344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027891201310931396,
+      "loss": 0.2755,
+      "step": 4345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002789023071050278,
+      "loss": 0.4741,
+      "step": 4346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027889259903656464,
+      "loss": 0.2919,
+      "step": 4347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027888288890407995,
+      "loss": 0.4332,
+      "step": 4348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027887317670772924,
+      "loss": 0.0037,
+      "step": 4349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000278863462447668,
+      "loss": 0.6553,
+      "step": 4350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002788537461240519,
+      "loss": 0.0056,
+      "step": 4351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002788440277370363,
+      "loss": 0.3714,
+      "step": 4352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000278834307286777,
+      "loss": 0.419,
+      "step": 4353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027882458477342966,
+      "loss": 0.0437,
+      "step": 4354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002788148601971499,
+      "loss": 0.3629,
+      "step": 4355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027880513355809354,
+      "loss": 0.1398,
+      "step": 4356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002787954048564162,
+      "loss": 0.5873,
+      "step": 4357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027878567409227373,
+      "loss": 0.4231,
+      "step": 4358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00027877594126582205,
+      "loss": 0.4535,
+      "step": 4359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002787662063772169,
+      "loss": 0.5649,
+      "step": 4360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002787564694266142,
+      "loss": 0.8206,
+      "step": 4361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002787467304141699,
+      "loss": 0.5902,
+      "step": 4362
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027873698934003993,
+      "loss": 2.0392,
+      "step": 4363
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002787272462043802,
+      "loss": 0.1958,
+      "step": 4364
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027871750100734693,
+      "loss": 0.163,
+      "step": 4365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000278707753749096,
+      "loss": 0.3599,
+      "step": 4366
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002786980044297836,
+      "loss": 0.7988,
+      "step": 4367
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002786882530495658,
+      "loss": 0.8592,
+      "step": 4368
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002786784996085987,
+      "loss": 0.0057,
+      "step": 4369
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002786687441070386,
+      "loss": 0.452,
+      "step": 4370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027865898654504166,
+      "loss": 0.9084,
+      "step": 4371
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027864922692276417,
+      "loss": 0.3247,
+      "step": 4372
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027863946524036236,
+      "loss": 0.35,
+      "step": 4373
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002786297014979925,
+      "loss": 0.5794,
+      "step": 4374
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027861993569581117,
+      "loss": 0.405,
+      "step": 4375
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027861016783397446,
+      "loss": 0.0104,
+      "step": 4376
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027860039791263903,
+      "loss": 0.0164,
+      "step": 4377
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027859062593196116,
+      "loss": 3.3765,
+      "step": 4378
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027858085189209745,
+      "loss": 0.7989,
+      "step": 4379
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002785710757932043,
+      "loss": 1.2049,
+      "step": 4380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027856129763543846,
+      "loss": 0.006,
+      "step": 4381
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027855151741895624,
+      "loss": 0.6369,
+      "step": 4382
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002785417351439145,
+      "loss": 3.0782,
+      "step": 4383
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002785319508104697,
+      "loss": 2.4709,
+      "step": 4384
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002785221644187787,
+      "loss": 0.6792,
+      "step": 4385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027851237596899803,
+      "loss": 0.7062,
+      "step": 4386
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002785025854612846,
+      "loss": 0.1333,
+      "step": 4387
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027849279289579506,
+      "loss": 0.4848,
+      "step": 4388
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027848299827268634,
+      "loss": 0.6412,
+      "step": 4389
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002784732015921152,
+      "loss": 0.4653,
+      "step": 4390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027846340285423853,
+      "loss": 0.2497,
+      "step": 4391
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002784536020592133,
+      "loss": 0.3355,
+      "step": 4392
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027844379920719637,
+      "loss": 0.8142,
+      "step": 4393
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027843399429834485,
+      "loss": 0.9562,
+      "step": 4394
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002784241873328156,
+      "loss": 0.6317,
+      "step": 4395
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002784143783107657,
+      "loss": 0.6153,
+      "step": 4396
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002784045672323523,
+      "loss": 0.0747,
+      "step": 4397
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027839475409773245,
+      "loss": 0.2239,
+      "step": 4398
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027838493890706327,
+      "loss": 0.5388,
+      "step": 4399
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000278375121660502,
+      "loss": 0.8335,
+      "step": 4400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002783653023582059,
+      "loss": 0.2119,
+      "step": 4401
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000278355481000332,
+      "loss": 0.0697,
+      "step": 4402
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027834565758703777,
+      "loss": 0.2172,
+      "step": 4403
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027833583211848045,
+      "loss": 0.7571,
+      "step": 4404
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002783260045948174,
+      "loss": 0.1159,
+      "step": 4405
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002783161750162059,
+      "loss": 0.6592,
+      "step": 4406
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027830634338280356,
+      "loss": 0.0361,
+      "step": 4407
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002782965096947676,
+      "loss": 0.3402,
+      "step": 4408
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002782866739522556,
+      "loss": 0.2689,
+      "step": 4409
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027827683615542505,
+      "loss": 0.34,
+      "step": 4410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002782669963044335,
+      "loss": 0.3058,
+      "step": 4411
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002782571543994385,
+      "loss": 3.178,
+      "step": 4412
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002782473104405977,
+      "loss": 2.172,
+      "step": 4413
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002782374644280686,
+      "loss": 0.3023,
+      "step": 4414
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027822761636200904,
+      "loss": 3.9289,
+      "step": 4415
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002782177662425766,
+      "loss": 1.3095,
+      "step": 4416
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002782079140699291,
+      "loss": 1.1827,
+      "step": 4417
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027819805984422426,
+      "loss": 0.1429,
+      "step": 4418
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781882035656199,
+      "loss": 0.7578,
+      "step": 4419
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781783452342738,
+      "loss": 0.2782,
+      "step": 4420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781684848503439,
+      "loss": 1.9076,
+      "step": 4421
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027815862241398805,
+      "loss": 1.1187,
+      "step": 4422
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781487579253642,
+      "loss": 0.6677,
+      "step": 4423
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781388913846304,
+      "loss": 0.2719,
+      "step": 4424
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027812902279194444,
+      "loss": 0.5352,
+      "step": 4425
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027811915214746456,
+      "loss": 0.4238,
+      "step": 4426
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002781092794513487,
+      "loss": 0.5649,
+      "step": 4427
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000278099404703755,
+      "loss": 0.1175,
+      "step": 4428
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002780895279048416,
+      "loss": 2.7913,
+      "step": 4429
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027807964905476667,
+      "loss": 0.9143,
+      "step": 4430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002780697681536883,
+      "loss": 0.1744,
+      "step": 4431
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027805988520176485,
+      "loss": 0.8424,
+      "step": 4432
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002780500001991545,
+      "loss": 0.6612,
+      "step": 4433
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027804011314601563,
+      "loss": 0.4019,
+      "step": 4434
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027803022404250646,
+      "loss": 0.5821,
+      "step": 4435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002780203328887854,
+      "loss": 0.0955,
+      "step": 4436
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027801043968501087,
+      "loss": 0.0732,
+      "step": 4437
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002780005444313413,
+      "loss": 0.7397,
+      "step": 4438
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027799064712793505,
+      "loss": 0.9572,
+      "step": 4439
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002779807477749507,
+      "loss": 0.9484,
+      "step": 4440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002779708463725467,
+      "loss": 1.2755,
+      "step": 4441
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002779609429208817,
+      "loss": 0.1855,
+      "step": 4442
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027795103742011423,
+      "loss": 0.5347,
+      "step": 4443
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002779411298704029,
+      "loss": 0.3326,
+      "step": 4444
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002779312202719064,
+      "loss": 0.364,
+      "step": 4445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027792130862478343,
+      "loss": 0.4792,
+      "step": 4446
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002779113949291927,
+      "loss": 0.864,
+      "step": 4447
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027790147918529284,
+      "loss": 0.5777,
+      "step": 4448
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002778915613932428,
+      "loss": 0.0255,
+      "step": 4449
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027788164155320146,
+      "loss": 0.4516,
+      "step": 4450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002778717196653274,
+      "loss": 0.3654,
+      "step": 4451
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027786179572977977,
+      "loss": 0.5517,
+      "step": 4452
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002778518697467173,
+      "loss": 0.1329,
+      "step": 4453
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027784194171629904,
+      "loss": 0.0712,
+      "step": 4454
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027783201163868393,
+      "loss": 0.4304,
+      "step": 4455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027782207951403104,
+      "loss": 0.0506,
+      "step": 4456
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002778121453424993,
+      "loss": 0.6375,
+      "step": 4457
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027780220912424793,
+      "loss": 0.6125,
+      "step": 4458
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000277792270859436,
+      "loss": 0.692,
+      "step": 4459
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027778233054822265,
+      "loss": 0.5271,
+      "step": 4460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027777238819076704,
+      "loss": 0.4906,
+      "step": 4461
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027776244378722835,
+      "loss": 0.0133,
+      "step": 4462
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002777524973377659,
+      "loss": 0.3871,
+      "step": 4463
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027774254884253895,
+      "loss": 0.1396,
+      "step": 4464
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002777325983017068,
+      "loss": 0.2706,
+      "step": 4465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002777226457154287,
+      "loss": 1.4036,
+      "step": 4466
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027771269108386423,
+      "loss": 2.9023,
+      "step": 4467
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002777027344071726,
+      "loss": 0.3598,
+      "step": 4468
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002776927756855134,
+      "loss": 0.2591,
+      "step": 4469
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027768281491904594,
+      "loss": 0.9047,
+      "step": 4470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002776728521079299,
+      "loss": 0.6496,
+      "step": 4471
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027766288725232475,
+      "loss": 0.5866,
+      "step": 4472
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027765292035239003,
+      "loss": 0.6965,
+      "step": 4473
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027764295140828536,
+      "loss": 2.0685,
+      "step": 4474
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027763298042017044,
+      "loss": 1.172,
+      "step": 4475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027762300738820486,
+      "loss": 0.3979,
+      "step": 4476
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002776130323125483,
+      "loss": 0.4309,
+      "step": 4477
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002776030551933606,
+      "loss": 0.3802,
+      "step": 4478
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027759307603080147,
+      "loss": 0.007,
+      "step": 4479
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002775830948250307,
+      "loss": 0.3232,
+      "step": 4480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027757311157620816,
+      "loss": 0.451,
+      "step": 4481
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027756312628449375,
+      "loss": 0.4759,
+      "step": 4482
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027755313895004726,
+      "loss": 0.5381,
+      "step": 4483
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027754314957302865,
+      "loss": 0.8698,
+      "step": 4484
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027753315815359797,
+      "loss": 0.4446,
+      "step": 4485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002775231646919151,
+      "loss": 0.3708,
+      "step": 4486
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002775131691881402,
+      "loss": 0.9157,
+      "step": 4487
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027750317164243325,
+      "loss": 0.5528,
+      "step": 4488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027749317205495434,
+      "loss": 2.487,
+      "step": 4489
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774831704258636,
+      "loss": 0.9498,
+      "step": 4490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774731667553213,
+      "loss": 0.5479,
+      "step": 4491
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774631610434875,
+      "loss": 0.2179,
+      "step": 4492
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774531532905224,
+      "loss": 0.4739,
+      "step": 4493
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774431434965864,
+      "loss": 0.3543,
+      "step": 4494
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027743313166183974,
+      "loss": 0.5958,
+      "step": 4495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027742311778644266,
+      "loss": 0.4422,
+      "step": 4496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002774131018705557,
+      "loss": 0.4815,
+      "step": 4497
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027740308391433903,
+      "loss": 0.3579,
+      "step": 4498
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002773930639179532,
+      "loss": 1.0003,
+      "step": 4499
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027738304188155863,
+      "loss": 0.9886,
+      "step": 4500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027737301780531584,
+      "loss": 2.2835,
+      "step": 4501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002773629916893853,
+      "loss": 0.5661,
+      "step": 4502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002773529635339277,
+      "loss": 0.6287,
+      "step": 4503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002773429333391034,
+      "loss": 0.6207,
+      "step": 4504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002773329011050732,
+      "loss": 0.751,
+      "step": 4505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002773228668319976,
+      "loss": 0.3909,
+      "step": 4506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027731283052003744,
+      "loss": 0.9652,
+      "step": 4507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002773027921693534,
+      "loss": 0.5336,
+      "step": 4508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027729275178010616,
+      "loss": 0.4778,
+      "step": 4509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027728270935245655,
+      "loss": 0.4051,
+      "step": 4510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027727266488656536,
+      "loss": 0.3682,
+      "step": 4511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772626183825934,
+      "loss": 0.425,
+      "step": 4512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772525698407017,
+      "loss": 0.7613,
+      "step": 4513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027724251926105107,
+      "loss": 0.3795,
+      "step": 4514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772324666438024,
+      "loss": 0.5998,
+      "step": 4515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027722241198911674,
+      "loss": 0.7604,
+      "step": 4516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772123552971551,
+      "loss": 0.2375,
+      "step": 4517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002772022965680785,
+      "loss": 0.7477,
+      "step": 4518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000277192235802048,
+      "loss": 0.1486,
+      "step": 4519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027718217299922476,
+      "loss": 0.4627,
+      "step": 4520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002771721081597699,
+      "loss": 0.2302,
+      "step": 4521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002771620412838446,
+      "loss": 0.0929,
+      "step": 4522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027715197237161005,
+      "loss": 0.3909,
+      "step": 4523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002771419014232274,
+      "loss": 0.6392,
+      "step": 4524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027713182843885816,
+      "loss": 0.7922,
+      "step": 4525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027712175341866336,
+      "loss": 0.4876,
+      "step": 4526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027711167636280455,
+      "loss": 0.2769,
+      "step": 4527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027710159727144294,
+      "loss": 0.0404,
+      "step": 4528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027709151614474005,
+      "loss": 1.3165,
+      "step": 4529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027708143298285724,
+      "loss": 0.3936,
+      "step": 4530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000277071347785956,
+      "loss": 0.1908,
+      "step": 4531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027706126055419784,
+      "loss": 0.5606,
+      "step": 4532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027705117128774435,
+      "loss": 0.2655,
+      "step": 4533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002770410799867569,
+      "loss": 1.4874,
+      "step": 4534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027703098665139735,
+      "loss": 0.1534,
+      "step": 4535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002770208912818271,
+      "loss": 0.2581,
+      "step": 4536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027701079387820797,
+      "loss": 1.2315,
+      "step": 4537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027700069444070157,
+      "loss": 1.0477,
+      "step": 4538
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027699059296946966,
+      "loss": 0.3013,
+      "step": 4539
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002769804894646739,
+      "loss": 0.2818,
+      "step": 4540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027697038392647625,
+      "loss": 0.5636,
+      "step": 4541
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002769602763550385,
+      "loss": 0.1991,
+      "step": 4542
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002769501667505224,
+      "loss": 0.0335,
+      "step": 4543
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027694005511308997,
+      "loss": 0.4355,
+      "step": 4544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000276929941442903,
+      "loss": 0.9795,
+      "step": 4545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002769198257401235,
+      "loss": 0.4668,
+      "step": 4546
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027690970800491353,
+      "loss": 1.2355,
+      "step": 4547
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000276899588237435,
+      "loss": 0.4232,
+      "step": 4548
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027688946643785004,
+      "loss": 0.7761,
+      "step": 4549
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002768793426063207,
+      "loss": 0.0287,
+      "step": 4550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027686921674300913,
+      "loss": 2.5073,
+      "step": 4551
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027685908884807745,
+      "loss": 0.4831,
+      "step": 4552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002768489589216878,
+      "loss": 0.5816,
+      "step": 4553
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027683882696400255,
+      "loss": 0.479,
+      "step": 4554
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027682869297518376,
+      "loss": 0.3167,
+      "step": 4555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002768185569553938,
+      "loss": 0.2036,
+      "step": 4556
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027680841890479504,
+      "loss": 0.3015,
+      "step": 4557
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002767982788235497,
+      "loss": 0.1785,
+      "step": 4558
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027678813671182023,
+      "loss": 0.592,
+      "step": 4559
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002767779925697691,
+      "loss": 2.0463,
+      "step": 4560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027676784639755857,
+      "loss": 1.2711,
+      "step": 4561
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027675769819535134,
+      "loss": 0.209,
+      "step": 4562
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027674754796330975,
+      "loss": 0.3874,
+      "step": 4563
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002767373957015964,
+      "loss": 0.5306,
+      "step": 4564
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002767272414103739,
+      "loss": 0.0265,
+      "step": 4565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002767170850898048,
+      "loss": 0.1205,
+      "step": 4566
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027670692674005177,
+      "loss": 0.1229,
+      "step": 4567
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027669676636127744,
+      "loss": 0.2017,
+      "step": 4568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027668660395364453,
+      "loss": 0.0966,
+      "step": 4569
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002766764395173159,
+      "loss": 0.3506,
+      "step": 4570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002766662730524541,
+      "loss": 1.0515,
+      "step": 4571
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027665610455922203,
+      "loss": 0.3124,
+      "step": 4572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027664593403778256,
+      "loss": 1.0887,
+      "step": 4573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027663576148829853,
+      "loss": 0.0447,
+      "step": 4574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027662558691093285,
+      "loss": 0.2446,
+      "step": 4575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002766154103058484,
+      "loss": 0.8983,
+      "step": 4576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027660523167320815,
+      "loss": 0.4363,
+      "step": 4577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002765950510131752,
+      "loss": 0.5447,
+      "step": 4578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027658486832591244,
+      "loss": 0.3466,
+      "step": 4579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000276574683611583,
+      "loss": 0.8632,
+      "step": 4580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027656449687034996,
+      "loss": 2.6544,
+      "step": 4581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002765543081023765,
+      "loss": 0.0464,
+      "step": 4582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002765441173078257,
+      "loss": 0.0081,
+      "step": 4583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002765339244868608,
+      "loss": 0.8443,
+      "step": 4584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027652372963964496,
+      "loss": 2.3009,
+      "step": 4585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027651353276634145,
+      "loss": 0.954,
+      "step": 4586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027650333386711363,
+      "loss": 0.9177,
+      "step": 4587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027649313294212475,
+      "loss": 1.4751,
+      "step": 4588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027648292999153825,
+      "loss": 0.0168,
+      "step": 4589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002764727250155174,
+      "loss": 0.614,
+      "step": 4590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027646251801422564,
+      "loss": 0.5569,
+      "step": 4591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027645230898782643,
+      "loss": 1.1219,
+      "step": 4592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027644209793648337,
+      "loss": 0.1616,
+      "step": 4593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002764318848603598,
+      "loss": 0.5348,
+      "step": 4594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027642166975961935,
+      "loss": 0.4356,
+      "step": 4595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002764114526344256,
+      "loss": 0.6636,
+      "step": 4596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027640123348494217,
+      "loss": 0.0339,
+      "step": 4597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002763910123113327,
+      "loss": 0.5001,
+      "step": 4598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002763807891137608,
+      "loss": 0.5303,
+      "step": 4599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027637056389239026,
+      "loss": 1.057,
+      "step": 4600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002763603366473848,
+      "loss": 0.7144,
+      "step": 4601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027635010737890814,
+      "loss": 0.37,
+      "step": 4602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027633987608712424,
+      "loss": 0.1531,
+      "step": 4603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027632964277219676,
+      "loss": 0.3102,
+      "step": 4604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002763194074342896,
+      "loss": 0.8325,
+      "step": 4605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002763091700735668,
+      "loss": 0.9156,
+      "step": 4606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002762989306901922,
+      "loss": 0.4611,
+      "step": 4607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002762886892843297,
+      "loss": 0.6333,
+      "step": 4608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027627844585614343,
+      "loss": 0.5614,
+      "step": 4609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00027626820040579735,
+      "loss": 0.6592,
+      "step": 4610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002762579529334556,
+      "loss": 0.7703,
+      "step": 4611
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002762477034392821,
+      "loss": 0.4122,
+      "step": 4612
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002762374519234412,
+      "loss": 0.5301,
+      "step": 4613
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002762271983860969,
+      "loss": 1.3707,
+      "step": 4614
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002762169428274135,
+      "loss": 0.8262,
+      "step": 4615
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002762066852475552,
+      "loss": 0.1937,
+      "step": 4616
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027619642564668623,
+      "loss": 0.3326,
+      "step": 4617
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002761861640249709,
+      "loss": 0.6488,
+      "step": 4618
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002761759003825735,
+      "loss": 0.4868,
+      "step": 4619
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002761656347196585,
+      "loss": 0.8142,
+      "step": 4620
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002761553670363901,
+      "loss": 0.5672,
+      "step": 4621
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002761450973329329,
+      "loss": 0.1039,
+      "step": 4622
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027613482560945123,
+      "loss": 1.6641,
+      "step": 4623
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027612455186610966,
+      "loss": 0.0757,
+      "step": 4624
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002761142761030727,
+      "loss": 0.6269,
+      "step": 4625
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002761039983205048,
+      "loss": 0.7405,
+      "step": 4626
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027609371851857075,
+      "loss": 0.4267,
+      "step": 4627
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027608343669743493,
+      "loss": 0.5557,
+      "step": 4628
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027607315285726216,
+      "loss": 0.5575,
+      "step": 4629
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000276062866998217,
+      "loss": 0.3985,
+      "step": 4630
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002760525791204643,
+      "loss": 1.904,
+      "step": 4631
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027604228922416865,
+      "loss": 0.5421,
+      "step": 4632
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027603199730949496,
+      "loss": 1.7813,
+      "step": 4633
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002760217033766079,
+      "loss": 0.447,
+      "step": 4634
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002760114074256724,
+      "loss": 0.7645,
+      "step": 4635
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002760011094568534,
+      "loss": 0.6218,
+      "step": 4636
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027599080947031575,
+      "loss": 1.9143,
+      "step": 4637
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027598050746622426,
+      "loss": 0.7057,
+      "step": 4638
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002759702034447441,
+      "loss": 0.6541,
+      "step": 4639
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027595989740604015,
+      "loss": 0.4559,
+      "step": 4640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002759495893502775,
+      "loss": 0.752,
+      "step": 4641
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027593927927762116,
+      "loss": 1.313,
+      "step": 4642
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027592896718823634,
+      "loss": 0.4693,
+      "step": 4643
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027591865308228806,
+      "loss": 0.3868,
+      "step": 4644
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027590833695994156,
+      "loss": 0.5622,
+      "step": 4645
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027589801882136196,
+      "loss": 0.2214,
+      "step": 4646
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027588769866671454,
+      "loss": 0.2573,
+      "step": 4647
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002758773764961646,
+      "loss": 0.5488,
+      "step": 4648
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027586705230987736,
+      "loss": 0.2722,
+      "step": 4649
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027585672610801816,
+      "loss": 0.7463,
+      "step": 4650
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002758463978907524,
+      "loss": 0.362,
+      "step": 4651
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002758360676582455,
+      "loss": 0.2928,
+      "step": 4652
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027582573541066276,
+      "loss": 1.4579,
+      "step": 4653
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002758154011481697,
+      "loss": 0.791,
+      "step": 4654
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002758050648709318,
+      "loss": 0.4922,
+      "step": 4655
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002757947265791147,
+      "loss": 0.8339,
+      "step": 4656
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027578438627288374,
+      "loss": 0.294,
+      "step": 4657
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027577404395240466,
+      "loss": 0.9584,
+      "step": 4658
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275763699617843,
+      "loss": 0.4774,
+      "step": 4659
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027575335326936445,
+      "loss": 0.5848,
+      "step": 4660
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027574300490713465,
+      "loss": 0.5713,
+      "step": 4661
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027573265453131933,
+      "loss": 0.4393,
+      "step": 4662
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002757223021420843,
+      "loss": 0.4246,
+      "step": 4663
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002757119477395953,
+      "loss": 0.2423,
+      "step": 4664
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027570159132401807,
+      "loss": 0.069,
+      "step": 4665
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027569123289551846,
+      "loss": 0.8783,
+      "step": 4666
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027568087245426247,
+      "loss": 0.0685,
+      "step": 4667
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002756705100004159,
+      "loss": 0.0695,
+      "step": 4668
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027566014553414474,
+      "loss": 0.6239,
+      "step": 4669
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027564977905561493,
+      "loss": 0.5238,
+      "step": 4670
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002756394105649925,
+      "loss": 0.1801,
+      "step": 4671
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027562904006244343,
+      "loss": 1.2727,
+      "step": 4672
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027561866754813384,
+      "loss": 0.4046,
+      "step": 4673
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002756082930222298,
+      "loss": 0.0415,
+      "step": 4674
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002755979164848975,
+      "loss": 0.3237,
+      "step": 4675
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275587537936303,
+      "loss": 0.6173,
+      "step": 4676
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027557715737661266,
+      "loss": 1.125,
+      "step": 4677
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002755667748059925,
+      "loss": 0.8653,
+      "step": 4678
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027555639022460896,
+      "loss": 0.6549,
+      "step": 4679
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027554600363262826,
+      "loss": 0.4592,
+      "step": 4680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027553561503021673,
+      "loss": 0.012,
+      "step": 4681
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027552522441754075,
+      "loss": 0.131,
+      "step": 4682
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002755148317947666,
+      "loss": 0.9353,
+      "step": 4683
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002755044371620609,
+      "loss": 0.6177,
+      "step": 4684
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002754940405195899,
+      "loss": 0.5473,
+      "step": 4685
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002754836418675202,
+      "loss": 0.7061,
+      "step": 4686
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002754732412060183,
+      "loss": 0.4406,
+      "step": 4687
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002754628385352508,
+      "loss": 0.7043,
+      "step": 4688
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027545243385538416,
+      "loss": 0.6108,
+      "step": 4689
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027544202716658515,
+      "loss": 0.5092,
+      "step": 4690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027543161846902027,
+      "loss": 0.6531,
+      "step": 4691
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002754212077628563,
+      "loss": 0.2792,
+      "step": 4692
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027541079504825994,
+      "loss": 0.0057,
+      "step": 4693
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002754003803253979,
+      "loss": 0.9297,
+      "step": 4694
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027538996359443695,
+      "loss": 3.6166,
+      "step": 4695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027537954485554395,
+      "loss": 4.293,
+      "step": 4696
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027536912410888563,
+      "loss": 0.5778,
+      "step": 4697
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275358701354629,
+      "loss": 0.5039,
+      "step": 4698
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002753482765929409,
+      "loss": 0.012,
+      "step": 4699
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027533784982398826,
+      "loss": 0.618,
+      "step": 4700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027532742104793807,
+      "loss": 0.7735,
+      "step": 4701
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027531699026495733,
+      "loss": 0.1862,
+      "step": 4702
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002753065574752131,
+      "loss": 1.042,
+      "step": 4703
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027529612267887234,
+      "loss": 1.1202,
+      "step": 4704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027528568587610223,
+      "loss": 0.2502,
+      "step": 4705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027527524706706994,
+      "loss": 0.0444,
+      "step": 4706
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027526480625194254,
+      "loss": 0.967,
+      "step": 4707
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002752543634308872,
+      "loss": 0.2039,
+      "step": 4708
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027524391860407133,
+      "loss": 0.4194,
+      "step": 4709
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275233471771662,
+      "loss": 0.9743,
+      "step": 4710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002752230229338266,
+      "loss": 0.4346,
+      "step": 4711
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002752125720907324,
+      "loss": 0.8012,
+      "step": 4712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027520211924254677,
+      "loss": 1.4912,
+      "step": 4713
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002751916643894371,
+      "loss": 0.2543,
+      "step": 4714
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027518120753157086,
+      "loss": 0.4594,
+      "step": 4715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002751707486691154,
+      "loss": 0.9184,
+      "step": 4716
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002751602878022383,
+      "loss": 0.2917,
+      "step": 4717
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000275149824931107,
+      "loss": 0.6251,
+      "step": 4718
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027513936005588906,
+      "loss": 0.3292,
+      "step": 4719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027512889317675215,
+      "loss": 0.4515,
+      "step": 4720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002751184242938637,
+      "loss": 2.0607,
+      "step": 4721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002751079534073915,
+      "loss": 0.6278,
+      "step": 4722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002750974805175032,
+      "loss": 0.1351,
+      "step": 4723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027508700562436656,
+      "loss": 0.3836,
+      "step": 4724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002750765287281492,
+      "loss": 0.2381,
+      "step": 4725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002750660498290189,
+      "loss": 0.4457,
+      "step": 4726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027505556892714353,
+      "loss": 0.6027,
+      "step": 4727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027504508602269094,
+      "loss": 0.4864,
+      "step": 4728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027503460111582894,
+      "loss": 0.0041,
+      "step": 4729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027502411420672547,
+      "loss": 0.14,
+      "step": 4730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002750136252955484,
+      "loss": 2.817,
+      "step": 4731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027500313438246583,
+      "loss": 0.0493,
+      "step": 4732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002749926414676456,
+      "loss": 0.1506,
+      "step": 4733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027498214655125584,
+      "loss": 0.0587,
+      "step": 4734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027497164963346453,
+      "loss": 0.1424,
+      "step": 4735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027496115071443984,
+      "loss": 0.3757,
+      "step": 4736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002749506497943499,
+      "loss": 0.5527,
+      "step": 4737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027494014687336275,
+      "loss": 0.1445,
+      "step": 4738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027492964195164666,
+      "loss": 0.1132,
+      "step": 4739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002749191350293699,
+      "loss": 0.7832,
+      "step": 4740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002749086261067006,
+      "loss": 0.7999,
+      "step": 4741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027489811518380714,
+      "loss": 0.0719,
+      "step": 4742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027488760226085786,
+      "loss": 0.7838,
+      "step": 4743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000274877087338021,
+      "loss": 0.2902,
+      "step": 4744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027486657041546505,
+      "loss": 2.3262,
+      "step": 4745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002748560514933584,
+      "loss": 0.0038,
+      "step": 4746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002748455305718694,
+      "loss": 1.9088,
+      "step": 4747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002748350076511666,
+      "loss": 1.0031,
+      "step": 4748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027482448273141854,
+      "loss": 0.7066,
+      "step": 4749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027481395581279366,
+      "loss": 0.0082,
+      "step": 4750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027480342689546066,
+      "loss": 0.2861,
+      "step": 4751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002747928959795881,
+      "loss": 1.064,
+      "step": 4752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027478236306534455,
+      "loss": 0.3487,
+      "step": 4753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027477182815289873,
+      "loss": 0.1907,
+      "step": 4754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027476129124241934,
+      "loss": 0.5524,
+      "step": 4755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002747507523340751,
+      "loss": 0.2912,
+      "step": 4756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027474021142803484,
+      "loss": 0.2136,
+      "step": 4757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027472966852446726,
+      "loss": 0.7964,
+      "step": 4758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027471912362354124,
+      "loss": 0.7936,
+      "step": 4759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027470857672542564,
+      "loss": 1.2169,
+      "step": 4760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002746980278302893,
+      "loss": 0.2389,
+      "step": 4761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027468747693830127,
+      "loss": 0.7077,
+      "step": 4762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027467692404963034,
+      "loss": 1.2584,
+      "step": 4763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027466636916444567,
+      "loss": 0.9652,
+      "step": 4764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002746558122829161,
+      "loss": 2.6954,
+      "step": 4765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002746452534052109,
+      "loss": 0.3592,
+      "step": 4766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027463469253149894,
+      "loss": 0.1239,
+      "step": 4767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002746241296619494,
+      "loss": 0.2166,
+      "step": 4768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027461356479673157,
+      "loss": 0.8415,
+      "step": 4769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027460299793601446,
+      "loss": 0.4783,
+      "step": 4770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027459242907996735,
+      "loss": 0.8305,
+      "step": 4771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027458185822875947,
+      "loss": 1.3848,
+      "step": 4772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002745712853825601,
+      "loss": 0.0085,
+      "step": 4773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027456071054153856,
+      "loss": 1.0671,
+      "step": 4774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002745501337058642,
+      "loss": 2.9146,
+      "step": 4775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002745395548757063,
+      "loss": 1.2473,
+      "step": 4776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002745289740512344,
+      "loss": 0.7451,
+      "step": 4777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002745183912326178,
+      "loss": 0.4809,
+      "step": 4778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027450780642002614,
+      "loss": 1.3155,
+      "step": 4779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002744972196136287,
+      "loss": 3.3868,
+      "step": 4780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027448663081359526,
+      "loss": 0.5588,
+      "step": 4781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002744760400200952,
+      "loss": 0.0057,
+      "step": 4782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027446544723329813,
+      "loss": 0.2865,
+      "step": 4783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002744548524533738,
+      "loss": 0.5808,
+      "step": 4784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027444425568049174,
+      "loss": 0.5499,
+      "step": 4785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027443365691482167,
+      "loss": 0.4741,
+      "step": 4786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002744230561565334,
+      "loss": 0.5561,
+      "step": 4787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002744124534057966,
+      "loss": 0.4342,
+      "step": 4788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027440184866278107,
+      "loss": 0.7738,
+      "step": 4789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743912419276566,
+      "loss": 0.0334,
+      "step": 4790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743806332005931,
+      "loss": 0.0221,
+      "step": 4791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743700224817604,
+      "loss": 0.4243,
+      "step": 4792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743594097713285,
+      "loss": 0.1018,
+      "step": 4793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027434879506946724,
+      "loss": 0.0681,
+      "step": 4794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743381783763467,
+      "loss": 0.3346,
+      "step": 4795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027432755969213676,
+      "loss": 0.8675,
+      "step": 4796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743169390170076,
+      "loss": 6.9629,
+      "step": 4797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002743063163511292,
+      "loss": 0.0362,
+      "step": 4798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027429569169467173,
+      "loss": 0.8943,
+      "step": 4799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027428506504780534,
+      "loss": 0.622,
+      "step": 4800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002742744364107001,
+      "loss": 0.0377,
+      "step": 4801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002742638057835263,
+      "loss": 0.0243,
+      "step": 4802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027425317316645407,
+      "loss": 0.2744,
+      "step": 4803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027424253855965386,
+      "loss": 0.5951,
+      "step": 4804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027423190196329577,
+      "loss": 0.4969,
+      "step": 4805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027422126337755024,
+      "loss": 1.6406,
+      "step": 4806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002742106228025876,
+      "loss": 0.5822,
+      "step": 4807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741999802385783,
+      "loss": 0.0178,
+      "step": 4808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027418933568569266,
+      "loss": 0.2156,
+      "step": 4809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741786891441012,
+      "loss": 0.4591,
+      "step": 4810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741680406139744,
+      "loss": 0.3907,
+      "step": 4811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741573900954828,
+      "loss": 0.8634,
+      "step": 4812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741467375887969,
+      "loss": 0.0036,
+      "step": 4813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741360830940874,
+      "loss": 0.0086,
+      "step": 4814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027412542661152476,
+      "loss": 0.266,
+      "step": 4815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741147681412797,
+      "loss": 1.0233,
+      "step": 4816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002741041076835229,
+      "loss": 0.7314,
+      "step": 4817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027409344523842506,
+      "loss": 0.4349,
+      "step": 4818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027408278080615696,
+      "loss": 0.7939,
+      "step": 4819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002740721143868893,
+      "loss": 0.6234,
+      "step": 4820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000274061445980793,
+      "loss": 0.1688,
+      "step": 4821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027405077558803885,
+      "loss": 0.5925,
+      "step": 4822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002740401032087976,
+      "loss": 0.0503,
+      "step": 4823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027402942884324034,
+      "loss": 0.0141,
+      "step": 4824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002740187524915379,
+      "loss": 1.665,
+      "step": 4825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027400807415386123,
+      "loss": 0.8385,
+      "step": 4826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002739973938303814,
+      "loss": 2.6535,
+      "step": 4827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002739867115212694,
+      "loss": 0.7462,
+      "step": 4828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027397602722669626,
+      "loss": 0.2765,
+      "step": 4829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002739653409468332,
+      "loss": 0.5416,
+      "step": 4830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027395465268185114,
+      "loss": 0.8032,
+      "step": 4831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027394396243192134,
+      "loss": 0.4546,
+      "step": 4832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027393327019721505,
+      "loss": 0.642,
+      "step": 4833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027392257597790346,
+      "loss": 0.2761,
+      "step": 4834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027391187977415777,
+      "loss": 0.3131,
+      "step": 4835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002739011815861493,
+      "loss": 4.2344,
+      "step": 4836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027389048141404935,
+      "loss": 0.6007,
+      "step": 4837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027387977925802924,
+      "loss": 0.7195,
+      "step": 4838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002738690751182604,
+      "loss": 1.0042,
+      "step": 4839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027385836899491425,
+      "loss": 0.7874,
+      "step": 4840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027384766088816214,
+      "loss": 0.3711,
+      "step": 4841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002738369507981757,
+      "loss": 0.2568,
+      "step": 4842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002738262387251263,
+      "loss": 0.5631,
+      "step": 4843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027381552466918555,
+      "loss": 0.0066,
+      "step": 4844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027380480863052495,
+      "loss": 0.2871,
+      "step": 4845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027379409060931617,
+      "loss": 0.8681,
+      "step": 4846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027378337060573083,
+      "loss": 0.8178,
+      "step": 4847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027377264861994056,
+      "loss": 1.1026,
+      "step": 4848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002737619246521171,
+      "loss": 0.341,
+      "step": 4849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027375119870243214,
+      "loss": 0.0097,
+      "step": 4850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027374047077105746,
+      "loss": 0.9279,
+      "step": 4851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027372974085816486,
+      "loss": 0.2899,
+      "step": 4852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027371900896392613,
+      "loss": 1.5195,
+      "step": 4853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027370827508851317,
+      "loss": 0.6507,
+      "step": 4854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002736975392320978,
+      "loss": 0.0165,
+      "step": 4855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027368680139485204,
+      "loss": 1.4561,
+      "step": 4856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002736760615769478,
+      "loss": 0.328,
+      "step": 4857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027366531977855696,
+      "loss": 0.5408,
+      "step": 4858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002736545759998517,
+      "loss": 0.5016,
+      "step": 4859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00027364383024100394,
+      "loss": 1.502,
+      "step": 4860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027363308250218576,
+      "loss": 1.2394,
+      "step": 4861
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002736223327835694,
+      "loss": 0.914,
+      "step": 4862
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002736115810853269,
+      "loss": 0.9011,
+      "step": 4863
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002736008274076304,
+      "loss": 0.5735,
+      "step": 4864
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002735900717506522,
+      "loss": 0.6193,
+      "step": 4865
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002735793141145644,
+      "loss": 0.4128,
+      "step": 4866
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002735685544995394,
+      "loss": 1.4158,
+      "step": 4867
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002735577929057495,
+      "loss": 0.3083,
+      "step": 4868
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002735470293333669,
+      "loss": 0.0344,
+      "step": 4869
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002735362637825641,
+      "loss": 0.6275,
+      "step": 4870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027352549625351343,
+      "loss": 0.449,
+      "step": 4871
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027351472674638734,
+      "loss": 0.0046,
+      "step": 4872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002735039552613582,
+      "loss": 0.0423,
+      "step": 4873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002734931817985986,
+      "loss": 0.5048,
+      "step": 4874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002734824063582811,
+      "loss": 0.1596,
+      "step": 4875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027347162894057807,
+      "loss": 0.7776,
+      "step": 4876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002734608495456622,
+      "loss": 0.646,
+      "step": 4877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027345006817370627,
+      "loss": 0.0373,
+      "step": 4878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027343928482488264,
+      "loss": 0.4276,
+      "step": 4879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027342849949936416,
+      "loss": 0.6492,
+      "step": 4880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002734177121973235,
+      "loss": 2.0039,
+      "step": 4881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002734069229189334,
+      "loss": 1.2055,
+      "step": 4882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002733961316643666,
+      "loss": 0.9631,
+      "step": 4883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000273385338433796,
+      "loss": 0.005,
+      "step": 4884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002733745432273944,
+      "loss": 0.4519,
+      "step": 4885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027336374604533455,
+      "loss": 0.2222,
+      "step": 4886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002733529468877895,
+      "loss": 1.114,
+      "step": 4887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002733421457549322,
+      "loss": 0.2547,
+      "step": 4888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002733313426469355,
+      "loss": 0.491,
+      "step": 4889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027332053756397236,
+      "loss": 0.0197,
+      "step": 4890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000273309730506216,
+      "loss": 0.3727,
+      "step": 4891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027329892147383934,
+      "loss": 0.1748,
+      "step": 4892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002732881104670155,
+      "loss": 0.0126,
+      "step": 4893
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027327729748591757,
+      "loss": 0.3502,
+      "step": 4894
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027326648253071876,
+      "loss": 0.8976,
+      "step": 4895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002732556656015922,
+      "loss": 0.127,
+      "step": 4896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027324484669871116,
+      "loss": 0.2634,
+      "step": 4897
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002732340258222489,
+      "loss": 0.0018,
+      "step": 4898
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002732232029723786,
+      "loss": 0.5519,
+      "step": 4899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002732123781492736,
+      "loss": 0.9941,
+      "step": 4900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027320155135310736,
+      "loss": 1.0035,
+      "step": 4901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027319072258405317,
+      "loss": 0.8511,
+      "step": 4902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027317989184228445,
+      "loss": 0.2233,
+      "step": 4903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027316905912797457,
+      "loss": 0.5012,
+      "step": 4904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027315822444129716,
+      "loss": 0.2864,
+      "step": 4905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027314738778242553,
+      "loss": 2.457,
+      "step": 4906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002731365491515333,
+      "loss": 0.6246,
+      "step": 4907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027312570854879406,
+      "loss": 0.4109,
+      "step": 4908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002731148659743814,
+      "loss": 0.0027,
+      "step": 4909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027310402142846895,
+      "loss": 2.127,
+      "step": 4910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730931749112303,
+      "loss": 0.9322,
+      "step": 4911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027308232642283924,
+      "loss": 0.6542,
+      "step": 4912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730714759634694,
+      "loss": 0.7795,
+      "step": 4913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730606235332946,
+      "loss": 0.8784,
+      "step": 4914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730497691324886,
+      "loss": 2.2207,
+      "step": 4915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730389127612252,
+      "loss": 0.426,
+      "step": 4916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730280544196783,
+      "loss": 0.0403,
+      "step": 4917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730171941080217,
+      "loss": 1.2227,
+      "step": 4918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002730063318264294,
+      "loss": 0.2917,
+      "step": 4919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027299546757507534,
+      "loss": 0.0301,
+      "step": 4920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027298460135413336,
+      "loss": 0.4721,
+      "step": 4921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002729737331637776,
+      "loss": 1.0159,
+      "step": 4922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002729628630041821,
+      "loss": 0.0467,
+      "step": 4923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027295199087552086,
+      "loss": 0.8412,
+      "step": 4924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000272941116777968,
+      "loss": 1.5362,
+      "step": 4925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027293024071169765,
+      "loss": 0.3249,
+      "step": 4926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000272919362676884,
+      "loss": 0.2499,
+      "step": 4927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002729084826737013,
+      "loss": 0.296,
+      "step": 4928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027289760070232363,
+      "loss": 1.2305,
+      "step": 4929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027288671676292534,
+      "loss": 0.3735,
+      "step": 4930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027287583085568076,
+      "loss": 0.6428,
+      "step": 4931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027286494298076416,
+      "loss": 0.4776,
+      "step": 4932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027285405313834984,
+      "loss": 0.2786,
+      "step": 4933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027284316132861224,
+      "loss": 0.4104,
+      "step": 4934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027283226755172576,
+      "loss": 0.5358,
+      "step": 4935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002728213718078649,
+      "loss": 0.6052,
+      "step": 4936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027281047409720405,
+      "loss": 0.0271,
+      "step": 4937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002727995744199179,
+      "loss": 0.4802,
+      "step": 4938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027278867277618073,
+      "loss": 0.5239,
+      "step": 4939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027277776916616726,
+      "loss": 0.4303,
+      "step": 4940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027276686359005214,
+      "loss": 0.0595,
+      "step": 4941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027275595604800987,
+      "loss": 0.3832,
+      "step": 4942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002727450465402152,
+      "loss": 0.8376,
+      "step": 4943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002727341350668429,
+      "loss": 0.4991,
+      "step": 4944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027272322162806757,
+      "loss": 0.3574,
+      "step": 4945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027271230622406405,
+      "loss": 1.6417,
+      "step": 4946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027270138885500714,
+      "loss": 0.4026,
+      "step": 4947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027269046952107156,
+      "loss": 0.8391,
+      "step": 4948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002726795482224323,
+      "loss": 0.3403,
+      "step": 4949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002726686249592642,
+      "loss": 0.0131,
+      "step": 4950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002726576997317421,
+      "loss": 0.1545,
+      "step": 4951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002726467725400411,
+      "loss": 0.1672,
+      "step": 4952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002726358433843361,
+      "loss": 1.9727,
+      "step": 4953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027262491226480207,
+      "loss": 0.4103,
+      "step": 4954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002726139791816141,
+      "loss": 0.2517,
+      "step": 4955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027260304413494733,
+      "loss": 0.266,
+      "step": 4956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027259210712497674,
+      "loss": 0.2906,
+      "step": 4957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002725811681518776,
+      "loss": 0.4815,
+      "step": 4958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000272570227215825,
+      "loss": 0.0149,
+      "step": 4959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002725592843169941,
+      "loss": 0.0135,
+      "step": 4960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002725483394555603,
+      "loss": 0.3552,
+      "step": 4961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027253739263169874,
+      "loss": 0.2171,
+      "step": 4962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027252644384558473,
+      "loss": 0.2128,
+      "step": 4963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027251549309739357,
+      "loss": 0.4463,
+      "step": 4964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002725045403873007,
+      "loss": 0.5123,
+      "step": 4965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027249358571548144,
+      "loss": 0.4418,
+      "step": 4966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027248262908211127,
+      "loss": 0.7798,
+      "step": 4967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002724716704873656,
+      "loss": 0.9526,
+      "step": 4968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027246070993141994,
+      "loss": 1.4038,
+      "step": 4969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002724497474144498,
+      "loss": 0.3216,
+      "step": 4970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027243878293663074,
+      "loss": 0.4654,
+      "step": 4971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002724278164981383,
+      "loss": 0.0053,
+      "step": 4972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002724168480991481,
+      "loss": 0.0422,
+      "step": 4973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002724058777398358,
+      "loss": 0.4291,
+      "step": 4974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723949054203771,
+      "loss": 0.2068,
+      "step": 4975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723839311409477,
+      "loss": 0.3646,
+      "step": 4976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723729549017233,
+      "loss": 0.3246,
+      "step": 4977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723619767028797,
+      "loss": 0.4006,
+      "step": 4978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723509965445926,
+      "loss": 0.0094,
+      "step": 4979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000272340014427038,
+      "loss": 0.5591,
+      "step": 4980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002723290303503916,
+      "loss": 0.8188,
+      "step": 4981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027231804431482944,
+      "loss": 0.8479,
+      "step": 4982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027230705632052735,
+      "loss": 0.0123,
+      "step": 4983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722960663676613,
+      "loss": 0.7712,
+      "step": 4984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722850744564073,
+      "loss": 1.0293,
+      "step": 4985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722740805869414,
+      "loss": 0.7461,
+      "step": 4986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722630847594395,
+      "loss": 0.4057,
+      "step": 4987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027225208697407785,
+      "loss": 0.1177,
+      "step": 4988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722410872310325,
+      "loss": 0.4867,
+      "step": 4989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027223008553047966,
+      "loss": 1.1861,
+      "step": 4990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722190818725953,
+      "loss": 0.1751,
+      "step": 4991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002722080762575559,
+      "loss": 0.6969,
+      "step": 4992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002721970686855375,
+      "loss": 1.5541,
+      "step": 4993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027218605915671647,
+      "loss": 0.0031,
+      "step": 4994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002721750476712691,
+      "loss": 1.969,
+      "step": 4995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002721640342293716,
+      "loss": 1.442,
+      "step": 4996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002721530188312005,
+      "loss": 0.2917,
+      "step": 4997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027214200147693214,
+      "loss": 0.3055,
+      "step": 4998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027213098216674294,
+      "loss": 0.2686,
+      "step": 4999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002721199609008093,
+      "loss": 0.215,
+      "step": 5000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002721089376793078,
+      "loss": 0.414,
+      "step": 5001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027209791250241495,
+      "loss": 0.0051,
+      "step": 5002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027208688537030723,
+      "loss": 0.7594,
+      "step": 5003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002720758562831613,
+      "loss": 0.6186,
+      "step": 5004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002720648252411537,
+      "loss": 0.0332,
+      "step": 5005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027205379224446113,
+      "loss": 1.767,
+      "step": 5006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002720427572932603,
+      "loss": 0.6813,
+      "step": 5007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002720317203877278,
+      "loss": 0.3675,
+      "step": 5008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027202068152804053,
+      "loss": 0.6247,
+      "step": 5009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002720096407143751,
+      "loss": 0.7072,
+      "step": 5010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719985979469084,
+      "loss": 0.5464,
+      "step": 5011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027198755322581727,
+      "loss": 0.0139,
+      "step": 5012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719765065512785,
+      "loss": 0.3164,
+      "step": 5013
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719654579234691,
+      "loss": 1.0372,
+      "step": 5014
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719544073425659,
+      "loss": 0.4983,
+      "step": 5015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027194335480874594,
+      "loss": 0.734,
+      "step": 5016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027193230032218613,
+      "loss": 0.7844,
+      "step": 5017
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027192124388306354,
+      "loss": 0.9148,
+      "step": 5018
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002719101854915552,
+      "loss": 0.4116,
+      "step": 5019
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027189912514783814,
+      "loss": 0.8859,
+      "step": 5020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002718880628520896,
+      "loss": 0.2659,
+      "step": 5021
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002718769986044867,
+      "loss": 2.3031,
+      "step": 5022
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027186593240520647,
+      "loss": 0.6975,
+      "step": 5023
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027185486425442636,
+      "loss": 0.3944,
+      "step": 5024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027184379415232335,
+      "loss": 0.727,
+      "step": 5025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002718327220990749,
+      "loss": 0.2972,
+      "step": 5026
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027182164809485827,
+      "loss": 0.3164,
+      "step": 5027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027181057213985073,
+      "loss": 0.4184,
+      "step": 5028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027179949423422975,
+      "loss": 1.7531,
+      "step": 5029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002717884143781726,
+      "loss": 0.6555,
+      "step": 5030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002717773325718568,
+      "loss": 0.3719,
+      "step": 5031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027176624881545977,
+      "loss": 0.4932,
+      "step": 5032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027175516310915904,
+      "loss": 1.0633,
+      "step": 5033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027174407545313206,
+      "loss": 0.0032,
+      "step": 5034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002717329858475565,
+      "loss": 4.0898,
+      "step": 5035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027172189429260974,
+      "loss": 0.0055,
+      "step": 5036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027171080078846965,
+      "loss": 0.5547,
+      "step": 5037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002716997053353136,
+      "loss": 0.0311,
+      "step": 5038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002716886079333196,
+      "loss": 0.3363,
+      "step": 5039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002716775085826651,
+      "loss": 0.4997,
+      "step": 5040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027166640728352786,
+      "loss": 0.2984,
+      "step": 5041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002716553040360857,
+      "loss": 0.0046,
+      "step": 5042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027164419884051646,
+      "loss": 0.0167,
+      "step": 5043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027163309169699794,
+      "loss": 0.0781,
+      "step": 5044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000271621982605708,
+      "loss": 0.3799,
+      "step": 5045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027161087156682455,
+      "loss": 0.528,
+      "step": 5046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002715997585805255,
+      "loss": 0.0022,
+      "step": 5047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002715886436469888,
+      "loss": 0.1259,
+      "step": 5048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027157752676639245,
+      "loss": 0.7868,
+      "step": 5049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027156640793891447,
+      "loss": 0.3035,
+      "step": 5050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027155528716473295,
+      "loss": 0.3669,
+      "step": 5051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027154416444402586,
+      "loss": 0.7415,
+      "step": 5052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027153303977697145,
+      "loss": 0.0027,
+      "step": 5053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002715219131637478,
+      "loss": 0.1052,
+      "step": 5054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027151078460453307,
+      "loss": 0.2668,
+      "step": 5055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002714996540995055,
+      "loss": 0.5208,
+      "step": 5056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002714885216488433,
+      "loss": 0.4799,
+      "step": 5057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002714773872527248,
+      "loss": 1.1736,
+      "step": 5058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027146625091132817,
+      "loss": 2.9192,
+      "step": 5059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027145511262483186,
+      "loss": 0.0083,
+      "step": 5060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002714439723934143,
+      "loss": 0.6565,
+      "step": 5061
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002714328302172536,
+      "loss": 0.4844,
+      "step": 5062
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027142168609652854,
+      "loss": 0.7202,
+      "step": 5063
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002714105400314173,
+      "loss": 1.548,
+      "step": 5064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002713993920220985,
+      "loss": 0.5505,
+      "step": 5065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027138824206875063,
+      "loss": 0.1686,
+      "step": 5066
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027137709017155225,
+      "loss": 1.1727,
+      "step": 5067
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027136593633068196,
+      "loss": 0.4322,
+      "step": 5068
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002713547805463183,
+      "loss": 0.3421,
+      "step": 5069
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027134362281863993,
+      "loss": 0.0198,
+      "step": 5070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027133246314782564,
+      "loss": 0.0023,
+      "step": 5071
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000271321301534054,
+      "loss": 0.3839,
+      "step": 5072
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002713101379775038,
+      "loss": 0.4803,
+      "step": 5073
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002712989724783538,
+      "loss": 0.3562,
+      "step": 5074
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002712878050367828,
+      "loss": 0.2093,
+      "step": 5075
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027127663565296964,
+      "loss": 0.4603,
+      "step": 5076
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027126546432709313,
+      "loss": 0.249,
+      "step": 5077
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027125429105933225,
+      "loss": 0.5766,
+      "step": 5078
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002712431158498659,
+      "loss": 0.7111,
+      "step": 5079
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002712319386988729,
+      "loss": 0.3979,
+      "step": 5080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027122075960653243,
+      "loss": 0.0052,
+      "step": 5081
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002712095785730234,
+      "loss": 0.3125,
+      "step": 5082
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027119839559852486,
+      "loss": 0.594,
+      "step": 5083
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000271187210683216,
+      "loss": 0.3727,
+      "step": 5084
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027117602382727574,
+      "loss": 0.0031,
+      "step": 5085
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027116483503088337,
+      "loss": 0.888,
+      "step": 5086
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027115364429421796,
+      "loss": 0.6689,
+      "step": 5087
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002711424516174588,
+      "loss": 0.2603,
+      "step": 5088
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027113125700078505,
+      "loss": 0.0134,
+      "step": 5089
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002711200604443761,
+      "loss": 0.5516,
+      "step": 5090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002711088619484111,
+      "loss": 0.3815,
+      "step": 5091
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002710976615130694,
+      "loss": 0.3958,
+      "step": 5092
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002710864591385304,
+      "loss": 0.5315,
+      "step": 5093
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027107525482497353,
+      "loss": 0.3884,
+      "step": 5094
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002710640485725781,
+      "loss": 0.4723,
+      "step": 5095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002710528403815237,
+      "loss": 0.3273,
+      "step": 5096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027104163025198967,
+      "loss": 0.2883,
+      "step": 5097
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002710304181841556,
+      "loss": 0.0564,
+      "step": 5098
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027101920417820104,
+      "loss": 0.3169,
+      "step": 5099
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027100798823430556,
+      "loss": 1.3761,
+      "step": 5100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027099677035264877,
+      "loss": 0.8277,
+      "step": 5101
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027098555053341026,
+      "loss": 0.7294,
+      "step": 5102
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027097432877676975,
+      "loss": 0.6623,
+      "step": 5103
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002709631050829069,
+      "loss": 0.5063,
+      "step": 5104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027095187945200147,
+      "loss": 0.916,
+      "step": 5105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027094065188423323,
+      "loss": 3.0999,
+      "step": 5106
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027092942237978197,
+      "loss": 0.1649,
+      "step": 5107
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00027091819093882745,
+      "loss": 0.7795,
+      "step": 5108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002709069575615496,
+      "loss": 0.3781,
+      "step": 5109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027089572224812833,
+      "loss": 0.702,
+      "step": 5110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027088448499874347,
+      "loss": 0.3973,
+      "step": 5111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000270873245813575,
+      "loss": 1.8497,
+      "step": 5112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027086200469280286,
+      "loss": 0.0541,
+      "step": 5113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027085076163660713,
+      "loss": 1.1935,
+      "step": 5114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002708395166451678,
+      "loss": 1.2148,
+      "step": 5115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000270828269718665,
+      "loss": 2.4962,
+      "step": 5116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027081702085727887,
+      "loss": 0.4765,
+      "step": 5117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002708057700611894,
+      "loss": 0.2961,
+      "step": 5118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002707945173305768,
+      "loss": 0.9672,
+      "step": 5119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002707832626656213,
+      "loss": 0.9065,
+      "step": 5120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027077200606650316,
+      "loss": 0.0646,
+      "step": 5121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027076074753340257,
+      "loss": 0.8794,
+      "step": 5122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027074948706649985,
+      "loss": 0.4423,
+      "step": 5123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027073822466597535,
+      "loss": 0.8058,
+      "step": 5124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027072696033200933,
+      "loss": 0.7195,
+      "step": 5125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027071569406478225,
+      "loss": 0.6618,
+      "step": 5126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027070442586447446,
+      "loss": 0.3325,
+      "step": 5127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002706931557312665,
+      "loss": 0.2873,
+      "step": 5128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027068188366533876,
+      "loss": 0.017,
+      "step": 5129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002706706096668718,
+      "loss": 0.2939,
+      "step": 5130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002706593337360461,
+      "loss": 0.5671,
+      "step": 5131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027064805587304226,
+      "loss": 0.3266,
+      "step": 5132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027063677607804085,
+      "loss": 1.0926,
+      "step": 5133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027062549435122253,
+      "loss": 0.7719,
+      "step": 5134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000270614210692768,
+      "loss": 0.552,
+      "step": 5135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002706029251028578,
+      "loss": 1.3692,
+      "step": 5136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002705916375816728,
+      "loss": 0.2909,
+      "step": 5137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002705803481293938,
+      "loss": 0.462,
+      "step": 5138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002705690567462014,
+      "loss": 0.4587,
+      "step": 5139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002705577634322765,
+      "loss": 0.4725,
+      "step": 5140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002705464681878,
+      "loss": 0.595,
+      "step": 5141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027053517101295263,
+      "loss": 0.7256,
+      "step": 5142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002705238719079155,
+      "loss": 0.3883,
+      "step": 5143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002705125708728694,
+      "loss": 0.4414,
+      "step": 5144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027050126790799535,
+      "loss": 0.0295,
+      "step": 5145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002704899630134743,
+      "loss": 0.3996,
+      "step": 5146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002704786561894874,
+      "loss": 0.2559,
+      "step": 5147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027046734743621563,
+      "loss": 1.1359,
+      "step": 5148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027045603675384,
+      "loss": 0.6631,
+      "step": 5149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002704447241425418,
+      "loss": 0.3537,
+      "step": 5150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000270433409602502,
+      "loss": 0.2166,
+      "step": 5151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027042209313390205,
+      "loss": 0.2566,
+      "step": 5152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002704107747369229,
+      "loss": 0.9312,
+      "step": 5153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002703994544117459,
+      "loss": 0.4554,
+      "step": 5154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027038813215855237,
+      "loss": 0.3764,
+      "step": 5155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027037680797752355,
+      "loss": 0.4714,
+      "step": 5156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002703654818688409,
+      "loss": 0.2031,
+      "step": 5157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027035415383268557,
+      "loss": 1.1316,
+      "step": 5158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002703428238692392,
+      "loss": 0.966,
+      "step": 5159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027033149197868306,
+      "loss": 0.4395,
+      "step": 5160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027032015816119866,
+      "loss": 0.0311,
+      "step": 5161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002703088224169675,
+      "loss": 0.4664,
+      "step": 5162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027029748474617117,
+      "loss": 0.5671,
+      "step": 5163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027028614514899114,
+      "loss": 3.0045,
+      "step": 5164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027027480362560896,
+      "loss": 0.8462,
+      "step": 5165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027026346017620636,
+      "loss": 0.3459,
+      "step": 5166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002702521148009649,
+      "loss": 0.154,
+      "step": 5167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002702407675000663,
+      "loss": 0.0469,
+      "step": 5168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027022941827369235,
+      "loss": 0.6406,
+      "step": 5169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002702180671220246,
+      "loss": 1.0015,
+      "step": 5170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027020671404524496,
+      "loss": 2.3184,
+      "step": 5171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002701953590435352,
+      "loss": 0.137,
+      "step": 5172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027018400211707716,
+      "loss": 0.1569,
+      "step": 5173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002701726432660527,
+      "loss": 0.149,
+      "step": 5174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027016128249064366,
+      "loss": 0.0115,
+      "step": 5175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002701499197910321,
+      "loss": 0.3667,
+      "step": 5176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027013855516739985,
+      "loss": 0.5227,
+      "step": 5177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027012718861992897,
+      "loss": 0.4261,
+      "step": 5178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002701158201488014,
+      "loss": 0.3587,
+      "step": 5179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027010444975419934,
+      "loss": 1.5499,
+      "step": 5180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027009307743630466,
+      "loss": 0.2272,
+      "step": 5181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002700817031952996,
+      "loss": 0.4869,
+      "step": 5182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002700703270313663,
+      "loss": 0.2982,
+      "step": 5183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002700589489446869,
+      "loss": 0.7136,
+      "step": 5184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002700475689354436,
+      "loss": 0.4295,
+      "step": 5185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002700361870038187,
+      "loss": 1.0464,
+      "step": 5186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027002480314999443,
+      "loss": 0.6562,
+      "step": 5187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000270013417374153,
+      "loss": 0.6703,
+      "step": 5188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00027000202967647686,
+      "loss": 0.7066,
+      "step": 5189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026999064005714824,
+      "loss": 0.6726,
+      "step": 5190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002699792485163497,
+      "loss": 1.5038,
+      "step": 5191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002699678550542635,
+      "loss": 0.6065,
+      "step": 5192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002699564596710721,
+      "loss": 2.1863,
+      "step": 5193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002699450623669581,
+      "loss": 1.4364,
+      "step": 5194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002699336631421039,
+      "loss": 0.5962,
+      "step": 5195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002699222619966921,
+      "loss": 0.6432,
+      "step": 5196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002699108589309053,
+      "loss": 0.2823,
+      "step": 5197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000269899453944926,
+      "loss": 1.2481,
+      "step": 5198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002698880470389369,
+      "loss": 0.4015,
+      "step": 5199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026987663821312065,
+      "loss": 0.3033,
+      "step": 5200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026986522746765993,
+      "loss": 4.9241,
+      "step": 5201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002698538148027375,
+      "loss": 0.5229,
+      "step": 5202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002698424002185361,
+      "loss": 0.4157,
+      "step": 5203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026983098371523854,
+      "loss": 0.3099,
+      "step": 5204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026981956529302764,
+      "loss": 0.008,
+      "step": 5205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026980814495208616,
+      "loss": 0.0047,
+      "step": 5206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026979672269259705,
+      "loss": 0.1613,
+      "step": 5207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002697852985147433,
+      "loss": 0.3621,
+      "step": 5208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026977387241870765,
+      "loss": 1.6095,
+      "step": 5209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026976244440467327,
+      "loss": 0.4214,
+      "step": 5210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026975101447282305,
+      "loss": 0.3521,
+      "step": 5211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026973958262334,
+      "loss": 0.4248,
+      "step": 5212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026972814885640735,
+      "loss": 0.0126,
+      "step": 5213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026971671317220806,
+      "loss": 1.2097,
+      "step": 5214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002697052755709252,
+      "loss": 0.6315,
+      "step": 5215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026969383605274206,
+      "loss": 1.2428,
+      "step": 5216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026968239461784175,
+      "loss": 1.4551,
+      "step": 5217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696709512664075,
+      "loss": 0.3401,
+      "step": 5218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696595059986226,
+      "loss": 0.4584,
+      "step": 5219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696480588146703,
+      "loss": 0.3943,
+      "step": 5220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026963660971473385,
+      "loss": 0.1584,
+      "step": 5221
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696251586989967,
+      "loss": 0.7672,
+      "step": 5222
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696137057676421,
+      "loss": 0.3504,
+      "step": 5223
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002696022509208536,
+      "loss": 0.1306,
+      "step": 5224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026959079415881456,
+      "loss": 0.0056,
+      "step": 5225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002695793354817084,
+      "loss": 0.1923,
+      "step": 5226
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026956787488971863,
+      "loss": 0.0771,
+      "step": 5227
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026955641238302877,
+      "loss": 0.7919,
+      "step": 5228
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002695449479618224,
+      "loss": 0.6207,
+      "step": 5229
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026953348162628315,
+      "loss": 0.3058,
+      "step": 5230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002695220133765945,
+      "loss": 0.1591,
+      "step": 5231
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002695105432129403,
+      "loss": 0.8658,
+      "step": 5232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026949907113550404,
+      "loss": 0.8408,
+      "step": 5233
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026948759714446945,
+      "loss": 0.3608,
+      "step": 5234
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002694761212400204,
+      "loss": 0.4757,
+      "step": 5235
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002694646434223406,
+      "loss": 0.5935,
+      "step": 5236
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002694531636916137,
+      "loss": 0.004,
+      "step": 5237
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026944168204802373,
+      "loss": 0.1961,
+      "step": 5238
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026943019849175453,
+      "loss": 0.3013,
+      "step": 5239
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026941871302298994,
+      "loss": 1.2863,
+      "step": 5240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002694072256419138,
+      "loss": 0.4762,
+      "step": 5241
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002693957363487102,
+      "loss": 0.8075,
+      "step": 5242
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000269384245143563,
+      "loss": 1.0787,
+      "step": 5243
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026937275202665637,
+      "loss": 0.2623,
+      "step": 5244
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026936125699817423,
+      "loss": 0.9414,
+      "step": 5245
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002693497600583007,
+      "loss": 0.6723,
+      "step": 5246
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002693382612072199,
+      "loss": 0.0034,
+      "step": 5247
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026932676044511596,
+      "loss": 0.3038,
+      "step": 5248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000269315257772173,
+      "loss": 0.4563,
+      "step": 5249
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026930375318857524,
+      "loss": 0.7673,
+      "step": 5250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692922466945069,
+      "loss": 0.4377,
+      "step": 5251
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026928073829015237,
+      "loss": 0.6639,
+      "step": 5252
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026926922797569576,
+      "loss": 0.2943,
+      "step": 5253
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692577157513215,
+      "loss": 0.6372,
+      "step": 5254
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026924620161721387,
+      "loss": 0.513,
+      "step": 5255
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026923468557355727,
+      "loss": 0.2722,
+      "step": 5256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692231676205361,
+      "loss": 0.6129,
+      "step": 5257
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692116477583349,
+      "loss": 0.9037,
+      "step": 5258
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002692001259871381,
+      "loss": 0.689,
+      "step": 5259
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026918860230713007,
+      "loss": 2.5001,
+      "step": 5260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002691770767184955,
+      "loss": 0.6996,
+      "step": 5261
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026916554922141897,
+      "loss": 0.5809,
+      "step": 5262
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026915401981608494,
+      "loss": 0.1609,
+      "step": 5263
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002691424885026781,
+      "loss": 0.618,
+      "step": 5264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026913095528138315,
+      "loss": 0.4191,
+      "step": 5265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026911942015238474,
+      "loss": 1.2627,
+      "step": 5266
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002691078831158676,
+      "loss": 1.1788,
+      "step": 5267
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026909634417201643,
+      "loss": 3.6446,
+      "step": 5268
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026908480332101604,
+      "loss": 0.2226,
+      "step": 5269
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026907326056305127,
+      "loss": 0.4881,
+      "step": 5270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026906171589830696,
+      "loss": 0.5733,
+      "step": 5271
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026905016932696793,
+      "loss": 0.2333,
+      "step": 5272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002690386208492191,
+      "loss": 0.63,
+      "step": 5273
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002690270704652454,
+      "loss": 0.3544,
+      "step": 5274
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026901551817523183,
+      "loss": 0.9786,
+      "step": 5275
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002690039639793633,
+      "loss": 0.4536,
+      "step": 5276
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026899240787782496,
+      "loss": 0.2141,
+      "step": 5277
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026898084987080176,
+      "loss": 0.797,
+      "step": 5278
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689692899584788,
+      "loss": 0.5984,
+      "step": 5279
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026895772814104123,
+      "loss": 0.2884,
+      "step": 5280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689461644186742,
+      "loss": 0.2932,
+      "step": 5281
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026893459879156285,
+      "loss": 0.3264,
+      "step": 5282
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002689230312598923,
+      "loss": 0.3919,
+      "step": 5283
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000268911461823848,
+      "loss": 3.2369,
+      "step": 5284
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026889989048361503,
+      "loss": 0.1846,
+      "step": 5285
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002688883172393788,
+      "loss": 0.0762,
+      "step": 5286
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026887674209132457,
+      "loss": 0.5065,
+      "step": 5287
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026886516503963774,
+      "loss": 0.9943,
+      "step": 5288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002688535860845037,
+      "loss": 0.857,
+      "step": 5289
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002688420052261078,
+      "loss": 0.0364,
+      "step": 5290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002688304224646356,
+      "loss": 0.4896,
+      "step": 5291
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002688188378002725,
+      "loss": 0.33,
+      "step": 5292
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026880725123320403,
+      "loss": 1.3077,
+      "step": 5293
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002687956627636157,
+      "loss": 1.1622,
+      "step": 5294
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026878407239169317,
+      "loss": 0.2962,
+      "step": 5295
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000268772480117622,
+      "loss": 0.3257,
+      "step": 5296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002687608859415878,
+      "loss": 0.3188,
+      "step": 5297
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026874928986377625,
+      "loss": 0.5823,
+      "step": 5298
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000268737691884373,
+      "loss": 1.1319,
+      "step": 5299
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026872609200356384,
+      "loss": 0.3356,
+      "step": 5300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002687144902215345,
+      "loss": 0.0028,
+      "step": 5301
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002687028865384708,
+      "loss": 0.2925,
+      "step": 5302
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026869128095455844,
+      "loss": 0.9093,
+      "step": 5303
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002686796734699834,
+      "loss": 0.3799,
+      "step": 5304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002686680640849315,
+      "loss": 0.1776,
+      "step": 5305
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026865645279958866,
+      "loss": 0.3699,
+      "step": 5306
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002686448396141408,
+      "loss": 0.4009,
+      "step": 5307
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002686332245287739,
+      "loss": 0.0179,
+      "step": 5308
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026862160754367393,
+      "loss": 0.2594,
+      "step": 5309
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026860998865902694,
+      "loss": 0.576,
+      "step": 5310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000268598367875019,
+      "loss": 0.294,
+      "step": 5311
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685867451918362,
+      "loss": 0.6305,
+      "step": 5312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685751206096646,
+      "loss": 2.3575,
+      "step": 5313
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685634941286905,
+      "loss": 0.4648,
+      "step": 5314
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685518657490999,
+      "loss": 1.3284,
+      "step": 5315
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685402354710791,
+      "loss": 0.8474,
+      "step": 5316
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026852860329481436,
+      "loss": 2.5704,
+      "step": 5317
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002685169692204919,
+      "loss": 0.4805,
+      "step": 5318
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026850533324829803,
+      "loss": 0.5606,
+      "step": 5319
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684936953784192,
+      "loss": 0.6657,
+      "step": 5320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026848205561104155,
+      "loss": 0.0801,
+      "step": 5321
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684704139463516,
+      "loss": 1.7422,
+      "step": 5322
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026845877038453585,
+      "loss": 0.5296,
+      "step": 5323
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026844712492578066,
+      "loss": 0.4269,
+      "step": 5324
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026843547757027247,
+      "loss": 0.1891,
+      "step": 5325
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684238283181979,
+      "loss": 0.3454,
+      "step": 5326
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684121771697435,
+      "loss": 0.0649,
+      "step": 5327
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002684005241250957,
+      "loss": 0.601,
+      "step": 5328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683888691844412,
+      "loss": 0.7187,
+      "step": 5329
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683772123479667,
+      "loss": 0.2996,
+      "step": 5330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683655536158588,
+      "loss": 0.3693,
+      "step": 5331
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026835389298830416,
+      "loss": 0.3369,
+      "step": 5332
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026834223046548955,
+      "loss": 0.2014,
+      "step": 5333
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026833056604760175,
+      "loss": 0.4297,
+      "step": 5334
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683188997348275,
+      "loss": 0.5795,
+      "step": 5335
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002683072315273537,
+      "loss": 0.852,
+      "step": 5336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682955614253671,
+      "loss": 1.1702,
+      "step": 5337
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682838894290546,
+      "loss": 0.3264,
+      "step": 5338
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026827221553860314,
+      "loss": 0.2603,
+      "step": 5339
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026826053975419965,
+      "loss": 0.4701,
+      "step": 5340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026824886207603107,
+      "loss": 0.2345,
+      "step": 5341
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682371825042845,
+      "loss": 0.669,
+      "step": 5342
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682255010391469,
+      "loss": 0.6379,
+      "step": 5343
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002682138176808053,
+      "loss": 0.7395,
+      "step": 5344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026820213242944685,
+      "loss": 0.8577,
+      "step": 5345
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002681904452852586,
+      "loss": 0.0133,
+      "step": 5346
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026817875624842775,
+      "loss": 0.8394,
+      "step": 5347
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002681670653191415,
+      "loss": 2.4528,
+      "step": 5348
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026815537249758704,
+      "loss": 2.1485,
+      "step": 5349
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026814367778395167,
+      "loss": 0.019,
+      "step": 5350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026813198117842254,
+      "loss": 1.5479,
+      "step": 5351
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000268120282681187,
+      "loss": 0.2561,
+      "step": 5352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002681085822924325,
+      "loss": 0.84,
+      "step": 5353
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00026809688001234625,
+      "loss": 0.4201,
+      "step": 5354
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002680851758411157,
+      "loss": 0.6751,
+      "step": 5355
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002680734697789283,
+      "loss": 0.3545,
+      "step": 5356
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002680617618259715,
+      "loss": 0.2078,
+      "step": 5357
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002680500519824327,
+      "loss": 0.8314,
+      "step": 5358
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002680383402484996,
+      "loss": 0.599,
+      "step": 5359
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026802662662435954,
+      "loss": 0.4058,
+      "step": 5360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002680149111102002,
+      "loss": 1.7032,
+      "step": 5361
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026800319370620917,
+      "loss": 0.9379,
+      "step": 5362
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026799147441257414,
+      "loss": 0.5892,
+      "step": 5363
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002679797532294827,
+      "loss": 0.4544,
+      "step": 5364
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026796803015712256,
+      "loss": 0.3688,
+      "step": 5365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026795630519568147,
+      "loss": 0.3301,
+      "step": 5366
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002679445783453472,
+      "loss": 0.4323,
+      "step": 5367
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002679328496063075,
+      "loss": 0.7034,
+      "step": 5368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026792111897875015,
+      "loss": 0.4994,
+      "step": 5369
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026790938646286316,
+      "loss": 0.0943,
+      "step": 5370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026789765205883424,
+      "loss": 0.4635,
+      "step": 5371
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678859157668514,
+      "loss": 0.7838,
+      "step": 5372
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678741775871025,
+      "loss": 0.4908,
+      "step": 5373
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026786243751977557,
+      "loss": 0.4053,
+      "step": 5374
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678506955650586,
+      "loss": 1.1606,
+      "step": 5375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678389517231396,
+      "loss": 0.5101,
+      "step": 5376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678272059942066,
+      "loss": 0.1985,
+      "step": 5377
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678154583784478,
+      "loss": 0.9099,
+      "step": 5378
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002678037088760512,
+      "loss": 0.4567,
+      "step": 5379
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000267791957487205,
+      "loss": 0.4331,
+      "step": 5380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026778020421209744,
+      "loss": 0.5019,
+      "step": 5381
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002677684490509166,
+      "loss": 0.0421,
+      "step": 5382
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002677566920038509,
+      "loss": 0.6336,
+      "step": 5383
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002677449330710884,
+      "loss": 0.3754,
+      "step": 5384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002677331722528175,
+      "loss": 0.7366,
+      "step": 5385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002677214095492266,
+      "loss": 1.5245,
+      "step": 5386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000267709644960504,
+      "loss": 0.2858,
+      "step": 5387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002676978784868381,
+      "loss": 2.0918,
+      "step": 5388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026768611012841723,
+      "loss": 0.5204,
+      "step": 5389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026767433988543,
+      "loss": 0.9762,
+      "step": 5390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002676625677580648,
+      "loss": 0.0425,
+      "step": 5391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026765079374651017,
+      "loss": 0.0273,
+      "step": 5392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026763901785095467,
+      "loss": 0.1676,
+      "step": 5393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002676272400715868,
+      "loss": 0.1877,
+      "step": 5394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002676154604085953,
+      "loss": 0.2754,
+      "step": 5395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002676036788621686,
+      "loss": 0.272,
+      "step": 5396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002675918954324956,
+      "loss": 0.2238,
+      "step": 5397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026758011011976477,
+      "loss": 0.5024,
+      "step": 5398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026756832292416504,
+      "loss": 0.0502,
+      "step": 5399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000267556533845885,
+      "loss": 0.9217,
+      "step": 5400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002675447428851135,
+      "loss": 0.0453,
+      "step": 5401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026753295004203935,
+      "loss": 0.6314,
+      "step": 5402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002675211553168514,
+      "loss": 0.0874,
+      "step": 5403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026750935870973854,
+      "loss": 0.3272,
+      "step": 5404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002674975602208897,
+      "loss": 0.4336,
+      "step": 5405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002674857598504937,
+      "loss": 0.939,
+      "step": 5406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026747395759873957,
+      "loss": 0.219,
+      "step": 5407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002674621534658164,
+      "loss": 1.2029,
+      "step": 5408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000267450347451913,
+      "loss": 0.2999,
+      "step": 5409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026743853955721867,
+      "loss": 0.1183,
+      "step": 5410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002674267297819223,
+      "loss": 0.357,
+      "step": 5411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026741491812621313,
+      "loss": 0.3573,
+      "step": 5412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026740310459028027,
+      "loss": 2.2657,
+      "step": 5413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002673912891743129,
+      "loss": 1.5705,
+      "step": 5414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026737947187850013,
+      "loss": 0.9312,
+      "step": 5415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026736765270303134,
+      "loss": 0.0199,
+      "step": 5416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026735583164809576,
+      "loss": 0.1629,
+      "step": 5417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002673440087138826,
+      "loss": 0.0223,
+      "step": 5418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002673321839005813,
+      "loss": 0.0444,
+      "step": 5419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026732035720838113,
+      "loss": 0.4048,
+      "step": 5420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026730852863747153,
+      "loss": 0.4271,
+      "step": 5421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672966981880419,
+      "loss": 0.3899,
+      "step": 5422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672848658602817,
+      "loss": 1.9162,
+      "step": 5423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026727303165438034,
+      "loss": 0.0237,
+      "step": 5424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672611955705274,
+      "loss": 0.7174,
+      "step": 5425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672493576089125,
+      "loss": 0.4343,
+      "step": 5426
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026723751776972497,
+      "loss": 1.0501,
+      "step": 5427
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002672256760531546,
+      "loss": 0.681,
+      "step": 5428
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026721383245939095,
+      "loss": 0.0058,
+      "step": 5429
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026720198698862365,
+      "loss": 0.2324,
+      "step": 5430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002671901396410425,
+      "loss": 0.2105,
+      "step": 5431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002671782904168371,
+      "loss": 1.1403,
+      "step": 5432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026716643931619724,
+      "loss": 0.1758,
+      "step": 5433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002671545863393127,
+      "loss": 0.0428,
+      "step": 5434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026714273148637326,
+      "loss": 0.7647,
+      "step": 5435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026713087475756876,
+      "loss": 1.5645,
+      "step": 5436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002671190161530892,
+      "loss": 0.0434,
+      "step": 5437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002671071556731243,
+      "loss": 0.1111,
+      "step": 5438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026709529331786405,
+      "loss": 1.7327,
+      "step": 5439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026708342908749843,
+      "loss": 0.7881,
+      "step": 5440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002670715629822174,
+      "loss": 0.011,
+      "step": 5441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026705969500221094,
+      "loss": 0.4664,
+      "step": 5442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026704782514766923,
+      "loss": 0.9047,
+      "step": 5443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026703595341878223,
+      "loss": 1.0668,
+      "step": 5444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026702407981574005,
+      "loss": 0.3375,
+      "step": 5445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002670122043387329,
+      "loss": 0.2352,
+      "step": 5446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002670003269879509,
+      "loss": 0.4915,
+      "step": 5447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026698844776358425,
+      "loss": 0.0075,
+      "step": 5448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002669765666658232,
+      "loss": 0.7483,
+      "step": 5449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026696468369485796,
+      "loss": 0.7692,
+      "step": 5450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002669527988508789,
+      "loss": 0.2391,
+      "step": 5451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026694091213407624,
+      "loss": 0.5048,
+      "step": 5452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026692902354464036,
+      "loss": 0.6993,
+      "step": 5453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026691713308276175,
+      "loss": 0.769,
+      "step": 5454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002669052407486306,
+      "loss": 0.4241,
+      "step": 5455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002668933465424376,
+      "loss": 0.3509,
+      "step": 5456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266881450464373,
+      "loss": 0.0042,
+      "step": 5457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002668695525146274,
+      "loss": 0.6369,
+      "step": 5458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026685765269339137,
+      "loss": 0.2164,
+      "step": 5459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002668457510008553,
+      "loss": 0.0013,
+      "step": 5460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026683384743720994,
+      "loss": 0.5716,
+      "step": 5461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026682194200264585,
+      "loss": 0.1546,
+      "step": 5462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026681003469735374,
+      "loss": 0.4197,
+      "step": 5463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002667981255215242,
+      "loss": 0.523,
+      "step": 5464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026678621447534794,
+      "loss": 2.8133,
+      "step": 5465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002667743015590157,
+      "loss": 0.317,
+      "step": 5466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002667623867727184,
+      "loss": 0.1106,
+      "step": 5467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026675047011664657,
+      "loss": 0.5836,
+      "step": 5468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026673855159099123,
+      "loss": 0.189,
+      "step": 5469
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002667266311959432,
+      "loss": 0.3522,
+      "step": 5470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026671470893169333,
+      "loss": 0.4839,
+      "step": 5471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026670278479843263,
+      "loss": 0.3838,
+      "step": 5472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026669085879635194,
+      "loss": 0.3783,
+      "step": 5473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002666789309256422,
+      "loss": 0.6567,
+      "step": 5474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002666670011864946,
+      "loss": 0.4799,
+      "step": 5475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026665506957910005,
+      "loss": 0.2919,
+      "step": 5476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026664313610364957,
+      "loss": 3.598,
+      "step": 5477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026663120076033433,
+      "loss": 1.4423,
+      "step": 5478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026661926354934554,
+      "loss": 1.3551,
+      "step": 5479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002666073244708742,
+      "loss": 0.0096,
+      "step": 5480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002665953835251115,
+      "loss": 0.4268,
+      "step": 5481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026658344071224885,
+      "loss": 1.144,
+      "step": 5482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002665714960324773,
+      "loss": 0.919,
+      "step": 5483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026655954948598816,
+      "loss": 0.2671,
+      "step": 5484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026654760107297276,
+      "loss": 0.7987,
+      "step": 5485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002665356507936225,
+      "loss": 2.961,
+      "step": 5486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026652369864812867,
+      "loss": 0.0098,
+      "step": 5487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002665117446366827,
+      "loss": 0.0057,
+      "step": 5488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266499788759476,
+      "loss": 0.3567,
+      "step": 5489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002664878310167,
+      "loss": 0.6006,
+      "step": 5490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026647587140854614,
+      "loss": 1.2958,
+      "step": 5491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002664639099352061,
+      "loss": 0.185,
+      "step": 5492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002664519465968713,
+      "loss": 0.06,
+      "step": 5493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026643998139373336,
+      "loss": 0.3065,
+      "step": 5494
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002664280143259839,
+      "loss": 0.3581,
+      "step": 5495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026641604539381447,
+      "loss": 0.2479,
+      "step": 5496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002664040745974168,
+      "loss": 0.8986,
+      "step": 5497
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026639210193698257,
+      "loss": 0.0586,
+      "step": 5498
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663801274127035,
+      "loss": 0.2407,
+      "step": 5499
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663681510247714,
+      "loss": 1.9321,
+      "step": 5500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266356172773378,
+      "loss": 0.3389,
+      "step": 5501
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663441926587151,
+      "loss": 2.3635,
+      "step": 5502
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026633221068097455,
+      "loss": 0.036,
+      "step": 5503
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663202268403482,
+      "loss": 0.4173,
+      "step": 5504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002663082411370281,
+      "loss": 0.9021,
+      "step": 5505
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266296253571206,
+      "loss": 0.501,
+      "step": 5506
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002662842641430739,
+      "loss": 0.3068,
+      "step": 5507
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002662722728528239,
+      "loss": 2.2949,
+      "step": 5508
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002662602797006479,
+      "loss": 0.6985,
+      "step": 5509
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266248284686738,
+      "loss": 0.8566,
+      "step": 5510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026623628781128635,
+      "loss": 2.9541,
+      "step": 5511
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026622428907448496,
+      "loss": 2.0647,
+      "step": 5512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026621228847652597,
+      "loss": 1.2902,
+      "step": 5513
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026620028601760166,
+      "loss": 0.2296,
+      "step": 5514
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661882816979041,
+      "loss": 0.5915,
+      "step": 5515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661762755176256,
+      "loss": 0.5072,
+      "step": 5516
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661642674769584,
+      "loss": 0.0092,
+      "step": 5517
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661522575760948,
+      "loss": 0.769,
+      "step": 5518
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026614024581522716,
+      "loss": 0.7514,
+      "step": 5519
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661282321945477,
+      "loss": 0.3787,
+      "step": 5520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266116216714249,
+      "loss": 0.0256,
+      "step": 5521
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002661041993745232,
+      "loss": 0.5377,
+      "step": 5522
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000266092180175563,
+      "loss": 0.4487,
+      "step": 5523
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002660801591175608,
+      "loss": 0.0069,
+      "step": 5524
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026606813620070897,
+      "loss": 0.8953,
+      "step": 5525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026605611142520015,
+      "loss": 0.7779,
+      "step": 5526
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026604408479122694,
+      "loss": 0.3774,
+      "step": 5527
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026603205629898184,
+      "loss": 0.8944,
+      "step": 5528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026602002594865745,
+      "loss": 0.6213,
+      "step": 5529
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026600799374044656,
+      "loss": 4.6896,
+      "step": 5530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026599595967454165,
+      "loss": 0.6279,
+      "step": 5531
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002659839237511356,
+      "loss": 0.5865,
+      "step": 5532
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002659718859704211,
+      "loss": 0.7039,
+      "step": 5533
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026595984633259083,
+      "loss": 0.1343,
+      "step": 5534
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002659478048378377,
+      "loss": 0.0165,
+      "step": 5535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026593576148635446,
+      "loss": 0.3843,
+      "step": 5536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000265923716278334,
+      "loss": 0.7082,
+      "step": 5537
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002659116692139692,
+      "loss": 0.2677,
+      "step": 5538
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026589962029345294,
+      "loss": 0.3956,
+      "step": 5539
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658875695169783,
+      "loss": 0.4227,
+      "step": 5540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658755168847381,
+      "loss": 0.5371,
+      "step": 5541
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026586346239692535,
+      "loss": 0.7336,
+      "step": 5542
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658514060537332,
+      "loss": 0.0063,
+      "step": 5543
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658393478553547,
+      "loss": 0.007,
+      "step": 5544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026582728780198277,
+      "loss": 0.6149,
+      "step": 5545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026581522589381074,
+      "loss": 0.3336,
+      "step": 5546
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002658031621310317,
+      "loss": 0.4898,
+      "step": 5547
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002657910965138387,
+      "loss": 0.0127,
+      "step": 5548
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002657790290424252,
+      "loss": 0.7222,
+      "step": 5549
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026576695971698423,
+      "loss": 0.0239,
+      "step": 5550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026575488853770917,
+      "loss": 0.6981,
+      "step": 5551
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026574281550479327,
+      "loss": 0.3653,
+      "step": 5552
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002657307406184299,
+      "loss": 0.3818,
+      "step": 5553
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026571866387881246,
+      "loss": 0.3818,
+      "step": 5554
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002657065852861342,
+      "loss": 0.1669,
+      "step": 5555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026569450484058873,
+      "loss": 0.5682,
+      "step": 5556
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002656824225423693,
+      "loss": 0.6906,
+      "step": 5557
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002656703383916695,
+      "loss": 0.2699,
+      "step": 5558
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002656582523886829,
+      "loss": 0.6384,
+      "step": 5559
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026564616453360286,
+      "loss": 0.503,
+      "step": 5560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026563407482662314,
+      "loss": 0.7014,
+      "step": 5561
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026562198326793724,
+      "loss": 1.0608,
+      "step": 5562
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002656098898577388,
+      "loss": 0.013,
+      "step": 5563
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026559779459622146,
+      "loss": 0.2676,
+      "step": 5564
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026558569748357896,
+      "loss": 2.3829,
+      "step": 5565
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026557359852000495,
+      "loss": 0.1762,
+      "step": 5566
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002655614977056932,
+      "loss": 0.6785,
+      "step": 5567
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026554939504083757,
+      "loss": 1.6163,
+      "step": 5568
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002655372905256317,
+      "loss": 0.1739,
+      "step": 5569
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026552518416026957,
+      "loss": 0.2668,
+      "step": 5570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000265513075944945,
+      "loss": 0.1858,
+      "step": 5571
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002655009658798519,
+      "loss": 0.5063,
+      "step": 5572
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026548885396518413,
+      "loss": 0.3505,
+      "step": 5573
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002654767402011357,
+      "loss": 0.7825,
+      "step": 5574
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002654646245879006,
+      "loss": 0.2978,
+      "step": 5575
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002654525071256728,
+      "loss": 0.3954,
+      "step": 5576
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026544038781464637,
+      "loss": 0.7083,
+      "step": 5577
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026542826665501535,
+      "loss": 0.3952,
+      "step": 5578
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026541614364697396,
+      "loss": 0.0929,
+      "step": 5579
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002654040187907162,
+      "loss": 1.9708,
+      "step": 5580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026539189208643626,
+      "loss": 0.3527,
+      "step": 5581
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026537976353432837,
+      "loss": 0.2514,
+      "step": 5582
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653676331345867,
+      "loss": 0.3298,
+      "step": 5583
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026535550088740555,
+      "loss": 0.7406,
+      "step": 5584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026534336679297915,
+      "loss": 1.0579,
+      "step": 5585
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026533123085150187,
+      "loss": 1.1444,
+      "step": 5586
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026531909306316804,
+      "loss": 0.0937,
+      "step": 5587
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002653069534281719,
+      "loss": 0.3469,
+      "step": 5588
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002652948119467081,
+      "loss": 0.8589,
+      "step": 5589
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026528266861897083,
+      "loss": 0.6113,
+      "step": 5590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026527052344515464,
+      "loss": 0.8873,
+      "step": 5591
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000265258376425454,
+      "loss": 0.762,
+      "step": 5592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002652462275600635,
+      "loss": 0.9296,
+      "step": 5593
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026523407684917756,
+      "loss": 0.8063,
+      "step": 5594
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002652219242929909,
+      "loss": 0.4092,
+      "step": 5595
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026520976989169795,
+      "loss": 0.6565,
+      "step": 5596
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026519761364549347,
+      "loss": 0.3957,
+      "step": 5597
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002651854555545721,
+      "loss": 0.2035,
+      "step": 5598
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002651732956191286,
+      "loss": 1.1846,
+      "step": 5599
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026516113383935754,
+      "loss": 0.4088,
+      "step": 5600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026514897021545376,
+      "loss": 0.3095,
+      "step": 5601
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002651368047476121,
+      "loss": 0.2869,
+      "step": 5602
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026512463743602724,
+      "loss": 0.4348,
+      "step": 5603
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002651124682808941,
+      "loss": 0.393,
+      "step": 5604
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026510029728240756,
+      "loss": 0.3377,
+      "step": 5605
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002650881244407625,
+      "loss": 0.3109,
+      "step": 5606
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026507594975615386,
+      "loss": 0.5403,
+      "step": 5607
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00026506377322877653,
+      "loss": 0.42,
+      "step": 5608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026505159485882564,
+      "loss": 4.1254,
+      "step": 5609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002650394146464961,
+      "loss": 0.343,
+      "step": 5610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026502723259198293,
+      "loss": 0.848,
+      "step": 5611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002650150486954813,
+      "loss": 1.4893,
+      "step": 5612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002650028629571863,
+      "loss": 0.7835,
+      "step": 5613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026499067537729304,
+      "loss": 0.2249,
+      "step": 5614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026497848595599666,
+      "loss": 0.0271,
+      "step": 5615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002649662946934924,
+      "loss": 1.3565,
+      "step": 5616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026495410158997544,
+      "loss": 0.2453,
+      "step": 5617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026494190664564114,
+      "loss": 0.3931,
+      "step": 5618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002649297098606847,
+      "loss": 0.3271,
+      "step": 5619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026491751123530136,
+      "loss": 0.1528,
+      "step": 5620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002649053107696866,
+      "loss": 0.2288,
+      "step": 5621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026489310846403575,
+      "loss": 0.079,
+      "step": 5622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026488090431854415,
+      "loss": 0.0719,
+      "step": 5623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026486869833340735,
+      "loss": 0.0767,
+      "step": 5624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002648564905088206,
+      "loss": 0.6749,
+      "step": 5625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002648442808449797,
+      "loss": 0.0773,
+      "step": 5626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026483206934207993,
+      "loss": 0.8949,
+      "step": 5627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026481985600031687,
+      "loss": 0.0062,
+      "step": 5628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002648076408198862,
+      "loss": 1.2764,
+      "step": 5629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002647954238009834,
+      "loss": 0.5771,
+      "step": 5630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026478320494380427,
+      "loss": 0.2582,
+      "step": 5631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026477098424854427,
+      "loss": 0.2218,
+      "step": 5632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002647587617153993,
+      "loss": 0.0142,
+      "step": 5633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002647465373445649,
+      "loss": 0.2855,
+      "step": 5634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000264734311136237,
+      "loss": 0.0066,
+      "step": 5635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002647220830906112,
+      "loss": 0.4785,
+      "step": 5636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002647098532078835,
+      "loss": 1.239,
+      "step": 5637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026469762148824963,
+      "loss": 0.3771,
+      "step": 5638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002646853879319055,
+      "loss": 0.5727,
+      "step": 5639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000264673152539047,
+      "loss": 0.2253,
+      "step": 5640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026466091530987,
+      "loss": 0.007,
+      "step": 5641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002646486762445706,
+      "loss": 1.2354,
+      "step": 5642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026463643534334466,
+      "loss": 0.9444,
+      "step": 5643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002646241926063883,
+      "loss": 0.5984,
+      "step": 5644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026461194803389747,
+      "loss": 1.124,
+      "step": 5645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002645997016260683,
+      "loss": 0.9825,
+      "step": 5646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026458745338309695,
+      "loss": 0.7355,
+      "step": 5647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002645752033051795,
+      "loss": 0.5966,
+      "step": 5648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002645629513925121,
+      "loss": 0.6704,
+      "step": 5649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026455069764529094,
+      "loss": 0.866,
+      "step": 5650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026453844206371233,
+      "loss": 0.5221,
+      "step": 5651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002645261846479725,
+      "loss": 0.2055,
+      "step": 5652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002645139253982676,
+      "loss": 0.0035,
+      "step": 5653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002645016643147941,
+      "loss": 0.4753,
+      "step": 5654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026448940139774834,
+      "loss": 0.0442,
+      "step": 5655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026447713664732653,
+      "loss": 1.3765,
+      "step": 5656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026446487006372523,
+      "loss": 0.0024,
+      "step": 5657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026445260164714083,
+      "loss": 1.5645,
+      "step": 5658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002644403313977698,
+      "loss": 0.7166,
+      "step": 5659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002644280593158086,
+      "loss": 0.4859,
+      "step": 5660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002644157854014538,
+      "loss": 0.9595,
+      "step": 5661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026440350965490185,
+      "loss": 0.6802,
+      "step": 5662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002643912320763494,
+      "loss": 1.1673,
+      "step": 5663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026437895266599303,
+      "loss": 0.3396,
+      "step": 5664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002643666714240295,
+      "loss": 0.3873,
+      "step": 5665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026435438835065525,
+      "loss": 0.5128,
+      "step": 5666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002643421034460671,
+      "loss": 0.1248,
+      "step": 5667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026432981671046184,
+      "loss": 0.3231,
+      "step": 5668
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026431752814403607,
+      "loss": 0.2673,
+      "step": 5669
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002643052377469867,
+      "loss": 0.266,
+      "step": 5670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026429294551951046,
+      "loss": 2.1544,
+      "step": 5671
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002642806514618043,
+      "loss": 0.2961,
+      "step": 5672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026426835557406494,
+      "loss": 0.4332,
+      "step": 5673
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002642560578564894,
+      "loss": 1.0128,
+      "step": 5674
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026424375830927453,
+      "loss": 0.6778,
+      "step": 5675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026423145693261737,
+      "loss": 0.7006,
+      "step": 5676
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026421915372671484,
+      "loss": 0.437,
+      "step": 5677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000264206848691764,
+      "loss": 0.3902,
+      "step": 5678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002641945418279618,
+      "loss": 0.4595,
+      "step": 5679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002641822331355055,
+      "loss": 0.1923,
+      "step": 5680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002641699226145921,
+      "loss": 0.3024,
+      "step": 5681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026415761026541866,
+      "loss": 1.2288,
+      "step": 5682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002641452960881825,
+      "loss": 1.0265,
+      "step": 5683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026413298008308065,
+      "loss": 0.5068,
+      "step": 5684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026412066225031043,
+      "loss": 0.0067,
+      "step": 5685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002641083425900691,
+      "loss": 0.7042,
+      "step": 5686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026409602110255396,
+      "loss": 0.2132,
+      "step": 5687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002640836977879622,
+      "loss": 0.0147,
+      "step": 5688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002640713726464913,
+      "loss": 1.5071,
+      "step": 5689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002640590456783385,
+      "loss": 0.0843,
+      "step": 5690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026404671688370126,
+      "loss": 2.2267,
+      "step": 5691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026403438626277704,
+      "loss": 0.5867,
+      "step": 5692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002640220538157633,
+      "loss": 0.447,
+      "step": 5693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002640097195428574,
+      "loss": 0.7049,
+      "step": 5694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026399738344425706,
+      "loss": 1.465,
+      "step": 5695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026398504552015964,
+      "loss": 0.085,
+      "step": 5696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002639727057707628,
+      "loss": 0.6358,
+      "step": 5697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002639603641962641,
+      "loss": 0.6089,
+      "step": 5698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026394802079686126,
+      "loss": 0.6478,
+      "step": 5699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026393567557275185,
+      "loss": 0.3666,
+      "step": 5700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002639233285241335,
+      "loss": 1.6602,
+      "step": 5701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026391097965120414,
+      "loss": 0.5042,
+      "step": 5702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026389862895416136,
+      "loss": 0.5682,
+      "step": 5703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026388627643320295,
+      "loss": 0.8292,
+      "step": 5704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026387392208852676,
+      "loss": 0.3038,
+      "step": 5705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002638615659203306,
+      "loss": 0.4086,
+      "step": 5706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002638492079288124,
+      "loss": 0.6633,
+      "step": 5707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002638368481141699,
+      "loss": 0.4378,
+      "step": 5708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002638244864766012,
+      "loss": 0.0946,
+      "step": 5709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026381212301630416,
+      "loss": 0.4374,
+      "step": 5710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026379975773347677,
+      "loss": 0.2052,
+      "step": 5711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026378739062831705,
+      "loss": 0.896,
+      "step": 5712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026377502170102303,
+      "loss": 1.4325,
+      "step": 5713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002637626509517928,
+      "loss": 0.5507,
+      "step": 5714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026375027838082446,
+      "loss": 0.4014,
+      "step": 5715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002637379039883161,
+      "loss": 0.5438,
+      "step": 5716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002637255277744659,
+      "loss": 1.0293,
+      "step": 5717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026371314973947207,
+      "loss": 0.6124,
+      "step": 5718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026370076988353286,
+      "loss": 0.7583,
+      "step": 5719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002636883882068464,
+      "loss": 0.1772,
+      "step": 5720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026367600470961097,
+      "loss": 0.8267,
+      "step": 5721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000263663619392025,
+      "loss": 1.0812,
+      "step": 5722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026365123225428677,
+      "loss": 0.3607,
+      "step": 5723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026363884329659457,
+      "loss": 0.7836,
+      "step": 5724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026362645251914687,
+      "loss": 0.5798,
+      "step": 5725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002636140599221421,
+      "loss": 0.2857,
+      "step": 5726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002636016655057786,
+      "loss": 0.8928,
+      "step": 5727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026358926927025496,
+      "loss": 0.3722,
+      "step": 5728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002635768712157696,
+      "loss": 0.2363,
+      "step": 5729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026356447134252115,
+      "loss": 0.5108,
+      "step": 5730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002635520696507081,
+      "loss": 0.6161,
+      "step": 5731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002635396661405291,
+      "loss": 0.2065,
+      "step": 5732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002635272608121828,
+      "loss": 0.3296,
+      "step": 5733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026351485366586766,
+      "loss": 0.3828,
+      "step": 5734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002635024447017826,
+      "loss": 0.4292,
+      "step": 5735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634900339201262,
+      "loss": 0.69,
+      "step": 5736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026347762132109724,
+      "loss": 0.5092,
+      "step": 5737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634652069048945,
+      "loss": 1.084,
+      "step": 5738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634527906717167,
+      "loss": 0.4775,
+      "step": 5739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634403726217628,
+      "loss": 1.2544,
+      "step": 5740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634279527552315,
+      "loss": 0.4922,
+      "step": 5741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026341553107232183,
+      "loss": 0.9375,
+      "step": 5742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002634031075732326,
+      "loss": 0.8388,
+      "step": 5743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026339068225816284,
+      "loss": 0.4014,
+      "step": 5744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026337825512731146,
+      "loss": 0.4025,
+      "step": 5745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002633658261808775,
+      "loss": 0.8751,
+      "step": 5746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026335339541905993,
+      "loss": 0.6274,
+      "step": 5747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026334096284205786,
+      "loss": 0.6754,
+      "step": 5748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002633285284500704,
+      "loss": 0.6675,
+      "step": 5749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026331609224329664,
+      "loss": 0.3241,
+      "step": 5750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002633036542219357,
+      "loss": 0.0115,
+      "step": 5751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002632912143861868,
+      "loss": 0.6226,
+      "step": 5752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026327877273624916,
+      "loss": 0.3789,
+      "step": 5753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026326632927232193,
+      "loss": 0.1635,
+      "step": 5754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026325388399460443,
+      "loss": 0.6678,
+      "step": 5755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002632414369032959,
+      "loss": 0.4334,
+      "step": 5756
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002632289879985958,
+      "loss": 0.048,
+      "step": 5757
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026321653728070336,
+      "loss": 2.1211,
+      "step": 5758
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000263204084749818,
+      "loss": 0.3057,
+      "step": 5759
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002631916304061391,
+      "loss": 0.5707,
+      "step": 5760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002631791742498661,
+      "loss": 0.9201,
+      "step": 5761
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026316671628119853,
+      "loss": 0.0102,
+      "step": 5762
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002631542565003358,
+      "loss": 0.4461,
+      "step": 5763
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002631417949074775,
+      "loss": 0.2523,
+      "step": 5764
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026312933150282315,
+      "loss": 0.2958,
+      "step": 5765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026311686628657235,
+      "loss": 0.2544,
+      "step": 5766
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026310439925892465,
+      "loss": 0.3043,
+      "step": 5767
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026309193042007976,
+      "loss": 0.3443,
+      "step": 5768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002630794597702374,
+      "loss": 0.0735,
+      "step": 5769
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002630669873095971,
+      "loss": 0.0688,
+      "step": 5770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026305451303835874,
+      "loss": 0.4408,
+      "step": 5771
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026304203695672203,
+      "loss": 0.0702,
+      "step": 5772
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026302955906488674,
+      "loss": 0.5171,
+      "step": 5773
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002630170793630527,
+      "loss": 0.2857,
+      "step": 5774
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026300459785141974,
+      "loss": 0.3306,
+      "step": 5775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002629921145301877,
+      "loss": 0.1914,
+      "step": 5776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002629796293995566,
+      "loss": 0.2673,
+      "step": 5777
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026296714245972627,
+      "loss": 0.1562,
+      "step": 5778
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002629546537108967,
+      "loss": 1.4614,
+      "step": 5779
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026294216315326786,
+      "loss": 0.4539,
+      "step": 5780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026292967078703975,
+      "loss": 0.8104,
+      "step": 5781
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026291717661241246,
+      "loss": 0.4194,
+      "step": 5782
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026290468062958606,
+      "loss": 0.5295,
+      "step": 5783
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026289218283876065,
+      "loss": 0.8313,
+      "step": 5784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002628796832401364,
+      "loss": 0.237,
+      "step": 5785
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002628671818339134,
+      "loss": 0.0618,
+      "step": 5786
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002628546786202919,
+      "loss": 0.0021,
+      "step": 5787
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026284217359947204,
+      "loss": 0.4667,
+      "step": 5788
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026282966677165407,
+      "loss": 1.2786,
+      "step": 5789
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026281715813703845,
+      "loss": 0.584,
+      "step": 5790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026280464769582525,
+      "loss": 0.3922,
+      "step": 5791
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000262792135448215,
+      "loss": 1.2522,
+      "step": 5792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002627796213944079,
+      "loss": 2.0645,
+      "step": 5793
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002627671055346045,
+      "loss": 0.3853,
+      "step": 5794
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002627545878690051,
+      "loss": 0.7081,
+      "step": 5795
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002627420683978102,
+      "loss": 0.6748,
+      "step": 5796
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026272954712122025,
+      "loss": 0.007,
+      "step": 5797
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026271702403943585,
+      "loss": 0.06,
+      "step": 5798
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002627044991526574,
+      "loss": 0.1737,
+      "step": 5799
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002626919724610856,
+      "loss": 0.5375,
+      "step": 5800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026267944396492096,
+      "loss": 0.2663,
+      "step": 5801
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026266691366436416,
+      "loss": 0.3092,
+      "step": 5802
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002626543815596158,
+      "loss": 0.4834,
+      "step": 5803
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002626418476508766,
+      "loss": 0.4743,
+      "step": 5804
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026262931193834723,
+      "loss": 0.7136,
+      "step": 5805
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026261677442222857,
+      "loss": 1.1129,
+      "step": 5806
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026260423510272114,
+      "loss": 0.5759,
+      "step": 5807
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026259169398002596,
+      "loss": 0.0316,
+      "step": 5808
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026257915105434376,
+      "loss": 0.3307,
+      "step": 5809
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002625666063258754,
+      "loss": 0.0298,
+      "step": 5810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026255405979482184,
+      "loss": 0.0685,
+      "step": 5811
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026254151146138386,
+      "loss": 1.9578,
+      "step": 5812
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002625289613257625,
+      "loss": 0.3447,
+      "step": 5813
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002625164093881587,
+      "loss": 1.0424,
+      "step": 5814
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002625038556487735,
+      "loss": 0.3069,
+      "step": 5815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026249130010780786,
+      "loss": 0.5055,
+      "step": 5816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002624787427654629,
+      "loss": 0.007,
+      "step": 5817
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026246618362193964,
+      "loss": 0.0128,
+      "step": 5818
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002624536226774393,
+      "loss": 0.4852,
+      "step": 5819
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002624410599321629,
+      "loss": 4.7412,
+      "step": 5820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026242849538631166,
+      "loss": 0.014,
+      "step": 5821
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002624159290400869,
+      "loss": 0.0067,
+      "step": 5822
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002624033608936897,
+      "loss": 0.0634,
+      "step": 5823
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026239079094732135,
+      "loss": 0.5168,
+      "step": 5824
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026237821920118315,
+      "loss": 0.3707,
+      "step": 5825
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026236564565547643,
+      "loss": 0.3484,
+      "step": 5826
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002623530703104026,
+      "loss": 0.0093,
+      "step": 5827
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026234049316616287,
+      "loss": 1.4932,
+      "step": 5828
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026232791422295875,
+      "loss": 1.1062,
+      "step": 5829
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026231533348099173,
+      "loss": 0.0039,
+      "step": 5830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026230275094046314,
+      "loss": 0.0256,
+      "step": 5831
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002622901666015746,
+      "loss": 0.5801,
+      "step": 5832
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026227758046452757,
+      "loss": 0.3892,
+      "step": 5833
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026226499252952353,
+      "loss": 0.2341,
+      "step": 5834
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002622524027967642,
+      "loss": 0.5934,
+      "step": 5835
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026223981126645104,
+      "loss": 0.3539,
+      "step": 5836
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026222721793878575,
+      "loss": 0.1678,
+      "step": 5837
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026221462281397,
+      "loss": 0.843,
+      "step": 5838
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002622020258922055,
+      "loss": 0.2319,
+      "step": 5839
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002621894271736939,
+      "loss": 0.4036,
+      "step": 5840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002621768266586371,
+      "loss": 0.4581,
+      "step": 5841
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026216422434723665,
+      "loss": 1.2901,
+      "step": 5842
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002621516202396946,
+      "loss": 0.1361,
+      "step": 5843
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026213901433621255,
+      "loss": 0.1104,
+      "step": 5844
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026212640663699253,
+      "loss": 0.4992,
+      "step": 5845
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002621137971422364,
+      "loss": 1.0512,
+      "step": 5846
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026210118585214605,
+      "loss": 0.2401,
+      "step": 5847
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002620885727669235,
+      "loss": 0.0182,
+      "step": 5848
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002620759578867706,
+      "loss": 0.5635,
+      "step": 5849
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002620633412118895,
+      "loss": 0.2643,
+      "step": 5850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002620507227424822,
+      "loss": 0.1339,
+      "step": 5851
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002620381024787506,
+      "loss": 0.3118,
+      "step": 5852
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026202548042089706,
+      "loss": 0.9502,
+      "step": 5853
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002620128565691235,
+      "loss": 0.2101,
+      "step": 5854
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00026200023092363224,
+      "loss": 0.3894,
+      "step": 5855
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002619876034846254,
+      "loss": 1.1029,
+      "step": 5856
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000261974974252305,
+      "loss": 0.0313,
+      "step": 5857
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026196234322687354,
+      "loss": 0.9858,
+      "step": 5858
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026194971040853314,
+      "loss": 0.3536,
+      "step": 5859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002619370757974862,
+      "loss": 1.8643,
+      "step": 5860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026192443939393494,
+      "loss": 0.4938,
+      "step": 5861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002619118011980818,
+      "loss": 1.0845,
+      "step": 5862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002618991612101291,
+      "loss": 0.1458,
+      "step": 5863
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002618865194302792,
+      "loss": 0.1591,
+      "step": 5864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002618738758587347,
+      "loss": 0.56,
+      "step": 5865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000261861230495698,
+      "loss": 0.2768,
+      "step": 5866
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002618485833413715,
+      "loss": 0.3722,
+      "step": 5867
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002618359343959578,
+      "loss": 0.7039,
+      "step": 5868
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026182328365965946,
+      "loss": 0.3355,
+      "step": 5869
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026181063113267907,
+      "loss": 0.5071,
+      "step": 5870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026179797681521923,
+      "loss": 0.3851,
+      "step": 5871
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002617853207074826,
+      "loss": 1.2149,
+      "step": 5872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026177266280967174,
+      "loss": 0.1414,
+      "step": 5873
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002617600031219895,
+      "loss": 0.5786,
+      "step": 5874
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002617473416446386,
+      "loss": 0.5348,
+      "step": 5875
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002617346783778216,
+      "loss": 0.4309,
+      "step": 5876
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002617220133217415,
+      "loss": 1.0506,
+      "step": 5877
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000261709346476601,
+      "loss": 0.0282,
+      "step": 5878
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026169667784260297,
+      "loss": 0.4379,
+      "step": 5879
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026168400741995027,
+      "loss": 0.3945,
+      "step": 5880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002616713352088458,
+      "loss": 0.0022,
+      "step": 5881
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026165866120949255,
+      "loss": 2.2618,
+      "step": 5882
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026164598542209334,
+      "loss": 0.5163,
+      "step": 5883
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002616333078468513,
+      "loss": 1.6818,
+      "step": 5884
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026162062848396935,
+      "loss": 0.5133,
+      "step": 5885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002616079473336505,
+      "loss": 0.8463,
+      "step": 5886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026159526439609794,
+      "loss": 0.5826,
+      "step": 5887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002615825796715147,
+      "loss": 0.2557,
+      "step": 5888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002615698931601039,
+      "loss": 0.0486,
+      "step": 5889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026155720486206865,
+      "loss": 0.5561,
+      "step": 5890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026154451477761225,
+      "loss": 0.9544,
+      "step": 5891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002615318229069378,
+      "loss": 0.7517,
+      "step": 5892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002615191292502487,
+      "loss": 0.7169,
+      "step": 5893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026150643380774797,
+      "loss": 0.7872,
+      "step": 5894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026149373657963915,
+      "loss": 0.2281,
+      "step": 5895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002614810375661254,
+      "loss": 0.2968,
+      "step": 5896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002614683367674102,
+      "loss": 0.0133,
+      "step": 5897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026145563418369683,
+      "loss": 0.1495,
+      "step": 5898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026144292981518873,
+      "loss": 0.6649,
+      "step": 5899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026143022366208944,
+      "loss": 0.0163,
+      "step": 5900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002614175157246023,
+      "loss": 0.2932,
+      "step": 5901
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026140480600293087,
+      "loss": 0.2181,
+      "step": 5902
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002613920944972787,
+      "loss": 2.8184,
+      "step": 5903
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026137938120784926,
+      "loss": 1.7178,
+      "step": 5904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026136666613484625,
+      "loss": 0.1884,
+      "step": 5905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026135394927847313,
+      "loss": 0.8113,
+      "step": 5906
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026134123063893374,
+      "loss": 0.9957,
+      "step": 5907
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002613285102164315,
+      "loss": 0.4598,
+      "step": 5908
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002613157880111704,
+      "loss": 0.2123,
+      "step": 5909
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026130306402335395,
+      "loss": 0.8738,
+      "step": 5910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026129033825318595,
+      "loss": 0.0209,
+      "step": 5911
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612776107008702,
+      "loss": 0.6367,
+      "step": 5912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612648813666106,
+      "loss": 0.7332,
+      "step": 5913
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612521502506108,
+      "loss": 2.435,
+      "step": 5914
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026123941735307487,
+      "loss": 0.2492,
+      "step": 5915
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612266826742066,
+      "loss": 0.7809,
+      "step": 5916
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612139462142099,
+      "loss": 0.3045,
+      "step": 5917
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002612012079732888,
+      "loss": 0.603,
+      "step": 5918
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002611884679516472,
+      "loss": 0.0312,
+      "step": 5919
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002611757261494892,
+      "loss": 0.0511,
+      "step": 5920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002611629825670188,
+      "loss": 0.6671,
+      "step": 5921
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026115023720444004,
+      "loss": 0.2292,
+      "step": 5922
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026113749006195704,
+      "loss": 0.2476,
+      "step": 5923
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002611247411397739,
+      "loss": 0.4548,
+      "step": 5924
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002611119904380949,
+      "loss": 0.6519,
+      "step": 5925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026109923795712405,
+      "loss": 0.3302,
+      "step": 5926
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026108648369706564,
+      "loss": 0.3467,
+      "step": 5927
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002610737276581239,
+      "loss": 0.6338,
+      "step": 5928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002610609698405032,
+      "loss": 0.4657,
+      "step": 5929
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026104821024440766,
+      "loss": 0.3594,
+      "step": 5930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026103544887004173,
+      "loss": 0.0122,
+      "step": 5931
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002610226857176097,
+      "loss": 0.6286,
+      "step": 5932
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026100992078731606,
+      "loss": 0.3576,
+      "step": 5933
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026099715407936505,
+      "loss": 0.298,
+      "step": 5934
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026098438559396124,
+      "loss": 0.0552,
+      "step": 5935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026097161533130905,
+      "loss": 0.4508,
+      "step": 5936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000260958843291613,
+      "loss": 0.196,
+      "step": 5937
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002609460694750776,
+      "loss": 0.3478,
+      "step": 5938
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002609332938819074,
+      "loss": 0.5441,
+      "step": 5939
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000260920516512307,
+      "loss": 0.2328,
+      "step": 5940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002609077373664809,
+      "loss": 0.7882,
+      "step": 5941
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026089495644463395,
+      "loss": 0.3749,
+      "step": 5942
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026088217374697063,
+      "loss": 1.2696,
+      "step": 5943
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026086938927369574,
+      "loss": 0.0173,
+      "step": 5944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000260856603025014,
+      "loss": 0.6553,
+      "step": 5945
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002608438150011301,
+      "loss": 0.4971,
+      "step": 5946
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026083102520224887,
+      "loss": 0.3366,
+      "step": 5947
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002608182336285751,
+      "loss": 0.668,
+      "step": 5948
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002608054402803137,
+      "loss": 1.0264,
+      "step": 5949
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026079264515766935,
+      "loss": 0.1513,
+      "step": 5950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026077984826084713,
+      "loss": 0.8847,
+      "step": 5951
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026076704959005186,
+      "loss": 1.0264,
+      "step": 5952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026075424914548856,
+      "loss": 0.8011,
+      "step": 5953
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002607414469273622,
+      "loss": 1.1661,
+      "step": 5954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002607286429358777,
+      "loss": 0.848,
+      "step": 5955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002607158371712402,
+      "loss": 0.4924,
+      "step": 5956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002607030296336548,
+      "loss": 0.526,
+      "step": 5957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002606902203233264,
+      "loss": 0.8598,
+      "step": 5958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002606774092404603,
+      "loss": 0.265,
+      "step": 5959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002606645963852616,
+      "loss": 0.2982,
+      "step": 5960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002606517817579355,
+      "loss": 0.7018,
+      "step": 5961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026063896535868713,
+      "loss": 0.209,
+      "step": 5962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002606261471877218,
+      "loss": 0.2355,
+      "step": 5963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002606133272452447,
+      "loss": 0.4606,
+      "step": 5964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002606005055314612,
+      "loss": 0.0968,
+      "step": 5965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026058768204657666,
+      "loss": 0.4701,
+      "step": 5966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002605748567907963,
+      "loss": 1.2551,
+      "step": 5967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026056202976432554,
+      "loss": 0.4131,
+      "step": 5968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026054920096736986,
+      "loss": 0.4732,
+      "step": 5969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002605363704001346,
+      "loss": 0.2899,
+      "step": 5970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002605235380628252,
+      "loss": 0.3216,
+      "step": 5971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026051070395564734,
+      "loss": 0.3896,
+      "step": 5972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026049786807880637,
+      "loss": 0.5823,
+      "step": 5973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026048503043250783,
+      "loss": 0.721,
+      "step": 5974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002604721910169574,
+      "loss": 0.4804,
+      "step": 5975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026045934983236056,
+      "loss": 0.7918,
+      "step": 5976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000260446506878923,
+      "loss": 0.3646,
+      "step": 5977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002604336621568504,
+      "loss": 0.3436,
+      "step": 5978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002604208156663485,
+      "loss": 0.6238,
+      "step": 5979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002604079674076229,
+      "loss": 1.0117,
+      "step": 5980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026039511738087945,
+      "loss": 0.0833,
+      "step": 5981
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002603822655863239,
+      "loss": 0.3487,
+      "step": 5982
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026036941202416195,
+      "loss": 0.2645,
+      "step": 5983
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026035655669459954,
+      "loss": 1.2022,
+      "step": 5984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002603436995978425,
+      "loss": 0.9463,
+      "step": 5985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002603308407340967,
+      "loss": 0.3864,
+      "step": 5986
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026031798010356806,
+      "loss": 0.5237,
+      "step": 5987
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002603051177064626,
+      "loss": 1.0138,
+      "step": 5988
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026029225354298616,
+      "loss": 0.559,
+      "step": 5989
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002602793876133448,
+      "loss": 0.4502,
+      "step": 5990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026026651991774454,
+      "loss": 0.5618,
+      "step": 5991
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002602536504563915,
+      "loss": 1.0394,
+      "step": 5992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002602407792294917,
+      "loss": 0.9112,
+      "step": 5993
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026022790623725123,
+      "loss": 1.0748,
+      "step": 5994
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002602150314798763,
+      "loss": 0.0055,
+      "step": 5995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002602021549575731,
+      "loss": 1.6202,
+      "step": 5996
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002601892766705477,
+      "loss": 0.58,
+      "step": 5997
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026017639661900645,
+      "loss": 0.1929,
+      "step": 5998
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026016351480315556,
+      "loss": 0.6722,
+      "step": 5999
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002601506312232013,
+      "loss": 0.6363,
+      "step": 6000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026013774587935,
+      "loss": 0.8608,
+      "step": 6001
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000260124858771808,
+      "loss": 0.3725,
+      "step": 6002
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026011196990078164,
+      "loss": 0.543,
+      "step": 6003
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026009907926647735,
+      "loss": 2.2141,
+      "step": 6004
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026008618686910155,
+      "loss": 0.4775,
+      "step": 6005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026007329270886066,
+      "loss": 1.0373,
+      "step": 6006
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002600603967859612,
+      "loss": 0.2995,
+      "step": 6007
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026004749910060965,
+      "loss": 0.5742,
+      "step": 6008
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002600345996530126,
+      "loss": 0.0494,
+      "step": 6009
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002600216984433765,
+      "loss": 1.0243,
+      "step": 6010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00026000879547190807,
+      "loss": 0.5889,
+      "step": 6011
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002599958907388139,
+      "loss": 0.3191,
+      "step": 6012
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002599829842443006,
+      "loss": 0.4867,
+      "step": 6013
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002599700759885748,
+      "loss": 0.497,
+      "step": 6014
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025995716597184337,
+      "loss": 1.0391,
+      "step": 6015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002599442541943129,
+      "loss": 0.0326,
+      "step": 6016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025993134065619024,
+      "loss": 0.6313,
+      "step": 6017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002599184253576821,
+      "loss": 0.3638,
+      "step": 6018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002599055082989954,
+      "loss": 0.2729,
+      "step": 6019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002598925894803369,
+      "loss": 0.5051,
+      "step": 6020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025987966890191354,
+      "loss": 0.6741,
+      "step": 6021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025986674656393213,
+      "loss": 0.3952,
+      "step": 6022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025985382246659965,
+      "loss": 0.8054,
+      "step": 6023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025984089661012307,
+      "loss": 0.3258,
+      "step": 6024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025982796899470947,
+      "loss": 0.5311,
+      "step": 6025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025981503962056566,
+      "loss": 0.7238,
+      "step": 6026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002598021084878989,
+      "loss": 0.1642,
+      "step": 6027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002597891755969161,
+      "loss": 0.555,
+      "step": 6028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002597762409478244,
+      "loss": 0.253,
+      "step": 6029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025976330454083107,
+      "loss": 0.3327,
+      "step": 6030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259750366376143,
+      "loss": 2.1267,
+      "step": 6031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025973742645396764,
+      "loss": 0.2716,
+      "step": 6032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025972448477451206,
+      "loss": 0.066,
+      "step": 6033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002597115413379835,
+      "loss": 0.2109,
+      "step": 6034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025969859614458925,
+      "loss": 0.7317,
+      "step": 6035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002596856491945367,
+      "loss": 0.0349,
+      "step": 6036
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259672700488033,
+      "loss": 1.0347,
+      "step": 6037
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002596597500252856,
+      "loss": 0.7271,
+      "step": 6038
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025964679780650195,
+      "loss": 0.5219,
+      "step": 6039
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025963384383188935,
+      "loss": 0.0166,
+      "step": 6040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025962088810165527,
+      "loss": 1.2323,
+      "step": 6041
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002596079306160072,
+      "loss": 0.0334,
+      "step": 6042
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002595949713751526,
+      "loss": 0.8106,
+      "step": 6043
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025958201037929905,
+      "loss": 0.0422,
+      "step": 6044
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259569047628654,
+      "loss": 0.254,
+      "step": 6045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025955608312342517,
+      "loss": 0.4777,
+      "step": 6046
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025954311686382004,
+      "loss": 2.2679,
+      "step": 6047
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002595301488500463,
+      "loss": 0.7599,
+      "step": 6048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002595171790823116,
+      "loss": 0.0174,
+      "step": 6049
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025950420756082365,
+      "loss": 0.7514,
+      "step": 6050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002594912342857902,
+      "loss": 0.5205,
+      "step": 6051
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002594782592574189,
+      "loss": 0.3242,
+      "step": 6052
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002594652824759176,
+      "loss": 0.25,
+      "step": 6053
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002594523039414941,
+      "loss": 0.2332,
+      "step": 6054
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025943932365435617,
+      "loss": 0.1445,
+      "step": 6055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025942634161471177,
+      "loss": 0.2531,
+      "step": 6056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025941335782276876,
+      "loss": 0.2994,
+      "step": 6057
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259400372278735,
+      "loss": 0.0072,
+      "step": 6058
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002593873849828184,
+      "loss": 0.6095,
+      "step": 6059
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025937439593522703,
+      "loss": 0.2567,
+      "step": 6060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025936140513616886,
+      "loss": 0.7412,
+      "step": 6061
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025934841258585194,
+      "loss": 0.3664,
+      "step": 6062
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025933541828448433,
+      "loss": 0.0954,
+      "step": 6063
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025932242223227403,
+      "loss": 0.0472,
+      "step": 6064
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002593094244294292,
+      "loss": 0.143,
+      "step": 6065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259296424876158,
+      "loss": 0.0092,
+      "step": 6066
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002592834235726686,
+      "loss": 0.496,
+      "step": 6067
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002592704205191692,
+      "loss": 1.0961,
+      "step": 6068
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000259257415715868,
+      "loss": 0.9671,
+      "step": 6069
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025924440916297325,
+      "loss": 1.0786,
+      "step": 6070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025923140086069323,
+      "loss": 0.3191,
+      "step": 6071
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002592183908092363,
+      "loss": 0.0078,
+      "step": 6072
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002592053790088107,
+      "loss": 0.4218,
+      "step": 6073
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002591923654596249,
+      "loss": 0.2495,
+      "step": 6074
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025917935016188723,
+      "loss": 1.491,
+      "step": 6075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002591663331158061,
+      "loss": 0.7908,
+      "step": 6076
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025915331432159,
+      "loss": 1.1125,
+      "step": 6077
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025914029377944737,
+      "loss": 1.7452,
+      "step": 6078
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002591272714895867,
+      "loss": 0.5378,
+      "step": 6079
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002591142474522166,
+      "loss": 0.0052,
+      "step": 6080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002591012216675456,
+      "loss": 1.1652,
+      "step": 6081
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025908819413578225,
+      "loss": 0.3855,
+      "step": 6082
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002590751648571352,
+      "loss": 0.0687,
+      "step": 6083
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025906213383181304,
+      "loss": 0.0331,
+      "step": 6084
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002590491010600245,
+      "loss": 0.3106,
+      "step": 6085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025903606654197834,
+      "loss": 0.9116,
+      "step": 6086
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025902303027788316,
+      "loss": 0.3991,
+      "step": 6087
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025900999226794775,
+      "loss": 0.0255,
+      "step": 6088
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589969525123809,
+      "loss": 0.0038,
+      "step": 6089
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589839110113915,
+      "loss": 0.4511,
+      "step": 6090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589708677651883,
+      "loss": 0.4101,
+      "step": 6091
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589578227739802,
+      "loss": 0.6295,
+      "step": 6092
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025894477603797606,
+      "loss": 0.3421,
+      "step": 6093
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589317275573849,
+      "loss": 0.2559,
+      "step": 6094
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025891867733241554,
+      "loss": 0.2594,
+      "step": 6095
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002589056253632771,
+      "loss": 1.0034,
+      "step": 6096
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025889257165017846,
+      "loss": 0.339,
+      "step": 6097
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002588795161933287,
+      "loss": 0.4406,
+      "step": 6098
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025886645899293693,
+      "loss": 0.4891,
+      "step": 6099
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025885340004921215,
+      "loss": 0.3565,
+      "step": 6100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002588403393623636,
+      "loss": 0.3218,
+      "step": 6101
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002588272769326003,
+      "loss": 0.129,
+      "step": 6102
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002588142127601315,
+      "loss": 2.668,
+      "step": 6103
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00025880114684516647,
+      "loss": 0.3237,
+      "step": 6104
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002587880791879143,
+      "loss": 1.5649,
+      "step": 6105
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002587750097885843,
+      "loss": 0.079,
+      "step": 6106
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025876193864738577,
+      "loss": 0.0947,
+      "step": 6107
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025874886576452804,
+      "loss": 0.2987,
+      "step": 6108
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002587357911402204,
+      "loss": 0.6529,
+      "step": 6109
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025872271477467225,
+      "loss": 0.0043,
+      "step": 6110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000258709636668093,
+      "loss": 0.845,
+      "step": 6111
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025869655682069214,
+      "loss": 0.7469,
+      "step": 6112
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000258683475232679,
+      "loss": 0.4095,
+      "step": 6113
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002586703919042631,
+      "loss": 1.292,
+      "step": 6114
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000258657306835654,
+      "loss": 0.8651,
+      "step": 6115
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002586442200270612,
+      "loss": 0.2005,
+      "step": 6116
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025863113147869425,
+      "loss": 0.3522,
+      "step": 6117
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002586180411907628,
+      "loss": 0.5691,
+      "step": 6118
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025860494916347634,
+      "loss": 0.5802,
+      "step": 6119
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025859185539704466,
+      "loss": 0.4053,
+      "step": 6120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025857875989167744,
+      "loss": 0.9073,
+      "step": 6121
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002585656626475843,
+      "loss": 0.4838,
+      "step": 6122
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000258552563664975,
+      "loss": 0.4305,
+      "step": 6123
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002585394629440593,
+      "loss": 1.4721,
+      "step": 6124
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000258526360485047,
+      "loss": 0.5658,
+      "step": 6125
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002585132562881479,
+      "loss": 0.4341,
+      "step": 6126
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025850015035357187,
+      "loss": 0.0053,
+      "step": 6127
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002584870426815288,
+      "loss": 0.115,
+      "step": 6128
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025847393327222854,
+      "loss": 0.004,
+      "step": 6129
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000258460822125881,
+      "loss": 0.558,
+      "step": 6130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002584477092426962,
+      "loss": 0.2299,
+      "step": 6131
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025843459462288407,
+      "loss": 0.3775,
+      "step": 6132
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025842147826665463,
+      "loss": 0.0441,
+      "step": 6133
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000258408360174218,
+      "loss": 0.9039,
+      "step": 6134
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025839524034578415,
+      "loss": 1.3209,
+      "step": 6135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025838211878156323,
+      "loss": 1.206,
+      "step": 6136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025836899548176525,
+      "loss": 0.3233,
+      "step": 6137
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025835587044660053,
+      "loss": 0.0518,
+      "step": 6138
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025834274367627914,
+      "loss": 0.2436,
+      "step": 6139
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002583296151710113,
+      "loss": 0.7051,
+      "step": 6140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002583164849310073,
+      "loss": 1.3087,
+      "step": 6141
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002583033529564773,
+      "loss": 0.8149,
+      "step": 6142
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002582902192476316,
+      "loss": 0.4194,
+      "step": 6143
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025827708380468065,
+      "loss": 0.0342,
+      "step": 6144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025826394662783466,
+      "loss": 0.4432,
+      "step": 6145
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002582508077173041,
+      "loss": 0.6076,
+      "step": 6146
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025823766707329924,
+      "loss": 0.8785,
+      "step": 6147
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025822452469603064,
+      "loss": 0.8251,
+      "step": 6148
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025821138058570867,
+      "loss": 0.0037,
+      "step": 6149
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025819823474254383,
+      "loss": 0.9015,
+      "step": 6150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002581850871667467,
+      "loss": 0.2285,
+      "step": 6151
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002581719378585277,
+      "loss": 0.6392,
+      "step": 6152
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025815878681809747,
+      "loss": 0.4785,
+      "step": 6153
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025814563404566665,
+      "loss": 0.0973,
+      "step": 6154
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002581324795414457,
+      "loss": 0.0032,
+      "step": 6155
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025811932330564545,
+      "loss": 0.3941,
+      "step": 6156
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002581061653384765,
+      "loss": 0.5211,
+      "step": 6157
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002580930056401495,
+      "loss": 0.9853,
+      "step": 6158
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025807984421087534,
+      "loss": 0.2766,
+      "step": 6159
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002580666810508646,
+      "loss": 0.2893,
+      "step": 6160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002580535161603282,
+      "loss": 0.622,
+      "step": 6161
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002580403495394768,
+      "loss": 0.4279,
+      "step": 6162
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025802718118852146,
+      "loss": 0.7636,
+      "step": 6163
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025801401110767293,
+      "loss": 0.3198,
+      "step": 6164
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002580008392971421,
+      "loss": 0.1575,
+      "step": 6165
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002579876657571399,
+      "loss": 0.4311,
+      "step": 6166
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025797449048787726,
+      "loss": 0.3243,
+      "step": 6167
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025796131348956527,
+      "loss": 0.037,
+      "step": 6168
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025794813476241485,
+      "loss": 0.314,
+      "step": 6169
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002579349543066371,
+      "loss": 0.8732,
+      "step": 6170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025792177212244295,
+      "loss": 1.0067,
+      "step": 6171
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025790858821004366,
+      "loss": 1.2227,
+      "step": 6172
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025789540256965026,
+      "loss": 0.3125,
+      "step": 6173
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002578822152014739,
+      "loss": 0.9597,
+      "step": 6174
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002578690261057258,
+      "loss": 0.555,
+      "step": 6175
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025785583528261714,
+      "loss": 0.2005,
+      "step": 6176
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025784264273235913,
+      "loss": 0.235,
+      "step": 6177
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025782944845516303,
+      "loss": 0.5624,
+      "step": 6178
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025781625245124016,
+      "loss": 0.0936,
+      "step": 6179
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002578030547208018,
+      "loss": 0.3133,
+      "step": 6180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025778985526405936,
+      "loss": 0.8515,
+      "step": 6181
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002577766540812241,
+      "loss": 0.0019,
+      "step": 6182
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002577634511725075,
+      "loss": 0.6898,
+      "step": 6183
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000257750246538121,
+      "loss": 0.4124,
+      "step": 6184
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002577370401782759,
+      "loss": 0.0044,
+      "step": 6185
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025772383209318386,
+      "loss": 0.2049,
+      "step": 6186
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002577106222830563,
+      "loss": 0.9918,
+      "step": 6187
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025769741074810484,
+      "loss": 0.2607,
+      "step": 6188
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002576841974885409,
+      "loss": 0.118,
+      "step": 6189
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002576709825045762,
+      "loss": 0.6309,
+      "step": 6190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025765776579642226,
+      "loss": 0.8028,
+      "step": 6191
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025764454736429084,
+      "loss": 0.007,
+      "step": 6192
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002576313272083935,
+      "loss": 0.6699,
+      "step": 6193
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025761810532894195,
+      "loss": 0.3273,
+      "step": 6194
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000257604881726148,
+      "loss": 0.3734,
+      "step": 6195
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025759165640022334,
+      "loss": 0.274,
+      "step": 6196
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002575784293513798,
+      "loss": 0.3012,
+      "step": 6197
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002575652005798292,
+      "loss": 0.3136,
+      "step": 6198
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002575519700857833,
+      "loss": 0.7568,
+      "step": 6199
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025753873786945404,
+      "loss": 0.9055,
+      "step": 6200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025752550393105323,
+      "loss": 0.1207,
+      "step": 6201
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025751226827079295,
+      "loss": 0.5897,
+      "step": 6202
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025749903088888494,
+      "loss": 0.258,
+      "step": 6203
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025748579178554127,
+      "loss": 0.4646,
+      "step": 6204
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002574725509609741,
+      "loss": 0.4634,
+      "step": 6205
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002574593084153952,
+      "loss": 0.6171,
+      "step": 6206
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025744606414901676,
+      "loss": 1.35,
+      "step": 6207
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025743281816205084,
+      "loss": 0.4246,
+      "step": 6208
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025741957045470967,
+      "loss": 0.5953,
+      "step": 6209
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002574063210272052,
+      "loss": 3.2383,
+      "step": 6210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025739306987974974,
+      "loss": 0.0068,
+      "step": 6211
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573798170125554,
+      "loss": 1.0816,
+      "step": 6212
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573665624258345,
+      "loss": 0.596,
+      "step": 6213
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573533061197992,
+      "loss": 1.0931,
+      "step": 6214
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573400480946618,
+      "loss": 1.0832,
+      "step": 6215
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573267883506347,
+      "loss": 0.2659,
+      "step": 6216
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002573135268879301,
+      "loss": 0.3659,
+      "step": 6217
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025730026370676044,
+      "loss": 0.6608,
+      "step": 6218
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025728699880733807,
+      "loss": 0.3426,
+      "step": 6219
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025727373218987544,
+      "loss": 0.4554,
+      "step": 6220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025726046385458494,
+      "loss": 0.9682,
+      "step": 6221
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002572471938016792,
+      "loss": 0.3003,
+      "step": 6222
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002572339220313705,
+      "loss": 1.0809,
+      "step": 6223
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025722064854387155,
+      "loss": 0.442,
+      "step": 6224
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002572073733393948,
+      "loss": 0.2039,
+      "step": 6225
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025719409641815284,
+      "loss": 0.3258,
+      "step": 6226
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002571808177803583,
+      "loss": 0.3941,
+      "step": 6227
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025716753742622384,
+      "loss": 0.5106,
+      "step": 6228
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002571542553559621,
+      "loss": 0.6509,
+      "step": 6229
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025714097156978577,
+      "loss": 0.0097,
+      "step": 6230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002571276860679076,
+      "loss": 0.3035,
+      "step": 6231
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025711439885054034,
+      "loss": 0.4278,
+      "step": 6232
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025710110991789665,
+      "loss": 0.1662,
+      "step": 6233
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002570878192701895,
+      "loss": 0.1278,
+      "step": 6234
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002570745269076316,
+      "loss": 0.5239,
+      "step": 6235
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025706123283043586,
+      "loss": 0.0274,
+      "step": 6236
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002570479370388151,
+      "loss": 1.2978,
+      "step": 6237
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025703463953298235,
+      "loss": 0.859,
+      "step": 6238
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002570213403131505,
+      "loss": 0.3858,
+      "step": 6239
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002570080393795324,
+      "loss": 0.7022,
+      "step": 6240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002569947367323413,
+      "loss": 0.2556,
+      "step": 6241
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025698143237178994,
+      "loss": 0.7745,
+      "step": 6242
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002569681262980915,
+      "loss": 0.317,
+      "step": 6243
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002569548185114591,
+      "loss": 0.5999,
+      "step": 6244
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025694150901210585,
+      "loss": 1.1994,
+      "step": 6245
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025692819780024473,
+      "loss": 2.546,
+      "step": 6246
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025691488487608903,
+      "loss": 0.4703,
+      "step": 6247
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025690157023985195,
+      "loss": 1.0255,
+      "step": 6248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002568882538917466,
+      "loss": 0.5075,
+      "step": 6249
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002568749358319863,
+      "loss": 0.243,
+      "step": 6250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002568616160607843,
+      "loss": 0.6495,
+      "step": 6251
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025684829457835396,
+      "loss": 0.056,
+      "step": 6252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002568349713849085,
+      "loss": 1.0684,
+      "step": 6253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002568216464806612,
+      "loss": 0.1948,
+      "step": 6254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002568083198658257,
+      "loss": 0.0232,
+      "step": 6255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002567949915406152,
+      "loss": 1.2492,
+      "step": 6256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025678166150524314,
+      "loss": 0.9098,
+      "step": 6257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002567683297599231,
+      "loss": 0.3508,
+      "step": 6258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025675499630486846,
+      "loss": 0.3929,
+      "step": 6259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025674166114029274,
+      "loss": 0.1836,
+      "step": 6260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002567283242664096,
+      "loss": 0.6949,
+      "step": 6261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025671498568343245,
+      "loss": 0.3975,
+      "step": 6262
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025670164539157495,
+      "loss": 0.8349,
+      "step": 6263
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002566883033910508,
+      "loss": 0.0289,
+      "step": 6264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025667495968207353,
+      "loss": 0.2878,
+      "step": 6265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025666161426485696,
+      "loss": 1.1268,
+      "step": 6266
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025664826713961463,
+      "loss": 0.4279,
+      "step": 6267
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002566349183065604,
+      "loss": 0.0084,
+      "step": 6268
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000256621567765908,
+      "loss": 0.3434,
+      "step": 6269
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025660821551787115,
+      "loss": 0.565,
+      "step": 6270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025659486156266373,
+      "loss": 0.1728,
+      "step": 6271
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025658150590049964,
+      "loss": 2.7914,
+      "step": 6272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025656814853159267,
+      "loss": 0.5306,
+      "step": 6273
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002565547894561567,
+      "loss": 1.6872,
+      "step": 6274
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025654142867440576,
+      "loss": 0.3054,
+      "step": 6275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025652806618655364,
+      "loss": 0.9912,
+      "step": 6276
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025651470199281447,
+      "loss": 0.3081,
+      "step": 6277
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002565013360934022,
+      "loss": 0.0074,
+      "step": 6278
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002564879684885309,
+      "loss": 0.4994,
+      "step": 6279
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002564745991784146,
+      "loss": 0.9053,
+      "step": 6280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025646122816326734,
+      "loss": 0.2052,
+      "step": 6281
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002564478554433033,
+      "loss": 1.1348,
+      "step": 6282
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025643448101873666,
+      "loss": 0.2577,
+      "step": 6283
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002564211048897815,
+      "loss": 1.1391,
+      "step": 6284
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000256407727056652,
+      "loss": 0.4442,
+      "step": 6285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025639434751956256,
+      "loss": 0.7583,
+      "step": 6286
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002563809662787273,
+      "loss": 1.038,
+      "step": 6287
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025636758333436044,
+      "loss": 0.8222,
+      "step": 6288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025635419868667646,
+      "loss": 0.0226,
+      "step": 6289
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025634081233588956,
+      "loss": 1.3429,
+      "step": 6290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002563274242822141,
+      "loss": 0.6926,
+      "step": 6291
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025631403452586464,
+      "loss": 0.7001,
+      "step": 6292
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002563006430670554,
+      "loss": 0.0034,
+      "step": 6293
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025628724990600085,
+      "loss": 0.3529,
+      "step": 6294
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025627385504291554,
+      "loss": 0.1341,
+      "step": 6295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000256260458478014,
+      "loss": 0.8677,
+      "step": 6296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002562470602115106,
+      "loss": 0.0234,
+      "step": 6297
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025623366024362006,
+      "loss": 1.0967,
+      "step": 6298
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002562202585745568,
+      "loss": 0.7377,
+      "step": 6299
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025620685520453564,
+      "loss": 0.5886,
+      "step": 6300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025619345013377096,
+      "loss": 0.0111,
+      "step": 6301
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002561800433624776,
+      "loss": 0.3133,
+      "step": 6302
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025616663489087024,
+      "loss": 2.5051,
+      "step": 6303
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002561532247191635,
+      "loss": 1.259,
+      "step": 6304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002561398128475722,
+      "loss": 0.5868,
+      "step": 6305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002561263992763111,
+      "loss": 0.0093,
+      "step": 6306
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025611298400559496,
+      "loss": 0.5184,
+      "step": 6307
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025609956703563863,
+      "loss": 0.4438,
+      "step": 6308
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000256086148366657,
+      "loss": 1.2036,
+      "step": 6309
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025607272799886495,
+      "loss": 0.004,
+      "step": 6310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002560593059324773,
+      "loss": 2.5607,
+      "step": 6311
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000256045882167709,
+      "loss": 0.3034,
+      "step": 6312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025603245670477513,
+      "loss": 0.5475,
+      "step": 6313
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025601902954389057,
+      "loss": 0.5466,
+      "step": 6314
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025600560068527036,
+      "loss": 0.6142,
+      "step": 6315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025599217012912956,
+      "loss": 0.0108,
+      "step": 6316
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025597873787568316,
+      "loss": 0.6418,
+      "step": 6317
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002559653039251464,
+      "loss": 0.9479,
+      "step": 6318
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002559518682777343,
+      "loss": 0.0044,
+      "step": 6319
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025593843093366206,
+      "loss": 0.6389,
+      "step": 6320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025592499189314486,
+      "loss": 0.2462,
+      "step": 6321
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002559115511563978,
+      "loss": 1.2182,
+      "step": 6322
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025589810872363633,
+      "loss": 1.9983,
+      "step": 6323
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002558846645950755,
+      "loss": 0.9038,
+      "step": 6324
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025587121877093064,
+      "loss": 0.6327,
+      "step": 6325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025585777125141713,
+      "loss": 0.476,
+      "step": 6326
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002558443220367503,
+      "loss": 0.4483,
+      "step": 6327
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025583087112714554,
+      "loss": 0.7198,
+      "step": 6328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025581741852281816,
+      "loss": 0.5479,
+      "step": 6329
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002558039642239836,
+      "loss": 0.2878,
+      "step": 6330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002557905082308574,
+      "loss": 1.258,
+      "step": 6331
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000255777050543655,
+      "loss": 0.5118,
+      "step": 6332
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002557635911625918,
+      "loss": 0.1693,
+      "step": 6333
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002557501300878835,
+      "loss": 0.7386,
+      "step": 6334
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025573666731974553,
+      "loss": 2.3454,
+      "step": 6335
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025572320285839357,
+      "loss": 0.932,
+      "step": 6336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002557097367040431,
+      "loss": 0.442,
+      "step": 6337
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002556962688569099,
+      "loss": 0.5312,
+      "step": 6338
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025568279931720956,
+      "loss": 0.1945,
+      "step": 6339
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002556693280851578,
+      "loss": 0.3702,
+      "step": 6340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025565585516097033,
+      "loss": 0.286,
+      "step": 6341
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025564238054486297,
+      "loss": 0.7572,
+      "step": 6342
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002556289042370514,
+      "loss": 0.3497,
+      "step": 6343
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002556154262377514,
+      "loss": 0.9215,
+      "step": 6344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002556019465471789,
+      "loss": 0.3425,
+      "step": 6345
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025558846516554975,
+      "loss": 0.4574,
+      "step": 6346
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025557498209307977,
+      "loss": 0.2532,
+      "step": 6347
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002555614973299849,
+      "loss": 0.7067,
+      "step": 6348
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000255548010876481,
+      "loss": 0.8329,
+      "step": 6349
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025553452273278417,
+      "loss": 0.4474,
+      "step": 6350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002555210328991104,
+      "loss": 0.0043,
+      "step": 6351
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025550754137567555,
+      "loss": 0.4944,
+      "step": 6352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002554940481626958,
+      "loss": 0.6936,
+      "step": 6353
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025548055326038725,
+      "loss": 0.1145,
+      "step": 6354
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002554670566689659,
+      "loss": 0.5763,
+      "step": 6355
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00025545355838864787,
+      "loss": 0.2431,
+      "step": 6356
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002554400584196495,
+      "loss": 3.2052,
+      "step": 6357
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025542655676218676,
+      "loss": 0.1077,
+      "step": 6358
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002554130534164759,
+      "loss": 1.0158,
+      "step": 6359
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002553995483827332,
+      "loss": 2.261,
+      "step": 6360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025538604166117497,
+      "loss": 0.3241,
+      "step": 6361
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025537253325201743,
+      "loss": 0.6693,
+      "step": 6362
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002553590231554769,
+      "loss": 0.5348,
+      "step": 6363
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002553455113717697,
+      "loss": 2.463,
+      "step": 6364
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002553319979011123,
+      "loss": 0.2014,
+      "step": 6365
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000255318482743721,
+      "loss": 0.5552,
+      "step": 6366
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002553049658998123,
+      "loss": 1.155,
+      "step": 6367
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025529144736960256,
+      "loss": 0.8787,
+      "step": 6368
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002552779271533083,
+      "loss": 0.0208,
+      "step": 6369
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002552644052511461,
+      "loss": 0.5809,
+      "step": 6370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002552508816633324,
+      "loss": 0.3491,
+      "step": 6371
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002552373563900838,
+      "loss": 0.2922,
+      "step": 6372
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002552238294316169,
+      "loss": 0.7052,
+      "step": 6373
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002552103007881482,
+      "loss": 0.4723,
+      "step": 6374
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025519677045989447,
+      "loss": 0.5668,
+      "step": 6375
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025518323844707234,
+      "loss": 0.3159,
+      "step": 6376
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025516970474989854,
+      "loss": 0.363,
+      "step": 6377
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025515616936858977,
+      "loss": 0.7672,
+      "step": 6378
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002551426323033627,
+      "loss": 0.3744,
+      "step": 6379
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002551290935544342,
+      "loss": 0.7155,
+      "step": 6380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025511555312202104,
+      "loss": 0.88,
+      "step": 6381
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025510201100634004,
+      "loss": 0.4947,
+      "step": 6382
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002550884672076081,
+      "loss": 0.6813,
+      "step": 6383
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025507492172604205,
+      "loss": 0.1428,
+      "step": 6384
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025506137456185885,
+      "loss": 0.0211,
+      "step": 6385
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002550478257152754,
+      "loss": 0.8307,
+      "step": 6386
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002550342751865087,
+      "loss": 0.6231,
+      "step": 6387
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025502072297577566,
+      "loss": 0.4428,
+      "step": 6388
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002550071690832934,
+      "loss": 0.1908,
+      "step": 6389
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002549936135092789,
+      "loss": 0.8768,
+      "step": 6390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025498005625394923,
+      "loss": 0.4323,
+      "step": 6391
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025496649731752155,
+      "loss": 1.176,
+      "step": 6392
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025495293670021297,
+      "loss": 0.6567,
+      "step": 6393
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002549393744022405,
+      "loss": 0.3195,
+      "step": 6394
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025492581042382156,
+      "loss": 1.1861,
+      "step": 6395
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025491224476517323,
+      "loss": 0.3392,
+      "step": 6396
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025489867742651276,
+      "loss": 0.2941,
+      "step": 6397
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002548851084080573,
+      "loss": 1.7386,
+      "step": 6398
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025487153771002434,
+      "loss": 0.4222,
+      "step": 6399
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002548579653326311,
+      "loss": 0.5007,
+      "step": 6400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025484439127609486,
+      "loss": 0.3249,
+      "step": 6401
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025483081554063307,
+      "loss": 0.2759,
+      "step": 6402
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002548172381264631,
+      "loss": 0.6943,
+      "step": 6403
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025480365903380236,
+      "loss": 0.0086,
+      "step": 6404
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025479007826286826,
+      "loss": 0.7897,
+      "step": 6405
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002547764958138784,
+      "loss": 0.6401,
+      "step": 6406
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025476291168705014,
+      "loss": 1.1216,
+      "step": 6407
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025474932588260115,
+      "loss": 2.1812,
+      "step": 6408
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002547357384007489,
+      "loss": 0.7792,
+      "step": 6409
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002547221492417109,
+      "loss": 0.3649,
+      "step": 6410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025470855840570495,
+      "loss": 2.014,
+      "step": 6411
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025469496589294855,
+      "loss": 0.9571,
+      "step": 6412
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025468137170365937,
+      "loss": 0.9263,
+      "step": 6413
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025466777583805513,
+      "loss": 0.1944,
+      "step": 6414
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025465417829635364,
+      "loss": 0.356,
+      "step": 6415
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025464057907877244,
+      "loss": 0.044,
+      "step": 6416
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002546269781855294,
+      "loss": 0.3138,
+      "step": 6417
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002546133756168424,
+      "loss": 0.0958,
+      "step": 6418
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002545997713729292,
+      "loss": 0.5653,
+      "step": 6419
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025458616545400754,
+      "loss": 0.3563,
+      "step": 6420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002545725578602955,
+      "loss": 1.0659,
+      "step": 6421
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002545589485920109,
+      "loss": 0.2388,
+      "step": 6422
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002545453376493715,
+      "loss": 0.0126,
+      "step": 6423
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025453172503259555,
+      "loss": 0.7707,
+      "step": 6424
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002545181107419009,
+      "loss": 0.0039,
+      "step": 6425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002545044947775056,
+      "loss": 0.2855,
+      "step": 6426
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025449087713962754,
+      "loss": 0.7046,
+      "step": 6427
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000254477257828485,
+      "loss": 0.1637,
+      "step": 6428
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002544636368442959,
+      "loss": 0.0057,
+      "step": 6429
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002544500141872785,
+      "loss": 0.4642,
+      "step": 6430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025443638985765086,
+      "loss": 0.6154,
+      "step": 6431
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025442276385563117,
+      "loss": 0.0235,
+      "step": 6432
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025440913618143764,
+      "loss": 0.5061,
+      "step": 6433
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025439550683528845,
+      "loss": 0.041,
+      "step": 6434
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002543818758174019,
+      "loss": 0.2987,
+      "step": 6435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025436824312799635,
+      "loss": 0.3673,
+      "step": 6436
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002543546087672899,
+      "loss": 0.3132,
+      "step": 6437
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002543409727355011,
+      "loss": 1.7221,
+      "step": 6438
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002543273350328482,
+      "loss": 0.2696,
+      "step": 6439
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002543136956595496,
+      "loss": 0.2934,
+      "step": 6440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002543000546158237,
+      "loss": 1.4502,
+      "step": 6441
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025428641190188895,
+      "loss": 0.0109,
+      "step": 6442
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025427276751796384,
+      "loss": 0.4706,
+      "step": 6443
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002542591214642668,
+      "loss": 0.5243,
+      "step": 6444
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025424547374101647,
+      "loss": 0.627,
+      "step": 6445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025423182434843137,
+      "loss": 0.283,
+      "step": 6446
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002542181732867299,
+      "loss": 0.8161,
+      "step": 6447
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025420452055613087,
+      "loss": 2.8732,
+      "step": 6448
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025419086615685284,
+      "loss": 0.0037,
+      "step": 6449
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002541772100891144,
+      "loss": 0.695,
+      "step": 6450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002541635523531343,
+      "loss": 1.1693,
+      "step": 6451
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002541498929491313,
+      "loss": 0.375,
+      "step": 6452
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025413623187732406,
+      "loss": 0.4115,
+      "step": 6453
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002541225691379313,
+      "loss": 0.2386,
+      "step": 6454
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025410890473117194,
+      "loss": 0.9096,
+      "step": 6455
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025409523865726466,
+      "loss": 0.0281,
+      "step": 6456
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025408157091642836,
+      "loss": 3.0434,
+      "step": 6457
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002540679015088819,
+      "loss": 0.238,
+      "step": 6458
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002540542304348442,
+      "loss": 0.0133,
+      "step": 6459
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025404055769453414,
+      "loss": 0.3087,
+      "step": 6460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002540268832881707,
+      "loss": 1.0115,
+      "step": 6461
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002540132072159728,
+      "loss": 0.3341,
+      "step": 6462
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025399952947815965,
+      "loss": 0.6891,
+      "step": 6463
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025398585007494997,
+      "loss": 0.3882,
+      "step": 6464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000253972169006563,
+      "loss": 0.3798,
+      "step": 6465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002539584862732178,
+      "loss": 0.2271,
+      "step": 6466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025394480187513344,
+      "loss": 3.7645,
+      "step": 6467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002539311158125291,
+      "loss": 0.3073,
+      "step": 6468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002539174280856239,
+      "loss": 3.9351,
+      "step": 6469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000253903738694637,
+      "loss": 0.2527,
+      "step": 6470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002538900476397878,
+      "loss": 1.4803,
+      "step": 6471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002538763549212953,
+      "loss": 0.9539,
+      "step": 6472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002538626605393789,
+      "loss": 0.2982,
+      "step": 6473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025384896449425783,
+      "loss": 0.6885,
+      "step": 6474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002538352667861515,
+      "loss": 0.0028,
+      "step": 6475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025382156741527914,
+      "loss": 0.0849,
+      "step": 6476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025380786638186023,
+      "loss": 0.5625,
+      "step": 6477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025379416368611415,
+      "loss": 0.0113,
+      "step": 6478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002537804593282603,
+      "loss": 0.4616,
+      "step": 6479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025376675330851815,
+      "loss": 0.0169,
+      "step": 6480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002537530456271072,
+      "loss": 1.2512,
+      "step": 6481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002537393362842469,
+      "loss": 0.8229,
+      "step": 6482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002537256252801568,
+      "loss": 0.4949,
+      "step": 6483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025371191261505655,
+      "loss": 0.232,
+      "step": 6484
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025369819828916564,
+      "loss": 0.6766,
+      "step": 6485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002536844823027037,
+      "loss": 0.304,
+      "step": 6486
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002536707646558904,
+      "loss": 0.3208,
+      "step": 6487
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002536570453489454,
+      "loss": 0.4091,
+      "step": 6488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025364332438208837,
+      "loss": 0.5747,
+      "step": 6489
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000253629601755539,
+      "loss": 0.6064,
+      "step": 6490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002536158774695172,
+      "loss": 2.7385,
+      "step": 6491
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025360215152424255,
+      "loss": 3.314,
+      "step": 6492
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002535884239199349,
+      "loss": 1.0819,
+      "step": 6493
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002535746946568142,
+      "loss": 0.2884,
+      "step": 6494
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002535609637351001,
+      "loss": 0.597,
+      "step": 6495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025354723115501263,
+      "loss": 2.306,
+      "step": 6496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002535334969167717,
+      "loss": 0.6371,
+      "step": 6497
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002535197610205972,
+      "loss": 0.4124,
+      "step": 6498
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025350602346670896,
+      "loss": 0.0215,
+      "step": 6499
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025349228425532725,
+      "loss": 0.9634,
+      "step": 6500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025347854338667183,
+      "loss": 0.2838,
+      "step": 6501
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002534648008609629,
+      "loss": 0.7892,
+      "step": 6502
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002534510566784204,
+      "loss": 0.4297,
+      "step": 6503
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025343731083926456,
+      "loss": 0.7924,
+      "step": 6504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002534235633437154,
+      "loss": 0.8108,
+      "step": 6505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002534098141919931,
+      "loss": 0.5259,
+      "step": 6506
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025339606338431773,
+      "loss": 0.9629,
+      "step": 6507
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002533823109209097,
+      "loss": 0.2242,
+      "step": 6508
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025336855680198903,
+      "loss": 0.0175,
+      "step": 6509
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025335480102777613,
+      "loss": 0.8807,
+      "step": 6510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025334104359849116,
+      "loss": 0.0265,
+      "step": 6511
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025332728451435443,
+      "loss": 1.8916,
+      "step": 6512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002533135237755864,
+      "loss": 0.7596,
+      "step": 6513
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025329976138240733,
+      "loss": 0.8454,
+      "step": 6514
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025328599733503754,
+      "loss": 0.0204,
+      "step": 6515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025327223163369754,
+      "loss": 0.5592,
+      "step": 6516
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002532584642786078,
+      "loss": 0.5859,
+      "step": 6517
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002532446952699887,
+      "loss": 0.0283,
+      "step": 6518
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002532309246080607,
+      "loss": 0.5509,
+      "step": 6519
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002532171522930444,
+      "loss": 0.0968,
+      "step": 6520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002532033783251603,
+      "loss": 0.0236,
+      "step": 6521
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000253189602704629,
+      "loss": 0.0405,
+      "step": 6522
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002531758254316711,
+      "loss": 1.6274,
+      "step": 6523
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025316204650650717,
+      "loss": 0.6326,
+      "step": 6524
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002531482659293579,
+      "loss": 0.2168,
+      "step": 6525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002531344837004439,
+      "loss": 0.3704,
+      "step": 6526
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025312069981998605,
+      "loss": 0.2643,
+      "step": 6527
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002531069142882049,
+      "loss": 0.3465,
+      "step": 6528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002530931271053212,
+      "loss": 0.4584,
+      "step": 6529
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025307933827155586,
+      "loss": 3.5586,
+      "step": 6530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002530655477871296,
+      "loss": 0.2096,
+      "step": 6531
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025305175565226327,
+      "loss": 3.2761,
+      "step": 6532
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025303796186717766,
+      "loss": 0.6692,
+      "step": 6533
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002530241664320938,
+      "loss": 0.0081,
+      "step": 6534
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025301036934723256,
+      "loss": 0.3289,
+      "step": 6535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025299657061281477,
+      "loss": 0.6849,
+      "step": 6536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002529827702290616,
+      "loss": 0.6338,
+      "step": 6537
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025296896819619385,
+      "loss": 1.0297,
+      "step": 6538
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002529551645144326,
+      "loss": 0.4744,
+      "step": 6539
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002529413591839989,
+      "loss": 0.3844,
+      "step": 6540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002529275522051138,
+      "loss": 0.2672,
+      "step": 6541
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002529137435779985,
+      "loss": 0.5912,
+      "step": 6542
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000252899933302874,
+      "loss": 1.0303,
+      "step": 6543
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002528861213799615,
+      "loss": 0.2164,
+      "step": 6544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002528723078094822,
+      "loss": 1.7707,
+      "step": 6545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025285849259165725,
+      "loss": 0.8184,
+      "step": 6546
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025284467572670794,
+      "loss": 0.4772,
+      "step": 6547
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025283085721485547,
+      "loss": 0.6848,
+      "step": 6548
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002528170370563212,
+      "loss": 0.8682,
+      "step": 6549
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002528032152513263,
+      "loss": 0.2428,
+      "step": 6550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025278939180009227,
+      "loss": 0.4286,
+      "step": 6551
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002527755667028403,
+      "loss": 0.5398,
+      "step": 6552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000252761739959792,
+      "loss": 0.1621,
+      "step": 6553
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002527479115711686,
+      "loss": 0.1643,
+      "step": 6554
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002527340815371916,
+      "loss": 0.5768,
+      "step": 6555
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025272024985808247,
+      "loss": 2.7004,
+      "step": 6556
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002527064165340627,
+      "loss": 0.3112,
+      "step": 6557
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002526925815653538,
+      "loss": 0.6827,
+      "step": 6558
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025267874495217736,
+      "loss": 0.4907,
+      "step": 6559
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025266490669475486,
+      "loss": 0.5592,
+      "step": 6560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025265106679330803,
+      "loss": 0.3761,
+      "step": 6561
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002526372252480584,
+      "loss": 0.4632,
+      "step": 6562
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002526233820592276,
+      "loss": 0.3664,
+      "step": 6563
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025260953722703737,
+      "loss": 0.1897,
+      "step": 6564
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002525956907517094,
+      "loss": 1.1718,
+      "step": 6565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025258184263346543,
+      "loss": 0.015,
+      "step": 6566
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002525679928725272,
+      "loss": 0.2266,
+      "step": 6567
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025255414146911647,
+      "loss": 1.3886,
+      "step": 6568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025254028842345514,
+      "loss": 0.5345,
+      "step": 6569
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002525264337357649,
+      "loss": 0.4204,
+      "step": 6570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025251257740626775,
+      "loss": 0.0215,
+      "step": 6571
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002524987194351855,
+      "loss": 0.9515,
+      "step": 6572
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002524848598227401,
+      "loss": 2.5522,
+      "step": 6573
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002524709985691535,
+      "loss": 0.8953,
+      "step": 6574
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002524571356746476,
+      "loss": 0.2917,
+      "step": 6575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002524432711394445,
+      "loss": 0.0216,
+      "step": 6576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002524294049637661,
+      "loss": 0.9363,
+      "step": 6577
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025241553714783454,
+      "loss": 0.0133,
+      "step": 6578
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002524016676918718,
+      "loss": 0.1265,
+      "step": 6579
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002523877965961001,
+      "loss": 0.8745,
+      "step": 6580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002523739238607415,
+      "loss": 0.5799,
+      "step": 6581
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002523600494860181,
+      "loss": 1.0207,
+      "step": 6582
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002523461734721522,
+      "loss": 1.5298,
+      "step": 6583
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002523322958193658,
+      "loss": 0.714,
+      "step": 6584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025231841652788143,
+      "loss": 0.6866,
+      "step": 6585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025230453559792106,
+      "loss": 0.4121,
+      "step": 6586
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002522906530297071,
+      "loss": 1.095,
+      "step": 6587
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002522767688234619,
+      "loss": 0.3505,
+      "step": 6588
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002522628829794077,
+      "loss": 2.4573,
+      "step": 6589
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002522489954977669,
+      "loss": 0.2687,
+      "step": 6590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025223510637876193,
+      "loss": 0.0053,
+      "step": 6591
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002522212156226152,
+      "loss": 1.0917,
+      "step": 6592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025220732322954904,
+      "loss": 0.7717,
+      "step": 6593
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000252193429199786,
+      "loss": 4.0628,
+      "step": 6594
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025217953353354856,
+      "loss": 1.2158,
+      "step": 6595
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025216563623105926,
+      "loss": 0.4095,
+      "step": 6596
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002521517372925406,
+      "loss": 1.1725,
+      "step": 6597
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002521378367182152,
+      "loss": 0.6407,
+      "step": 6598
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002521239345083057,
+      "loss": 0.6649,
+      "step": 6599
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025211003066303455,
+      "loss": 1.5421,
+      "step": 6600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002520961251826245,
+      "loss": 0.3022,
+      "step": 6601
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002520822180672983,
+      "loss": 1.0069,
+      "step": 6602
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002520683093172786,
+      "loss": 0.6184,
+      "step": 6603
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00025205439893278804,
+      "loss": 0.3694,
+      "step": 6604
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002520404869140494,
+      "loss": 0.4024,
+      "step": 6605
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025202657326128564,
+      "loss": 0.0209,
+      "step": 6606
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025201265797471933,
+      "loss": 0.9746,
+      "step": 6607
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002519987410545734,
+      "loss": 0.7044,
+      "step": 6608
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025198482250107074,
+      "loss": 1.5031,
+      "step": 6609
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025197090231443423,
+      "loss": 0.3292,
+      "step": 6610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025195698049488674,
+      "loss": 0.364,
+      "step": 6611
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002519430570426512,
+      "loss": 0.2532,
+      "step": 6612
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002519291319579506,
+      "loss": 0.7526,
+      "step": 6613
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025191520524100796,
+      "loss": 0.1781,
+      "step": 6614
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025190127689204625,
+      "loss": 0.1674,
+      "step": 6615
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002518873469112885,
+      "loss": 0.2938,
+      "step": 6616
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025187341529895777,
+      "loss": 0.131,
+      "step": 6617
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025185948205527726,
+      "loss": 0.8844,
+      "step": 6618
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025184554718047,
+      "loss": 0.3935,
+      "step": 6619
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025183161067475913,
+      "loss": 0.6024,
+      "step": 6620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002518176725383678,
+      "loss": 0.0392,
+      "step": 6621
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002518037327715193,
+      "loss": 0.3518,
+      "step": 6622
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025178979137443683,
+      "loss": 0.9208,
+      "step": 6623
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002517758483473436,
+      "loss": 0.5875,
+      "step": 6624
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002517619036904629,
+      "loss": 0.3099,
+      "step": 6625
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000251747957404018,
+      "loss": 0.4198,
+      "step": 6626
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025173400948823234,
+      "loss": 0.1125,
+      "step": 6627
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002517200599433291,
+      "loss": 0.6485,
+      "step": 6628
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025170610876953177,
+      "loss": 0.741,
+      "step": 6629
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025169215596706384,
+      "loss": 0.1609,
+      "step": 6630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002516782015361486,
+      "loss": 2.4036,
+      "step": 6631
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002516642454770095,
+      "loss": 1.8219,
+      "step": 6632
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002516502877898702,
+      "loss": 0.8431,
+      "step": 6633
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025163632847495407,
+      "loss": 0.8826,
+      "step": 6634
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002516223675324846,
+      "loss": 0.8546,
+      "step": 6635
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025160840496268546,
+      "loss": 0.0135,
+      "step": 6636
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025159444076578027,
+      "loss": 0.7565,
+      "step": 6637
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002515804749419925,
+      "loss": 1.1947,
+      "step": 6638
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025156650749154585,
+      "loss": 0.0202,
+      "step": 6639
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025155253841466414,
+      "loss": 0.4071,
+      "step": 6640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002515385677115709,
+      "loss": 0.1676,
+      "step": 6641
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025152459538248977,
+      "loss": 0.177,
+      "step": 6642
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025151062142764473,
+      "loss": 0.2526,
+      "step": 6643
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002514966458472594,
+      "loss": 0.1236,
+      "step": 6644
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025148266864155757,
+      "loss": 0.7906,
+      "step": 6645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025146868981076317,
+      "loss": 0.2465,
+      "step": 6646
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025145470935509996,
+      "loss": 0.6085,
+      "step": 6647
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025144072727479184,
+      "loss": 0.626,
+      "step": 6648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025142674357006274,
+      "loss": 0.1788,
+      "step": 6649
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025141275824113656,
+      "loss": 0.3715,
+      "step": 6650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002513987712882372,
+      "loss": 1.6299,
+      "step": 6651
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002513847827115888,
+      "loss": 0.8116,
+      "step": 6652
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025137079251141516,
+      "loss": 0.557,
+      "step": 6653
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002513568006879404,
+      "loss": 2.252,
+      "step": 6654
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002513428072413887,
+      "loss": 0.4143,
+      "step": 6655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025132881217198395,
+      "loss": 0.3199,
+      "step": 6656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002513148154799504,
+      "loss": 0.1545,
+      "step": 6657
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025130081716551205,
+      "loss": 0.7193,
+      "step": 6658
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002512868172288932,
+      "loss": 0.5294,
+      "step": 6659
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000251272815670318,
+      "loss": 0.1543,
+      "step": 6660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002512588124900107,
+      "loss": 0.941,
+      "step": 6661
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002512448076881954,
+      "loss": 0.0323,
+      "step": 6662
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025123080126509645,
+      "loss": 0.5004,
+      "step": 6663
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002512167932209381,
+      "loss": 0.221,
+      "step": 6664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002512027835559448,
+      "loss": 0.3307,
+      "step": 6665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002511887722703408,
+      "loss": 0.2637,
+      "step": 6666
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025117475936435043,
+      "loss": 1.7852,
+      "step": 6667
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025116074483819817,
+      "loss": 0.2167,
+      "step": 6668
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002511467286921084,
+      "loss": 0.5465,
+      "step": 6669
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025113271092630556,
+      "loss": 2.2735,
+      "step": 6670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002511186915410141,
+      "loss": 0.395,
+      "step": 6671
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002511046705364586,
+      "loss": 0.2875,
+      "step": 6672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025109064791286353,
+      "loss": 0.4564,
+      "step": 6673
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002510766236704534,
+      "loss": 0.0173,
+      "step": 6674
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002510625978094529,
+      "loss": 0.0942,
+      "step": 6675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025104857033008657,
+      "loss": 0.511,
+      "step": 6676
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000251034541232579,
+      "loss": 0.2016,
+      "step": 6677
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002510205105171549,
+      "loss": 0.2132,
+      "step": 6678
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025100647818403893,
+      "loss": 1.1191,
+      "step": 6679
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002509924442334558,
+      "loss": 0.4368,
+      "step": 6680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025097840866563026,
+      "loss": 0.6582,
+      "step": 6681
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025096437148078705,
+      "loss": 0.7959,
+      "step": 6682
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025095033267915094,
+      "loss": 0.1699,
+      "step": 6683
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002509362922609467,
+      "loss": 0.6207,
+      "step": 6684
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025092225022639933,
+      "loss": 0.244,
+      "step": 6685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025090820657573356,
+      "loss": 0.1402,
+      "step": 6686
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002508941613091743,
+      "loss": 0.0556,
+      "step": 6687
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002508801144269464,
+      "loss": 0.4042,
+      "step": 6688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000250866065929275,
+      "loss": 0.4801,
+      "step": 6689
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002508520158163848,
+      "loss": 0.413,
+      "step": 6690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000250837964088501,
+      "loss": 0.2833,
+      "step": 6691
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025082391074584854,
+      "loss": 1.0782,
+      "step": 6692
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025080985578865246,
+      "loss": 0.624,
+      "step": 6693
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002507957992171378,
+      "loss": 0.6182,
+      "step": 6694
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002507817410315297,
+      "loss": 2.0314,
+      "step": 6695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002507676812320533,
+      "loss": 1.0968,
+      "step": 6696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002507536198189337,
+      "loss": 0.2978,
+      "step": 6697
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002507395567923961,
+      "loss": 0.4151,
+      "step": 6698
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025072549215266566,
+      "loss": 0.362,
+      "step": 6699
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025071142589996767,
+      "loss": 0.5702,
+      "step": 6700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002506973580345273,
+      "loss": 0.6079,
+      "step": 6701
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002506832885565699,
+      "loss": 0.5605,
+      "step": 6702
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002506692174663207,
+      "loss": 0.4069,
+      "step": 6703
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025065514476400506,
+      "loss": 0.547,
+      "step": 6704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002506410704498484,
+      "loss": 0.8013,
+      "step": 6705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000250626994524076,
+      "loss": 1.0477,
+      "step": 6706
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025061291698691326,
+      "loss": 0.0564,
+      "step": 6707
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002505988378385857,
+      "loss": 1.1389,
+      "step": 6708
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025058475707931876,
+      "loss": 1.1692,
+      "step": 6709
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002505706747093378,
+      "loss": 0.4691,
+      "step": 6710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002505565907288685,
+      "loss": 0.2545,
+      "step": 6711
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002505425051381363,
+      "loss": 0.6327,
+      "step": 6712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002505284179373668,
+      "loss": 1.1813,
+      "step": 6713
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025051432912678547,
+      "loss": 0.2034,
+      "step": 6714
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025050023870661813,
+      "loss": 0.0803,
+      "step": 6715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002504861466770902,
+      "loss": 0.8307,
+      "step": 6716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002504720530384275,
+      "loss": 0.2514,
+      "step": 6717
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025045795779085557,
+      "loss": 0.0027,
+      "step": 6718
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002504438609346003,
+      "loss": 0.3593,
+      "step": 6719
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025042976246988734,
+      "loss": 0.9719,
+      "step": 6720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025041566239694245,
+      "loss": 0.3573,
+      "step": 6721
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025040156071599136,
+      "loss": 0.7524,
+      "step": 6722
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025038745742726005,
+      "loss": 0.5005,
+      "step": 6723
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002503733525309742,
+      "loss": 0.7097,
+      "step": 6724
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002503592460273598,
+      "loss": 0.2721,
+      "step": 6725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025034513791664264,
+      "loss": 0.2703,
+      "step": 6726
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025033102819904874,
+      "loss": 0.3648,
+      "step": 6727
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000250316916874804,
+      "loss": 0.4894,
+      "step": 6728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025030280394413434,
+      "loss": 0.0615,
+      "step": 6729
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025028868940726585,
+      "loss": 0.4345,
+      "step": 6730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025027457326442447,
+      "loss": 0.5727,
+      "step": 6731
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025026045551583635,
+      "loss": 0.1357,
+      "step": 6732
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025024633616172744,
+      "loss": 0.0555,
+      "step": 6733
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000250232215202324,
+      "loss": 0.3952,
+      "step": 6734
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025021809263785195,
+      "loss": 0.274,
+      "step": 6735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025020396846853763,
+      "loss": 0.8609,
+      "step": 6736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025018984269460713,
+      "loss": 0.5045,
+      "step": 6737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002501757153162866,
+      "loss": 0.3896,
+      "step": 6738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002501615863338024,
+      "loss": 1.1597,
+      "step": 6739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025014745574738074,
+      "loss": 1.589,
+      "step": 6740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002501333235572478,
+      "loss": 0.7823,
+      "step": 6741
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025011918976363003,
+      "loss": 0.635,
+      "step": 6742
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025010505436675364,
+      "loss": 0.0193,
+      "step": 6743
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025009091736684504,
+      "loss": 0.0234,
+      "step": 6744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025007677876413067,
+      "loss": 0.8658,
+      "step": 6745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025006263855883684,
+      "loss": 0.7374,
+      "step": 6746
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025004849675119004,
+      "loss": 0.3345,
+      "step": 6747
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002500343533414167,
+      "loss": 0.0177,
+      "step": 6748
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002500202083297433,
+      "loss": 0.8892,
+      "step": 6749
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00025000606171639647,
+      "loss": 0.8393,
+      "step": 6750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024999191350160257,
+      "loss": 0.6319,
+      "step": 6751
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024997776368558826,
+      "loss": 0.7129,
+      "step": 6752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002499636122685801,
+      "loss": 5.2671,
+      "step": 6753
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002499494592508047,
+      "loss": 0.3776,
+      "step": 6754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002499353046324888,
+      "loss": 0.5029,
+      "step": 6755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024992114841385885,
+      "loss": 0.152,
+      "step": 6756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002499069905951417,
+      "loss": 0.4112,
+      "step": 6757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002498928311765641,
+      "loss": 0.5773,
+      "step": 6758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002498786701583527,
+      "loss": 2.8068,
+      "step": 6759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024986450754073424,
+      "loss": 0.7678,
+      "step": 6760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024985034332393566,
+      "loss": 1.0945,
+      "step": 6761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024983617750818363,
+      "loss": 0.6977,
+      "step": 6762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002498220100937051,
+      "loss": 0.5358,
+      "step": 6763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002498078410807268,
+      "loss": 4.2771,
+      "step": 6764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002497936704694758,
+      "loss": 0.5908,
+      "step": 6765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002497794982601789,
+      "loss": 0.0714,
+      "step": 6766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024976532445306304,
+      "loss": 0.3699,
+      "step": 6767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002497511490483553,
+      "loss": 0.2326,
+      "step": 6768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002497369720462826,
+      "loss": 0.5193,
+      "step": 6769
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024972279344707195,
+      "loss": 0.3121,
+      "step": 6770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024970861325095047,
+      "loss": 0.362,
+      "step": 6771
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024969443145814516,
+      "loss": 0.2766,
+      "step": 6772
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002496802480688832,
+      "loss": 0.6257,
+      "step": 6773
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002496660630833916,
+      "loss": 1.8356,
+      "step": 6774
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002496518765018976,
+      "loss": 1.2719,
+      "step": 6775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002496376883246284,
+      "loss": 0.0278,
+      "step": 6776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024962349855181114,
+      "loss": 0.3998,
+      "step": 6777
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000249609307183673,
+      "loss": 0.2624,
+      "step": 6778
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024959511422044134,
+      "loss": 0.7651,
+      "step": 6779
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024958091966234336,
+      "loss": 0.5444,
+      "step": 6780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024956672350960645,
+      "loss": 0.8419,
+      "step": 6781
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002495525257624579,
+      "loss": 0.4826,
+      "step": 6782
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024953832642112504,
+      "loss": 0.7057,
+      "step": 6783
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002495241254858352,
+      "loss": 0.216,
+      "step": 6784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000249509922956816,
+      "loss": 0.4803,
+      "step": 6785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002494957188342946,
+      "loss": 0.5811,
+      "step": 6786
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024948151311849866,
+      "loss": 0.5971,
+      "step": 6787
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002494673058096555,
+      "loss": 0.3309,
+      "step": 6788
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002494530969079928,
+      "loss": 0.0935,
+      "step": 6789
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024943888641373803,
+      "loss": 2.3265,
+      "step": 6790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024942467432711866,
+      "loss": 0.22,
+      "step": 6791
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024941046064836243,
+      "loss": 0.4117,
+      "step": 6792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024939624537769677,
+      "loss": 0.2593,
+      "step": 6793
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024938202851534946,
+      "loss": 0.3858,
+      "step": 6794
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002493678100615481,
+      "loss": 1.0316,
+      "step": 6795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002493535900165204,
+      "loss": 0.3416,
+      "step": 6796
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002493393683804941,
+      "loss": 0.3928,
+      "step": 6797
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024932514515369684,
+      "loss": 0.4255,
+      "step": 6798
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002493109203363565,
+      "loss": 0.3017,
+      "step": 6799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002492966939287008,
+      "loss": 0.2958,
+      "step": 6800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024928246593095753,
+      "loss": 0.4665,
+      "step": 6801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002492682363433546,
+      "loss": 0.5831,
+      "step": 6802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002492540051661199,
+      "loss": 0.2361,
+      "step": 6803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002492397723994812,
+      "loss": 0.2647,
+      "step": 6804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002492255380436665,
+      "loss": 1.2032,
+      "step": 6805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024921130209890367,
+      "loss": 1.4281,
+      "step": 6806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002491970645654208,
+      "loss": 0.4353,
+      "step": 6807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002491828254434458,
+      "loss": 1.2652,
+      "step": 6808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002491685847332067,
+      "loss": 0.1805,
+      "step": 6809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024915434243493156,
+      "loss": 0.4518,
+      "step": 6810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002491400985488484,
+      "loss": 1.7065,
+      "step": 6811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024912585307518536,
+      "loss": 0.0336,
+      "step": 6812
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002491116060141706,
+      "loss": 0.4744,
+      "step": 6813
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002490973573660322,
+      "loss": 0.8213,
+      "step": 6814
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024908310713099834,
+      "loss": 0.4457,
+      "step": 6815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024906885530929715,
+      "loss": 0.8406,
+      "step": 6816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000249054601901157,
+      "loss": 0.255,
+      "step": 6817
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002490403469068061,
+      "loss": 0.3266,
+      "step": 6818
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002490260903264726,
+      "loss": 4.9414,
+      "step": 6819
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024901183216038486,
+      "loss": 0.2265,
+      "step": 6820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002489975724087713,
+      "loss": 0.2601,
+      "step": 6821
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024898331107186015,
+      "loss": 0.4454,
+      "step": 6822
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002489690481498799,
+      "loss": 0.2998,
+      "step": 6823
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002489547836430587,
+      "loss": 0.893,
+      "step": 6824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002489405175516253,
+      "loss": 3.5445,
+      "step": 6825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000248926249875808,
+      "loss": 0.2163,
+      "step": 6826
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002489119806158352,
+      "loss": 0.1729,
+      "step": 6827
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024889770977193555,
+      "loss": 0.9933,
+      "step": 6828
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002488834373443375,
+      "loss": 0.7112,
+      "step": 6829
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002488691633332696,
+      "loss": 0.1916,
+      "step": 6830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024885488773896037,
+      "loss": 0.1182,
+      "step": 6831
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002488406105616385,
+      "loss": 0.0997,
+      "step": 6832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002488263318015326,
+      "loss": 0.0356,
+      "step": 6833
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024881205145887135,
+      "loss": 0.4201,
+      "step": 6834
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002487977695338833,
+      "loss": 0.4307,
+      "step": 6835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002487834860267973,
+      "loss": 1.5319,
+      "step": 6836
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000248769200937842,
+      "loss": 0.4198,
+      "step": 6837
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002487549142672463,
+      "loss": 0.4538,
+      "step": 6838
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024874062601523873,
+      "loss": 0.7649,
+      "step": 6839
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002487263361820483,
+      "loss": 0.4266,
+      "step": 6840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002487120447679037,
+      "loss": 0.7988,
+      "step": 6841
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024869775177303383,
+      "loss": 0.4956,
+      "step": 6842
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024868345719766767,
+      "loss": 0.9541,
+      "step": 6843
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000248669161042034,
+      "loss": 0.4389,
+      "step": 6844
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002486548633063618,
+      "loss": 0.7926,
+      "step": 6845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024864056399088,
+      "loss": 1.7551,
+      "step": 6846
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024862626309581765,
+      "loss": 0.2904,
+      "step": 6847
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002486119606214037,
+      "loss": 0.3466,
+      "step": 6848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002485976565678672,
+      "loss": 0.0122,
+      "step": 6849
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024858335093543715,
+      "loss": 0.2016,
+      "step": 6850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002485690437243428,
+      "loss": 0.2711,
+      "step": 6851
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00024855473493481305,
+      "loss": 0.7385,
+      "step": 6852
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002485404245670772,
+      "loss": 0.4329,
+      "step": 6853
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002485261126213643,
+      "loss": 1.1471,
+      "step": 6854
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024851179909790355,
+      "loss": 0.4013,
+      "step": 6855
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002484974839969242,
+      "loss": 0.0691,
+      "step": 6856
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002484831673186555,
+      "loss": 0.3201,
+      "step": 6857
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002484688490633266,
+      "loss": 0.3637,
+      "step": 6858
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002484545292311669,
+      "loss": 0.7673,
+      "step": 6859
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024844020782240577,
+      "loss": 0.4772,
+      "step": 6860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024842588483727233,
+      "loss": 0.5665,
+      "step": 6861
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024841156027599615,
+      "loss": 0.4517,
+      "step": 6862
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024839723413880646,
+      "loss": 0.8146,
+      "step": 6863
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024838290642593275,
+      "loss": 0.0158,
+      "step": 6864
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002483685771376044,
+      "loss": 0.3642,
+      "step": 6865
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000248354246274051,
+      "loss": 0.4462,
+      "step": 6866
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024833991383550193,
+      "loss": 0.7851,
+      "step": 6867
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002483255798221867,
+      "loss": 0.1211,
+      "step": 6868
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002483112442343349,
+      "loss": 0.6889,
+      "step": 6869
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002482969070721761,
+      "loss": 0.0953,
+      "step": 6870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024828256833593984,
+      "loss": 0.1362,
+      "step": 6871
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024826822802585574,
+      "loss": 0.6201,
+      "step": 6872
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002482538861421534,
+      "loss": 0.8123,
+      "step": 6873
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002482395426850626,
+      "loss": 0.9424,
+      "step": 6874
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002482251976548129,
+      "loss": 0.0781,
+      "step": 6875
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002482108510516341,
+      "loss": 0.0138,
+      "step": 6876
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024819650287575596,
+      "loss": 0.1326,
+      "step": 6877
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024818215312740816,
+      "loss": 0.5888,
+      "step": 6878
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002481678018068205,
+      "loss": 0.0006,
+      "step": 6879
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002481534489142229,
+      "loss": 0.1525,
+      "step": 6880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024813909444984506,
+      "loss": 0.1162,
+      "step": 6881
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024812473841391694,
+      "loss": 0.6757,
+      "step": 6882
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002481103808066684,
+      "loss": 0.4326,
+      "step": 6883
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002480960216283293,
+      "loss": 0.5451,
+      "step": 6884
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024808166087912967,
+      "loss": 1.6702,
+      "step": 6885
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024806729855929944,
+      "loss": 1.1167,
+      "step": 6886
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002480529346690686,
+      "loss": 1.0437,
+      "step": 6887
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002480385692086672,
+      "loss": 0.188,
+      "step": 6888
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002480242021783252,
+      "loss": 2.2782,
+      "step": 6889
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024800983357827275,
+      "loss": 0.2826,
+      "step": 6890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002479954634087399,
+      "loss": 0.2869,
+      "step": 6891
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024798109166995676,
+      "loss": 0.8113,
+      "step": 6892
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002479667183621535,
+      "loss": 0.2399,
+      "step": 6893
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024795234348556027,
+      "loss": 0.3654,
+      "step": 6894
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002479379670404073,
+      "loss": 0.5109,
+      "step": 6895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002479235890269247,
+      "loss": 0.005,
+      "step": 6896
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024790920944534284,
+      "loss": 0.3798,
+      "step": 6897
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002478948282958919,
+      "loss": 1.0644,
+      "step": 6898
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002478804455788022,
+      "loss": 0.8043,
+      "step": 6899
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002478660612943041,
+      "loss": 0.7698,
+      "step": 6900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002478516754426279,
+      "loss": 1.2444,
+      "step": 6901
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024783728802400394,
+      "loss": 0.8269,
+      "step": 6902
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024782289903866266,
+      "loss": 0.9544,
+      "step": 6903
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002478085084868345,
+      "loss": 0.2887,
+      "step": 6904
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002477941163687498,
+      "loss": 0.0234,
+      "step": 6905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002477797226846392,
+      "loss": 0.4149,
+      "step": 6906
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024776532743473297,
+      "loss": 0.4223,
+      "step": 6907
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002477509306192618,
+      "loss": 0.7223,
+      "step": 6908
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024773653223845614,
+      "loss": 0.4125,
+      "step": 6909
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002477221322925466,
+      "loss": 0.3549,
+      "step": 6910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024770773078176383,
+      "loss": 0.749,
+      "step": 6911
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024769332770633836,
+      "loss": 0.6268,
+      "step": 6912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002476789230665009,
+      "loss": 0.4868,
+      "step": 6913
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024766451686248205,
+      "loss": 0.0906,
+      "step": 6914
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002476501090945125,
+      "loss": 0.4793,
+      "step": 6915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024763569976282305,
+      "loss": 0.7382,
+      "step": 6916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002476212888676444,
+      "loss": 0.2756,
+      "step": 6917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002476068764092073,
+      "loss": 0.5813,
+      "step": 6918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002475924623877426,
+      "loss": 0.4752,
+      "step": 6919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024757804680348105,
+      "loss": 0.0037,
+      "step": 6920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024756362965665355,
+      "loss": 1.5612,
+      "step": 6921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002475492109474909,
+      "loss": 0.5254,
+      "step": 6922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024753479067622403,
+      "loss": 2.0874,
+      "step": 6923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002475203688430839,
+      "loss": 0.7689,
+      "step": 6924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002475059454483014,
+      "loss": 0.2021,
+      "step": 6925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002474915204921075,
+      "loss": 0.3816,
+      "step": 6926
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002474770939747333,
+      "loss": 0.0006,
+      "step": 6927
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002474626658964097,
+      "loss": 0.3795,
+      "step": 6928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024744823625736767,
+      "loss": 0.2723,
+      "step": 6929
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002474338050578385,
+      "loss": 0.805,
+      "step": 6930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024741937229805314,
+      "loss": 0.4731,
+      "step": 6931
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024740493797824277,
+      "loss": 0.3955,
+      "step": 6932
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024739050209863844,
+      "loss": 1.1523,
+      "step": 6933
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002473760646594714,
+      "loss": 0.7466,
+      "step": 6934
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002473616256609728,
+      "loss": 0.6604,
+      "step": 6935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000247347185103374,
+      "loss": 0.3311,
+      "step": 6936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000247332742986906,
+      "loss": 0.5865,
+      "step": 6937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024731829931180023,
+      "loss": 0.0853,
+      "step": 6938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000247303854078288,
+      "loss": 0.6242,
+      "step": 6939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024728940728660053,
+      "loss": 0.4506,
+      "step": 6940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024727495893696924,
+      "loss": 2.3129,
+      "step": 6941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002472605090296255,
+      "loss": 0.4236,
+      "step": 6942
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024724605756480065,
+      "loss": 1.2407,
+      "step": 6943
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002472316045427261,
+      "loss": 1.0444,
+      "step": 6944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024721714996363335,
+      "loss": 0.2849,
+      "step": 6945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024720269382775385,
+      "loss": 0.4911,
+      "step": 6946
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002471882361353191,
+      "loss": 0.0118,
+      "step": 6947
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024717377688656056,
+      "loss": 0.3378,
+      "step": 6948
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024715931608170985,
+      "loss": 0.4846,
+      "step": 6949
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024714485372099855,
+      "loss": 0.0155,
+      "step": 6950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024713038980465814,
+      "loss": 0.0929,
+      "step": 6951
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002471159243329203,
+      "loss": 2.6414,
+      "step": 6952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002471014573060168,
+      "loss": 0.0007,
+      "step": 6953
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002470869887241791,
+      "loss": 0.31,
+      "step": 6954
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000247072518587639,
+      "loss": 0.2711,
+      "step": 6955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024705804689662816,
+      "loss": 0.3718,
+      "step": 6956
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002470435736513784,
+      "loss": 0.4503,
+      "step": 6957
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002470290988521215,
+      "loss": 0.9581,
+      "step": 6958
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002470146224990891,
+      "loss": 0.0014,
+      "step": 6959
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024700014459251324,
+      "loss": 0.0195,
+      "step": 6960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024698566513262554,
+      "loss": 0.7444,
+      "step": 6961
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000246971184119658,
+      "loss": 0.6958,
+      "step": 6962
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024695670155384247,
+      "loss": 0.2458,
+      "step": 6963
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002469422174354109,
+      "loss": 0.017,
+      "step": 6964
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002469277317645952,
+      "loss": 2.4927,
+      "step": 6965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002469132445416273,
+      "loss": 0.2022,
+      "step": 6966
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002468987557667393,
+      "loss": 3.0949,
+      "step": 6967
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002468842654401631,
+      "loss": 0.5417,
+      "step": 6968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024686977356213084,
+      "loss": 0.7519,
+      "step": 6969
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024685528013287456,
+      "loss": 0.7821,
+      "step": 6970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002468407851526263,
+      "loss": 0.6378,
+      "step": 6971
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024682628862161815,
+      "loss": 0.2009,
+      "step": 6972
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024681179054008234,
+      "loss": 0.0146,
+      "step": 6973
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000246797290908251,
+      "loss": 0.5861,
+      "step": 6974
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024678278972635633,
+      "loss": 0.2873,
+      "step": 6975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002467682869946305,
+      "loss": 0.531,
+      "step": 6976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024675378271330585,
+      "loss": 0.661,
+      "step": 6977
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002467392768826145,
+      "loss": 0.5754,
+      "step": 6978
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024672476950278883,
+      "loss": 0.3853,
+      "step": 6979
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024671026057406116,
+      "loss": 0.8004,
+      "step": 6980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024669575009666383,
+      "loss": 0.5344,
+      "step": 6981
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024668123807082917,
+      "loss": 0.4411,
+      "step": 6982
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024666672449678953,
+      "loss": 0.8123,
+      "step": 6983
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002466522093747774,
+      "loss": 0.3163,
+      "step": 6984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024663769270502525,
+      "loss": 0.0006,
+      "step": 6985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002466231744877654,
+      "loss": 0.8237,
+      "step": 6986
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002466086547232305,
+      "loss": 0.6726,
+      "step": 6987
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002465941334116529,
+      "loss": 2.2121,
+      "step": 6988
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024657961055326525,
+      "loss": 0.238,
+      "step": 6989
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002465650861483001,
+      "loss": 1.1923,
+      "step": 6990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002465505601969901,
+      "loss": 0.2538,
+      "step": 6991
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002465360326995676,
+      "loss": 0.4141,
+      "step": 6992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024652150365626554,
+      "loss": 0.8267,
+      "step": 6993
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024650697306731646,
+      "loss": 0.5889,
+      "step": 6994
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000246492440932953,
+      "loss": 2.9564,
+      "step": 6995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002464779072534079,
+      "loss": 0.5922,
+      "step": 6996
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024646337202891396,
+      "loss": 4.0059,
+      "step": 6997
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024644883525970387,
+      "loss": 2.249,
+      "step": 6998
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024643429694601044,
+      "loss": 0.3839,
+      "step": 6999
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024641975708806645,
+      "loss": 0.4758,
+      "step": 7000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024640521568610477,
+      "loss": 0.0303,
+      "step": 7001
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002463906727403582,
+      "loss": 0.4622,
+      "step": 7002
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002463761282510597,
+      "loss": 0.8994,
+      "step": 7003
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024636158221844214,
+      "loss": 2.0879,
+      "step": 7004
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002463470346427385,
+      "loss": 1.2821,
+      "step": 7005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024633248552418166,
+      "loss": 0.7498,
+      "step": 7006
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002463179348630047,
+      "loss": 0.0069,
+      "step": 7007
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024630338265944045,
+      "loss": 0.7316,
+      "step": 7008
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024628882891372216,
+      "loss": 0.2566,
+      "step": 7009
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002462742736260827,
+      "loss": 0.6863,
+      "step": 7010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024625971679675533,
+      "loss": 0.4601,
+      "step": 7011
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000246245158425973,
+      "loss": 0.4068,
+      "step": 7012
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002462305985139689,
+      "loss": 0.0767,
+      "step": 7013
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024621603706097625,
+      "loss": 0.9261,
+      "step": 7014
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002462014740672281,
+      "loss": 0.6088,
+      "step": 7015
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002461869095329577,
+      "loss": 0.3076,
+      "step": 7016
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002461723434583984,
+      "loss": 2.3542,
+      "step": 7017
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024615777584378326,
+      "loss": 0.8049,
+      "step": 7018
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002461432066893457,
+      "loss": 0.9644,
+      "step": 7019
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024612863599531894,
+      "loss": 0.265,
+      "step": 7020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024611406376193635,
+      "loss": 0.5692,
+      "step": 7021
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002460994899894313,
+      "loss": 0.7085,
+      "step": 7022
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002460849146780371,
+      "loss": 0.5139,
+      "step": 7023
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002460703378279873,
+      "loss": 0.1767,
+      "step": 7024
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002460557594395151,
+      "loss": 0.414,
+      "step": 7025
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024604117951285414,
+      "loss": 0.0227,
+      "step": 7026
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002460265980482378,
+      "loss": 0.901,
+      "step": 7027
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024601201504589963,
+      "loss": 0.0146,
+      "step": 7028
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002459974305060731,
+      "loss": 0.3494,
+      "step": 7029
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024598284442899185,
+      "loss": 0.2329,
+      "step": 7030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024596825681488934,
+      "loss": 0.6865,
+      "step": 7031
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024595366766399926,
+      "loss": 0.2984,
+      "step": 7032
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002459390769765552,
+      "loss": 0.4552,
+      "step": 7033
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024592448475279076,
+      "loss": 0.8611,
+      "step": 7034
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002459098909929397,
+      "loss": 0.4576,
+      "step": 7035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002458952956972357,
+      "loss": 1.9207,
+      "step": 7036
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002458806988659124,
+      "loss": 0.1716,
+      "step": 7037
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002458661004992036,
+      "loss": 0.5852,
+      "step": 7038
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002458515005973431,
+      "loss": 0.6501,
+      "step": 7039
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024583689916056467,
+      "loss": 0.3121,
+      "step": 7040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024582229618910214,
+      "loss": 0.4411,
+      "step": 7041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002458076916831893,
+      "loss": 0.1753,
+      "step": 7042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024579308564306014,
+      "loss": 0.4314,
+      "step": 7043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024577847806894846,
+      "loss": 0.5531,
+      "step": 7044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024576386896108817,
+      "loss": 0.2841,
+      "step": 7045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024574925831971324,
+      "loss": 0.7177,
+      "step": 7046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024573464614505765,
+      "loss": 0.1352,
+      "step": 7047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024572003243735537,
+      "loss": 0.0388,
+      "step": 7048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002457054171968405,
+      "loss": 0.6362,
+      "step": 7049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024569080042374694,
+      "loss": 0.3863,
+      "step": 7050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002456761821183088,
+      "loss": 0.4866,
+      "step": 7051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002456615622807602,
+      "loss": 0.4171,
+      "step": 7052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024564694091133525,
+      "loss": 1.0318,
+      "step": 7053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002456323180102681,
+      "loss": 0.5723,
+      "step": 7054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002456176935777929,
+      "loss": 0.2184,
+      "step": 7055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002456030676141439,
+      "loss": 0.4307,
+      "step": 7056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024558844011955515,
+      "loss": 0.5292,
+      "step": 7057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000245573811094261,
+      "loss": 0.2929,
+      "step": 7058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002455591805384958,
+      "loss": 0.9302,
+      "step": 7059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024554454845249365,
+      "loss": 0.3704,
+      "step": 7060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000245529914836489,
+      "loss": 0.8311,
+      "step": 7061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002455152796907161,
+      "loss": 0.4689,
+      "step": 7062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002455006430154093,
+      "loss": 0.0034,
+      "step": 7063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002454860048108031,
+      "loss": 0.839,
+      "step": 7064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024547136507713186,
+      "loss": 0.3456,
+      "step": 7065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024545672381462994,
+      "loss": 1.0154,
+      "step": 7066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024544208102353186,
+      "loss": 0.4294,
+      "step": 7067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002454274367040721,
+      "loss": 0.0014,
+      "step": 7068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002454127908564851,
+      "loss": 0.1399,
+      "step": 7069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024539814348100553,
+      "loss": 2.9671,
+      "step": 7070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002453834945778678,
+      "loss": 0.4568,
+      "step": 7071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024536884414730657,
+      "loss": 0.7156,
+      "step": 7072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002453541921895565,
+      "loss": 0.746,
+      "step": 7073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024533953870485207,
+      "loss": 0.6926,
+      "step": 7074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024532488369342806,
+      "loss": 0.0017,
+      "step": 7075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024531022715551905,
+      "loss": 0.8131,
+      "step": 7076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002452955690913598,
+      "loss": 0.4339,
+      "step": 7077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002452809095011851,
+      "loss": 0.1032,
+      "step": 7078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024526624838522955,
+      "loss": 0.5049,
+      "step": 7079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024525158574372803,
+      "loss": 0.2118,
+      "step": 7080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002452369215769153,
+      "loss": 0.7615,
+      "step": 7081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024522225588502624,
+      "loss": 3.0391,
+      "step": 7082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024520758866829564,
+      "loss": 0.0004,
+      "step": 7083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024519291992695843,
+      "loss": 0.1521,
+      "step": 7084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024517824966124944,
+      "loss": 0.146,
+      "step": 7085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002451635778714036,
+      "loss": 0.5238,
+      "step": 7086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024514890455765593,
+      "loss": 0.7813,
+      "step": 7087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002451342297202413,
+      "loss": 0.4035,
+      "step": 7088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002451195533593948,
+      "loss": 0.0123,
+      "step": 7089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002451048754753514,
+      "loss": 0.6002,
+      "step": 7090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024509019606834613,
+      "loss": 0.0027,
+      "step": 7091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002450755151386141,
+      "loss": 0.1632,
+      "step": 7092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002450608326863904,
+      "loss": 0.4236,
+      "step": 7093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002450461487119101,
+      "loss": 0.408,
+      "step": 7094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024503146321540835,
+      "loss": 0.4134,
+      "step": 7095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002450167761971204,
+      "loss": 2.1332,
+      "step": 7096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024500208765728136,
+      "loss": 0.0417,
+      "step": 7097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024498739759612644,
+      "loss": 0.394,
+      "step": 7098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002449727060138909,
+      "loss": 0.0156,
+      "step": 7099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024495801291081004,
+      "loss": 0.2415,
+      "step": 7100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002449433182871191,
+      "loss": 1.3225,
+      "step": 7101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024492862214305343,
+      "loss": 0.5346,
+      "step": 7102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00024491392447884827,
+      "loss": 0.6947,
+      "step": 7103
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024489922529473916,
+      "loss": 1.3814,
+      "step": 7104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002448845245909613,
+      "loss": 0.2967,
+      "step": 7105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002448698223677502,
+      "loss": 0.1558,
+      "step": 7106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002448551186253413,
+      "loss": 0.0035,
+      "step": 7107
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024484041336397,
+      "loss": 0.7107,
+      "step": 7108
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002448257065838718,
+      "loss": 0.3494,
+      "step": 7109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002448109982852823,
+      "loss": 0.3274,
+      "step": 7110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002447962884684369,
+      "loss": 0.2904,
+      "step": 7111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002447815771335712,
+      "loss": 0.0231,
+      "step": 7112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024476686428092086,
+      "loss": 0.3028,
+      "step": 7113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002447521499107213,
+      "loss": 0.3031,
+      "step": 7114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024473743402320835,
+      "loss": 0.6175,
+      "step": 7115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024472271661861753,
+      "loss": 0.2198,
+      "step": 7116
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024470799769718457,
+      "loss": 0.002,
+      "step": 7117
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002446932772591452,
+      "loss": 0.9732,
+      "step": 7118
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024467855530473505,
+      "loss": 3.2504,
+      "step": 7119
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024466383183419,
+      "loss": 0.2965,
+      "step": 7120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002446491068477457,
+      "loss": 0.3536,
+      "step": 7121
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024463438034563804,
+      "loss": 0.3732,
+      "step": 7122
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002446196523281028,
+      "loss": 0.4633,
+      "step": 7123
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002446049227953758,
+      "loss": 0.0058,
+      "step": 7124
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024459019174769296,
+      "loss": 0.6168,
+      "step": 7125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024457545918529013,
+      "loss": 0.8526,
+      "step": 7126
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024456072510840334,
+      "loss": 2.2129,
+      "step": 7127
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002445459895172684,
+      "loss": 0.5467,
+      "step": 7128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002445312524121213,
+      "loss": 0.7988,
+      "step": 7129
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002445165137931981,
+      "loss": 0.2209,
+      "step": 7130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024450177366073476,
+      "loss": 0.1546,
+      "step": 7131
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002444870320149674,
+      "loss": 0.4724,
+      "step": 7132
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024447228885613197,
+      "loss": 2.3887,
+      "step": 7133
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024445754418446464,
+      "loss": 0.0134,
+      "step": 7134
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002444427980002015,
+      "loss": 0.1889,
+      "step": 7135
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024442805030357867,
+      "loss": 0.9952,
+      "step": 7136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002444133010948323,
+      "loss": 0.4584,
+      "step": 7137
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002443985503741987,
+      "loss": 0.65,
+      "step": 7138
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000244383798141914,
+      "loss": 0.6727,
+      "step": 7139
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024436904439821425,
+      "loss": 1.1005,
+      "step": 7140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000244354289143336,
+      "loss": 0.0142,
+      "step": 7141
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024433953237751544,
+      "loss": 0.5458,
+      "step": 7142
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024432477410098885,
+      "loss": 0.4165,
+      "step": 7143
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002443100143139925,
+      "loss": 0.0057,
+      "step": 7144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002442952530167629,
+      "loss": 0.6559,
+      "step": 7145
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002442804902095363,
+      "loss": 0.058,
+      "step": 7146
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002442657258925491,
+      "loss": 0.5093,
+      "step": 7147
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002442509600660378,
+      "loss": 0.0584,
+      "step": 7148
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024423619273023883,
+      "loss": 1.1182,
+      "step": 7149
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002442214238853887,
+      "loss": 0.3081,
+      "step": 7150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002442066535317238,
+      "loss": 0.5095,
+      "step": 7151
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024419188166948074,
+      "loss": 0.0356,
+      "step": 7152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002441771082988961,
+      "loss": 0.4173,
+      "step": 7153
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002441623334202064,
+      "loss": 0.8032,
+      "step": 7154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002441475570336482,
+      "loss": 0.7482,
+      "step": 7155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024413277913945823,
+      "loss": 0.0209,
+      "step": 7156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024411799973787307,
+      "loss": 0.1422,
+      "step": 7157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024410321882912937,
+      "loss": 0.0018,
+      "step": 7158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024408843641346385,
+      "loss": 0.907,
+      "step": 7159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024407365249111326,
+      "loss": 0.3552,
+      "step": 7160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024405886706231427,
+      "loss": 0.0994,
+      "step": 7161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024404408012730376,
+      "loss": 3.1876,
+      "step": 7162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024402929168631837,
+      "loss": 0.3056,
+      "step": 7163
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000244014501739595,
+      "loss": 0.4975,
+      "step": 7164
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024399971028737054,
+      "loss": 1.547,
+      "step": 7165
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024398491732988175,
+      "loss": 0.3512,
+      "step": 7166
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002439701228673655,
+      "loss": 2.506,
+      "step": 7167
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024395532690005885,
+      "loss": 0.904,
+      "step": 7168
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024394052942819863,
+      "loss": 0.524,
+      "step": 7169
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002439257304520218,
+      "loss": 0.4391,
+      "step": 7170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024391092997176535,
+      "loss": 0.4833,
+      "step": 7171
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024389612798766635,
+      "loss": 0.6354,
+      "step": 7172
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002438813244999617,
+      "loss": 1.0104,
+      "step": 7173
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024386651950888857,
+      "loss": 0.6287,
+      "step": 7174
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000243851713014684,
+      "loss": 0.4497,
+      "step": 7175
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024383690501758514,
+      "loss": 1.8516,
+      "step": 7176
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024382209551782904,
+      "loss": 0.8187,
+      "step": 7177
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024380728451565286,
+      "loss": 3.0606,
+      "step": 7178
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024379247201129385,
+      "loss": 0.0978,
+      "step": 7179
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024377765800498913,
+      "loss": 0.0093,
+      "step": 7180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024376284249697598,
+      "loss": 0.2057,
+      "step": 7181
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002437480254874916,
+      "loss": 0.0164,
+      "step": 7182
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024373320697677333,
+      "loss": 0.2939,
+      "step": 7183
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024371838696505839,
+      "loss": 0.6548,
+      "step": 7184
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024370356545258415,
+      "loss": 0.9416,
+      "step": 7185
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002436887424395879,
+      "loss": 0.8951,
+      "step": 7186
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024367391792630712,
+      "loss": 0.651,
+      "step": 7187
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024365909191297906,
+      "loss": 0.4152,
+      "step": 7188
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024364426439984124,
+      "loss": 0.4494,
+      "step": 7189
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024362943538713106,
+      "loss": 0.2932,
+      "step": 7190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024361460487508604,
+      "loss": 0.4334,
+      "step": 7191
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024359977286394358,
+      "loss": 0.0066,
+      "step": 7192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024358493935394118,
+      "loss": 2.1218,
+      "step": 7193
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024357010434531649,
+      "loss": 0.8543,
+      "step": 7194
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024355526783830696,
+      "loss": 0.9116,
+      "step": 7195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024354042983315026,
+      "loss": 0.7211,
+      "step": 7196
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024352559033008394,
+      "loss": 0.2289,
+      "step": 7197
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002435107493293457,
+      "loss": 3.0118,
+      "step": 7198
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002434959068311731,
+      "loss": 0.8221,
+      "step": 7199
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002434810628358039,
+      "loss": 0.7337,
+      "step": 7200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024346621734347573,
+      "loss": 0.4428,
+      "step": 7201
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002434513703544264,
+      "loss": 0.4898,
+      "step": 7202
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024343652186889362,
+      "loss": 0.689,
+      "step": 7203
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024342167188711516,
+      "loss": 0.7574,
+      "step": 7204
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024340682040932887,
+      "loss": 0.9228,
+      "step": 7205
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024339196743577244,
+      "loss": 0.4997,
+      "step": 7206
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002433771129666839,
+      "loss": 0.3777,
+      "step": 7207
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000243362257002301,
+      "loss": 0.6214,
+      "step": 7208
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002433473995428617,
+      "loss": 1.7728,
+      "step": 7209
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024333254058860385,
+      "loss": 0.6124,
+      "step": 7210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024331768013976547,
+      "loss": 0.0451,
+      "step": 7211
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024330281819658447,
+      "loss": 0.3859,
+      "step": 7212
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002432879547592989,
+      "loss": 0.4678,
+      "step": 7213
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002432730898281467,
+      "loss": 0.4368,
+      "step": 7214
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024325822340336595,
+      "loss": 2.2383,
+      "step": 7215
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024324335548519472,
+      "loss": 0.017,
+      "step": 7216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002432284860738711,
+      "loss": 1.2217,
+      "step": 7217
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024321361516963321,
+      "loss": 0.6574,
+      "step": 7218
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024319874277271917,
+      "loss": 0.441,
+      "step": 7219
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024318386888336712,
+      "loss": 0.0135,
+      "step": 7220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024316899350181525,
+      "loss": 0.3239,
+      "step": 7221
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024315411662830175,
+      "loss": 0.4976,
+      "step": 7222
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024313923826306492,
+      "loss": 0.4805,
+      "step": 7223
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024312435840634293,
+      "loss": 0.3744,
+      "step": 7224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024310947705837417,
+      "loss": 1.0363,
+      "step": 7225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024309459421939682,
+      "loss": 0.9122,
+      "step": 7226
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024307970988964926,
+      "loss": 1.7681,
+      "step": 7227
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024306482406936985,
+      "loss": 0.4661,
+      "step": 7228
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024304993675879695,
+      "loss": 0.3392,
+      "step": 7229
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000243035047958169,
+      "loss": 0.7955,
+      "step": 7230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002430201576677243,
+      "loss": 0.7465,
+      "step": 7231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024300526588770142,
+      "loss": 0.0262,
+      "step": 7232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002429903726183388,
+      "loss": 0.3413,
+      "step": 7233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024297547785987494,
+      "loss": 0.2462,
+      "step": 7234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024296058161254832,
+      "loss": 0.561,
+      "step": 7235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024294568387659744,
+      "loss": 0.8225,
+      "step": 7236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024293078465226103,
+      "loss": 0.5538,
+      "step": 7237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024291588393977748,
+      "loss": 0.2242,
+      "step": 7238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024290098173938553,
+      "loss": 0.7223,
+      "step": 7239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002428860780513238,
+      "loss": 0.5002,
+      "step": 7240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024287117287583095,
+      "loss": 0.5283,
+      "step": 7241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024285626621314558,
+      "loss": 0.7762,
+      "step": 7242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024284135806350654,
+      "loss": 0.3126,
+      "step": 7243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024282644842715248,
+      "loss": 0.4739,
+      "step": 7244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024281153730432208,
+      "loss": 0.5997,
+      "step": 7245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024279662469525425,
+      "loss": 0.2238,
+      "step": 7246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024278171060018776,
+      "loss": 3.5521,
+      "step": 7247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002427667950193614,
+      "loss": 0.41,
+      "step": 7248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024275187795301406,
+      "loss": 0.2566,
+      "step": 7249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024273695940138458,
+      "loss": 0.0817,
+      "step": 7250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024272203936471186,
+      "loss": 0.5226,
+      "step": 7251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024270711784323484,
+      "loss": 0.2447,
+      "step": 7252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002426921948371925,
+      "loss": 1.1643,
+      "step": 7253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024267727034682374,
+      "loss": 0.1857,
+      "step": 7254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002426623443723676,
+      "loss": 0.8883,
+      "step": 7255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024264741691406306,
+      "loss": 0.7865,
+      "step": 7256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024263248797214918,
+      "loss": 0.3053,
+      "step": 7257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024261755754686503,
+      "loss": 0.2912,
+      "step": 7258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002426026256384497,
+      "loss": 0.4194,
+      "step": 7259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024258769224714234,
+      "loss": 0.8171,
+      "step": 7260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000242572757373182,
+      "loss": 0.2824,
+      "step": 7261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024255782101680789,
+      "loss": 2.6202,
+      "step": 7262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024254288317825917,
+      "loss": 0.7708,
+      "step": 7263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002425279438577751,
+      "loss": 0.3803,
+      "step": 7264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024251300305559484,
+      "loss": 0.6344,
+      "step": 7265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002424980607719577,
+      "loss": 0.5432,
+      "step": 7266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024248311700710292,
+      "loss": 0.7063,
+      "step": 7267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002424681717612698,
+      "loss": 0.5807,
+      "step": 7268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024245322503469775,
+      "loss": 0.3055,
+      "step": 7269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000242438276827626,
+      "loss": 0.1548,
+      "step": 7270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000242423327140294,
+      "loss": 0.1508,
+      "step": 7271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002424083759729411,
+      "loss": 0.9721,
+      "step": 7272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024239342332580676,
+      "loss": 0.2206,
+      "step": 7273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002423784691991304,
+      "loss": 0.2834,
+      "step": 7274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024236351359315154,
+      "loss": 1.0512,
+      "step": 7275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002423485565081096,
+      "loss": 0.5724,
+      "step": 7276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002423335979442441,
+      "loss": 0.3442,
+      "step": 7277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002423186379017946,
+      "loss": 0.369,
+      "step": 7278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002423036763810007,
+      "loss": 0.4919,
+      "step": 7279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002422887133821019,
+      "loss": 2.7995,
+      "step": 7280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002422737489053379,
+      "loss": 0.3897,
+      "step": 7281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024225878295094835,
+      "loss": 0.0031,
+      "step": 7282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024224381551917276,
+      "loss": 0.1774,
+      "step": 7283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024222884661025094,
+      "loss": 0.4639,
+      "step": 7284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024221387622442255,
+      "loss": 0.0908,
+      "step": 7285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024219890436192736,
+      "loss": 0.7037,
+      "step": 7286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024218393102300504,
+      "loss": 0.2875,
+      "step": 7287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002421689562078954,
+      "loss": 1.4165,
+      "step": 7288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024215397991683832,
+      "loss": 0.2432,
+      "step": 7289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024213900215007354,
+      "loss": 0.7146,
+      "step": 7290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024212402290784088,
+      "loss": 0.256,
+      "step": 7291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024210904219038027,
+      "loss": 0.6162,
+      "step": 7292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002420940599979316,
+      "loss": 1.5662,
+      "step": 7293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024207907633073475,
+      "loss": 0.1989,
+      "step": 7294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002420640911890297,
+      "loss": 0.9334,
+      "step": 7295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024204910457305636,
+      "loss": 0.2403,
+      "step": 7296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002420341164830548,
+      "loss": 0.376,
+      "step": 7297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000242019126919265,
+      "loss": 1.3544,
+      "step": 7298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024200413588192698,
+      "loss": 0.1733,
+      "step": 7299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024198914337128074,
+      "loss": 0.6026,
+      "step": 7300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024197414938756652,
+      "loss": 0.3226,
+      "step": 7301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024195915393102424,
+      "loss": 0.6408,
+      "step": 7302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024194415700189416,
+      "loss": 1.2999,
+      "step": 7303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024192915860041643,
+      "loss": 1.3382,
+      "step": 7304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024191415872683116,
+      "loss": 0.7824,
+      "step": 7305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024189915738137858,
+      "loss": 0.6152,
+      "step": 7306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002418841545642989,
+      "loss": 0.4267,
+      "step": 7307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002418691502758324,
+      "loss": 0.0052,
+      "step": 7308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024185414451621934,
+      "loss": 0.6037,
+      "step": 7309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002418391372857,
+      "loss": 0.267,
+      "step": 7310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024182412858451472,
+      "loss": 0.5916,
+      "step": 7311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002418091184129038,
+      "loss": 0.9159,
+      "step": 7312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024179410677110767,
+      "loss": 0.0143,
+      "step": 7313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024177909365936666,
+      "loss": 1.0027,
+      "step": 7314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024176407907792124,
+      "loss": 0.3156,
+      "step": 7315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024174906302701176,
+      "loss": 0.3892,
+      "step": 7316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024173404550687877,
+      "loss": 0.3377,
+      "step": 7317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024171902651776272,
+      "loss": 0.2368,
+      "step": 7318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024170400605990407,
+      "loss": 0.5329,
+      "step": 7319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024168898413354345,
+      "loss": 0.3496,
+      "step": 7320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002416739607389213,
+      "loss": 0.3505,
+      "step": 7321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024165893587627826,
+      "loss": 0.0074,
+      "step": 7322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024164390954585496,
+      "loss": 0.4414,
+      "step": 7323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024162888174789196,
+      "loss": 0.6819,
+      "step": 7324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024161385248262993,
+      "loss": 0.3276,
+      "step": 7325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024159882175030952,
+      "loss": 0.002,
+      "step": 7326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024158378955117153,
+      "loss": 0.154,
+      "step": 7327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024156875588545658,
+      "loss": 0.4939,
+      "step": 7328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002415537207534054,
+      "loss": 0.3116,
+      "step": 7329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024153868415525875,
+      "loss": 0.2659,
+      "step": 7330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002415236460912575,
+      "loss": 0.7935,
+      "step": 7331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024150860656164242,
+      "loss": 0.3315,
+      "step": 7332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002414935655666543,
+      "loss": 0.0065,
+      "step": 7333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002414785231065341,
+      "loss": 0.2772,
+      "step": 7334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024146347918152262,
+      "loss": 0.5359,
+      "step": 7335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024144843379186076,
+      "loss": 1.2198,
+      "step": 7336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002414333869377895,
+      "loss": 0.4726,
+      "step": 7337
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024141833861954974,
+      "loss": 0.3655,
+      "step": 7338
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024140328883738254,
+      "loss": 0.1636,
+      "step": 7339
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024138823759152882,
+      "loss": 0.1121,
+      "step": 7340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002413731848822296,
+      "loss": 1.4652,
+      "step": 7341
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000241358130709726,
+      "loss": 0.0062,
+      "step": 7342
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024134307507425908,
+      "loss": 0.1679,
+      "step": 7343
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002413280179760698,
+      "loss": 0.0263,
+      "step": 7344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024131295941539946,
+      "loss": 0.4279,
+      "step": 7345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024129789939248905,
+      "loss": 1.2714,
+      "step": 7346
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024128283790757985,
+      "loss": 0.0046,
+      "step": 7347
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024126777496091297,
+      "loss": 0.5966,
+      "step": 7348
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024125271055272965,
+      "loss": 0.0041,
+      "step": 7349
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024123764468327117,
+      "loss": 0.2548,
+      "step": 7350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024122257735277865,
+      "loss": 0.374,
+      "step": 7351
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024120750856149355,
+      "loss": 0.4183,
+      "step": 7352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00024119243830965703,
+      "loss": 0.2615,
+      "step": 7353
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002411773665975105,
+      "loss": 0.5734,
+      "step": 7354
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024116229342529527,
+      "loss": 0.0006,
+      "step": 7355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002411472187932527,
+      "loss": 0.6792,
+      "step": 7356
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024113214270162427,
+      "loss": 0.001,
+      "step": 7357
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002411170651506513,
+      "loss": 3.3399,
+      "step": 7358
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024110198614057528,
+      "loss": 0.7808,
+      "step": 7359
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024108690567163768,
+      "loss": 0.0095,
+      "step": 7360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024107182374408004,
+      "loss": 0.0085,
+      "step": 7361
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024105674035814373,
+      "loss": 0.5612,
+      "step": 7362
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024104165551407043,
+      "loss": 1.1444,
+      "step": 7363
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024102656921210164,
+      "loss": 0.3731,
+      "step": 7364
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024101148145247897,
+      "loss": 0.7139,
+      "step": 7365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024099639223544402,
+      "loss": 0.088,
+      "step": 7366
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002409813015612384,
+      "loss": 0.2713,
+      "step": 7367
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024096620943010378,
+      "loss": 0.3565,
+      "step": 7368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024095111584228186,
+      "loss": 0.0042,
+      "step": 7369
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024093602079801427,
+      "loss": 0.6834,
+      "step": 7370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024092092429754282,
+      "loss": 1.6162,
+      "step": 7371
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002409058263411092,
+      "loss": 0.5323,
+      "step": 7372
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024089072692895527,
+      "loss": 0.5844,
+      "step": 7373
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024087562606132264,
+      "loss": 0.3905,
+      "step": 7374
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024086052373845332,
+      "loss": 0.5633,
+      "step": 7375
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024084541996058907,
+      "loss": 0.0015,
+      "step": 7376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024083031472797175,
+      "loss": 0.0205,
+      "step": 7377
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024081520804084322,
+      "loss": 0.9161,
+      "step": 7378
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024080009989944549,
+      "loss": 0.2656,
+      "step": 7379
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024078499030402042,
+      "loss": 2.2255,
+      "step": 7380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024076987925480996,
+      "loss": 0.3192,
+      "step": 7381
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002407547667520561,
+      "loss": 0.3571,
+      "step": 7382
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002407396527960009,
+      "loss": 0.672,
+      "step": 7383
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024072453738688626,
+      "loss": 1.271,
+      "step": 7384
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024070942052495438,
+      "loss": 0.4691,
+      "step": 7385
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024069430221044724,
+      "loss": 0.0644,
+      "step": 7386
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000240679182443607,
+      "loss": 0.4807,
+      "step": 7387
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024066406122467566,
+      "loss": 0.0104,
+      "step": 7388
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002406489385538955,
+      "loss": 1.1711,
+      "step": 7389
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024063381443150867,
+      "loss": 0.2766,
+      "step": 7390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024061868885775727,
+      "loss": 0.2492,
+      "step": 7391
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024060356183288362,
+      "loss": 0.2786,
+      "step": 7392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024058843335712985,
+      "loss": 0.1604,
+      "step": 7393
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024057330343073827,
+      "loss": 0.3527,
+      "step": 7394
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024055817205395123,
+      "loss": 0.3711,
+      "step": 7395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024054303922701095,
+      "loss": 0.6434,
+      "step": 7396
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024052790495015977,
+      "loss": 1.6504,
+      "step": 7397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024051276922364004,
+      "loss": 0.4158,
+      "step": 7398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024049763204769416,
+      "loss": 0.3905,
+      "step": 7399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024048249342256453,
+      "loss": 3.2118,
+      "step": 7400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024046735334849354,
+      "loss": 1.8907,
+      "step": 7401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002404522118257237,
+      "loss": 0.4611,
+      "step": 7402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024043706885449744,
+      "loss": 0.0008,
+      "step": 7403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002404219244350572,
+      "loss": 0.0677,
+      "step": 7404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024040677856764558,
+      "loss": 0.3654,
+      "step": 7405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024039163125250508,
+      "loss": 0.0056,
+      "step": 7406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024037648248987825,
+      "loss": 0.7414,
+      "step": 7407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024036133228000768,
+      "loss": 0.0488,
+      "step": 7408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000240346180623136,
+      "loss": 0.0061,
+      "step": 7409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024033102751950584,
+      "loss": 0.1206,
+      "step": 7410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024031587296935986,
+      "loss": 0.3319,
+      "step": 7411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024030071697294064,
+      "loss": 0.3178,
+      "step": 7412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024028555953049108,
+      "loss": 0.7305,
+      "step": 7413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024027040064225367,
+      "loss": 0.4635,
+      "step": 7414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024025524030847132,
+      "loss": 0.1361,
+      "step": 7415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024024007852938676,
+      "loss": 0.4068,
+      "step": 7416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024022491530524271,
+      "loss": 0.3285,
+      "step": 7417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024020975063628215,
+      "loss": 0.3464,
+      "step": 7418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024019458452274775,
+      "loss": 0.4207,
+      "step": 7419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002401794169648825,
+      "loss": 0.6688,
+      "step": 7420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002401642479629292,
+      "loss": 0.7946,
+      "step": 7421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024014907751713076,
+      "loss": 0.4222,
+      "step": 7422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024013390562773013,
+      "loss": 0.369,
+      "step": 7423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002401187322949703,
+      "loss": 1.8839,
+      "step": 7424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024010355751909423,
+      "loss": 0.3378,
+      "step": 7425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002400883813003449,
+      "loss": 0.5632,
+      "step": 7426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024007320363896535,
+      "loss": 0.0069,
+      "step": 7427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024005802453519863,
+      "loss": 0.4494,
+      "step": 7428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002400428439892878,
+      "loss": 0.9101,
+      "step": 7429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024002766200147593,
+      "loss": 0.009,
+      "step": 7430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00024001247857200622,
+      "loss": 0.2335,
+      "step": 7431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023999729370112173,
+      "loss": 0.7048,
+      "step": 7432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023998210738906565,
+      "loss": 0.0074,
+      "step": 7433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023996691963608117,
+      "loss": 2.326,
+      "step": 7434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002399517304424115,
+      "loss": 0.4084,
+      "step": 7435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023993653980829984,
+      "loss": 0.3635,
+      "step": 7436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023992134773398948,
+      "loss": 1.1045,
+      "step": 7437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002399061542197237,
+      "loss": 0.3905,
+      "step": 7438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023989095926574578,
+      "loss": 0.7959,
+      "step": 7439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023987576287229904,
+      "loss": 0.5296,
+      "step": 7440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002398605650396269,
+      "loss": 0.3348,
+      "step": 7441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023984536576797258,
+      "loss": 0.2659,
+      "step": 7442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023983016505757962,
+      "loss": 0.3323,
+      "step": 7443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023981496290869143,
+      "loss": 0.4353,
+      "step": 7444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023979975932155134,
+      "loss": 0.5294,
+      "step": 7445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023978455429640288,
+      "loss": 0.5703,
+      "step": 7446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023976934783348952,
+      "loss": 0.4562,
+      "step": 7447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023975413993305482,
+      "loss": 1.6831,
+      "step": 7448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023973893059534221,
+      "loss": 0.8889,
+      "step": 7449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023972371982059536,
+      "loss": 0.0686,
+      "step": 7450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002397085076090578,
+      "loss": 0.6512,
+      "step": 7451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023969329396097307,
+      "loss": 0.3744,
+      "step": 7452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023967807887658487,
+      "loss": 0.1919,
+      "step": 7453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023966286235613688,
+      "loss": 0.4845,
+      "step": 7454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002396476443998727,
+      "loss": 2.084,
+      "step": 7455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023963242500803597,
+      "loss": 0.195,
+      "step": 7456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002396172041808705,
+      "loss": 0.5895,
+      "step": 7457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023960198191862002,
+      "loss": 0.6425,
+      "step": 7458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023958675822152828,
+      "loss": 0.0163,
+      "step": 7459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023957153308983906,
+      "loss": 0.346,
+      "step": 7460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023955630652379613,
+      "loss": 0.495,
+      "step": 7461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023954107852364336,
+      "loss": 0.2631,
+      "step": 7462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023952584908962463,
+      "loss": 1.6757,
+      "step": 7463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023951061822198374,
+      "loss": 0.331,
+      "step": 7464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002394953859209647,
+      "loss": 0.3127,
+      "step": 7465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023948015218681133,
+      "loss": 1.7505,
+      "step": 7466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023946491701976763,
+      "loss": 3.1896,
+      "step": 7467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023944968042007752,
+      "loss": 0.2749,
+      "step": 7468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023943444238798506,
+      "loss": 0.7405,
+      "step": 7469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002394192029237342,
+      "loss": 1.0899,
+      "step": 7470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023940396202756903,
+      "loss": 0.5396,
+      "step": 7471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023938871969973356,
+      "loss": 0.0705,
+      "step": 7472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002393734759404719,
+      "loss": 0.3968,
+      "step": 7473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002393582307500282,
+      "loss": 0.2003,
+      "step": 7474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002393429841286465,
+      "loss": 0.3229,
+      "step": 7475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023932773607657099,
+      "loss": 0.2703,
+      "step": 7476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023931248659404586,
+      "loss": 1.8665,
+      "step": 7477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023929723568131528,
+      "loss": 0.3091,
+      "step": 7478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023928198333862352,
+      "loss": 0.4581,
+      "step": 7479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023926672956621476,
+      "loss": 0.4305,
+      "step": 7480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002392514743643333,
+      "loss": 0.387,
+      "step": 7481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023923621773322344,
+      "loss": 0.3487,
+      "step": 7482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023922095967312943,
+      "loss": 0.7781,
+      "step": 7483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023920570018429573,
+      "loss": 2.4124,
+      "step": 7484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023919043926696657,
+      "loss": 0.3287,
+      "step": 7485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023917517692138642,
+      "loss": 0.5436,
+      "step": 7486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023915991314779958,
+      "loss": 1.1887,
+      "step": 7487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023914464794645062,
+      "loss": 2.0439,
+      "step": 7488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023912938131758385,
+      "loss": 0.0163,
+      "step": 7489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023911411326144389,
+      "loss": 0.6463,
+      "step": 7490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023909884377827507,
+      "loss": 0.3822,
+      "step": 7491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023908357286832197,
+      "loss": 0.6765,
+      "step": 7492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023906830053182918,
+      "loss": 0.4323,
+      "step": 7493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023905302676904122,
+      "loss": 0.574,
+      "step": 7494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002390377515802027,
+      "loss": 0.6216,
+      "step": 7495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002390224749655582,
+      "loss": 1.556,
+      "step": 7496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023900719692535236,
+      "loss": 0.5541,
+      "step": 7497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023899191745982987,
+      "loss": 0.3794,
+      "step": 7498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023897663656923538,
+      "loss": 0.4356,
+      "step": 7499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023896135425381356,
+      "loss": 0.0332,
+      "step": 7500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002389460705138092,
+      "loss": 0.3133,
+      "step": 7501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023893078534946698,
+      "loss": 0.1624,
+      "step": 7502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002389154987610317,
+      "loss": 0.4666,
+      "step": 7503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023890021074874816,
+      "loss": 0.7078,
+      "step": 7504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023888492131286118,
+      "loss": 0.419,
+      "step": 7505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023886963045361556,
+      "loss": 1.1997,
+      "step": 7506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002388543381712562,
+      "loss": 1.5294,
+      "step": 7507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023883904446602797,
+      "loss": 0.432,
+      "step": 7508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023882374933817575,
+      "loss": 1.0415,
+      "step": 7509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023880845278794448,
+      "loss": 0.0646,
+      "step": 7510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023879315481557917,
+      "loss": 0.8472,
+      "step": 7511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023877785542132467,
+      "loss": 1.1573,
+      "step": 7512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023876255460542612,
+      "loss": 0.6338,
+      "step": 7513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023874725236812844,
+      "loss": 0.3815,
+      "step": 7514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023873194870967676,
+      "loss": 0.4384,
+      "step": 7515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023871664363031602,
+      "loss": 0.5096,
+      "step": 7516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002387013371302914,
+      "loss": 0.5487,
+      "step": 7517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023868602920984797,
+      "loss": 0.0247,
+      "step": 7518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023867071986923092,
+      "loss": 0.7243,
+      "step": 7519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023865540910868538,
+      "loss": 0.2763,
+      "step": 7520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023864009692845642,
+      "loss": 0.3061,
+      "step": 7521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023862478332878944,
+      "loss": 0.6115,
+      "step": 7522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002386094683099295,
+      "loss": 0.5056,
+      "step": 7523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000238594151872122,
+      "loss": 2.3577,
+      "step": 7524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023857883401561198,
+      "loss": 0.7804,
+      "step": 7525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023856351474064498,
+      "loss": 0.3971,
+      "step": 7526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023854819404746616,
+      "loss": 0.5058,
+      "step": 7527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002385328719363209,
+      "loss": 0.3223,
+      "step": 7528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002385175484074546,
+      "loss": 1.1656,
+      "step": 7529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002385022234611126,
+      "loss": 0.7911,
+      "step": 7530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023848689709754032,
+      "loss": 1.2368,
+      "step": 7531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023847156931698312,
+      "loss": 0.3622,
+      "step": 7532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002384562401196866,
+      "loss": 0.7032,
+      "step": 7533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002384409095058961,
+      "loss": 0.9037,
+      "step": 7534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002384255774758572,
+      "loss": 0.6926,
+      "step": 7535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023841024402981533,
+      "loss": 0.6494,
+      "step": 7536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002383949091680161,
+      "loss": 0.8902,
+      "step": 7537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002383795728907051,
+      "loss": 0.0886,
+      "step": 7538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023836423519812787,
+      "loss": 1.2459,
+      "step": 7539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023834889609053,
+      "loss": 0.0588,
+      "step": 7540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023833355556815718,
+      "loss": 0.219,
+      "step": 7541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023831821363125505,
+      "loss": 0.0211,
+      "step": 7542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023830287028006922,
+      "loss": 0.931,
+      "step": 7543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023828752551484546,
+      "loss": 0.3411,
+      "step": 7544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002382721793358295,
+      "loss": 0.7965,
+      "step": 7545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002382568317432671,
+      "loss": 1.6321,
+      "step": 7546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023824148273740394,
+      "loss": 2.7715,
+      "step": 7547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023822613231848589,
+      "loss": 0.4543,
+      "step": 7548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023821078048675876,
+      "loss": 0.3887,
+      "step": 7549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023819542724246834,
+      "loss": 0.3686,
+      "step": 7550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023818007258586053,
+      "loss": 0.3241,
+      "step": 7551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023816471651718114,
+      "loss": 0.1681,
+      "step": 7552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023814935903667618,
+      "loss": 0.4087,
+      "step": 7553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023813400014459157,
+      "loss": 0.0851,
+      "step": 7554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023811863984117315,
+      "loss": 1.0314,
+      "step": 7555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000238103278126667,
+      "loss": 0.3619,
+      "step": 7556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023808791500131904,
+      "loss": 0.6103,
+      "step": 7557
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023807255046537536,
+      "loss": 0.7349,
+      "step": 7558
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023805718451908194,
+      "loss": 0.4368,
+      "step": 7559
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023804181716268486,
+      "loss": 2.4767,
+      "step": 7560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023802644839643024,
+      "loss": 2.5255,
+      "step": 7561
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002380110782205641,
+      "loss": 0.7602,
+      "step": 7562
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002379957066353327,
+      "loss": 0.0423,
+      "step": 7563
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023798033364098208,
+      "loss": 0.9575,
+      "step": 7564
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023796495923775847,
+      "loss": 0.4476,
+      "step": 7565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023794958342590806,
+      "loss": 0.7078,
+      "step": 7566
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023793420620567705,
+      "loss": 0.3598,
+      "step": 7567
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002379188275773117,
+      "loss": 0.0142,
+      "step": 7568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023790344754105828,
+      "loss": 0.3979,
+      "step": 7569
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023788806609716304,
+      "loss": 0.1437,
+      "step": 7570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002378726832458724,
+      "loss": 0.0501,
+      "step": 7571
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023785729898743256,
+      "loss": 0.5597,
+      "step": 7572
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023784191332208992,
+      "loss": 0.7402,
+      "step": 7573
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002378265262500909,
+      "loss": 0.2971,
+      "step": 7574
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023781113777168186,
+      "loss": 0.6888,
+      "step": 7575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023779574788710924,
+      "loss": 1.9589,
+      "step": 7576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002377803565966195,
+      "loss": 0.468,
+      "step": 7577
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023776496390045905,
+      "loss": 0.0136,
+      "step": 7578
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023774956979887444,
+      "loss": 0.7419,
+      "step": 7579
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023773417429211218,
+      "loss": 0.6816,
+      "step": 7580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023771877738041873,
+      "loss": 0.4938,
+      "step": 7581
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023770337906404076,
+      "loss": 0.6057,
+      "step": 7582
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002376879793432248,
+      "loss": 0.3293,
+      "step": 7583
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023767257821821744,
+      "loss": 0.314,
+      "step": 7584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002376571756892653,
+      "loss": 0.5667,
+      "step": 7585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023764177175661507,
+      "loss": 0.6667,
+      "step": 7586
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023762636642051343,
+      "loss": 0.9569,
+      "step": 7587
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023761095968120695,
+      "loss": 0.331,
+      "step": 7588
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023759555153894248,
+      "loss": 0.461,
+      "step": 7589
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023758014199396675,
+      "loss": 0.4406,
+      "step": 7590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023756473104652646,
+      "loss": 0.2052,
+      "step": 7591
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023754931869686841,
+      "loss": 0.328,
+      "step": 7592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023753390494523943,
+      "loss": 0.0217,
+      "step": 7593
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023751848979188631,
+      "loss": 0.014,
+      "step": 7594
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023750307323705595,
+      "loss": 0.3466,
+      "step": 7595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023748765528099514,
+      "loss": 0.0109,
+      "step": 7596
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023747223592395089,
+      "loss": 0.0555,
+      "step": 7597
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002374568151661701,
+      "loss": 1.6433,
+      "step": 7598
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023744139300789957,
+      "loss": 0.0817,
+      "step": 7599
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023742596944938637,
+      "loss": 0.7099,
+      "step": 7600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023741054449087753,
+      "loss": 0.6735,
+      "step": 7601
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00023739511813262,
+      "loss": 0.6141,
+      "step": 7602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002373796903748608,
+      "loss": 0.4585,
+      "step": 7603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000237364261217847,
+      "loss": 0.2934,
+      "step": 7604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023734883066182567,
+      "loss": 0.6636,
+      "step": 7605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002373333987070439,
+      "loss": 1.6048,
+      "step": 7606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023731796535374883,
+      "loss": 0.1613,
+      "step": 7607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023730253060218757,
+      "loss": 0.1716,
+      "step": 7608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023728709445260734,
+      "loss": 0.7061,
+      "step": 7609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002372716569052553,
+      "loss": 0.4302,
+      "step": 7610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002372562179603786,
+      "loss": 0.4265,
+      "step": 7611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002372407776182245,
+      "loss": 0.8683,
+      "step": 7612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023722533587904038,
+      "loss": 0.2322,
+      "step": 7613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023720989274307337,
+      "loss": 0.5941,
+      "step": 7614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002371944482105708,
+      "loss": 0.0499,
+      "step": 7615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023717900228177995,
+      "loss": 1.1117,
+      "step": 7616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002371635549569483,
+      "loss": 0.3098,
+      "step": 7617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002371481062363231,
+      "loss": 0.4054,
+      "step": 7618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023713265612015173,
+      "loss": 1.5695,
+      "step": 7619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023711720460868168,
+      "loss": 0.0198,
+      "step": 7620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023710175170216038,
+      "loss": 0.0683,
+      "step": 7621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002370862974008352,
+      "loss": 0.4361,
+      "step": 7622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023707084170495365,
+      "loss": 0.5569,
+      "step": 7623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023705538461476328,
+      "loss": 2.3641,
+      "step": 7624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002370399261305116,
+      "loss": 0.0051,
+      "step": 7625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023702446625244606,
+      "loss": 0.5533,
+      "step": 7626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023700900498081434,
+      "loss": 0.5173,
+      "step": 7627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000236993542315864,
+      "loss": 0.6658,
+      "step": 7628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023697807825784266,
+      "loss": 0.6016,
+      "step": 7629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023696261280699787,
+      "loss": 0.4612,
+      "step": 7630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023694714596357735,
+      "loss": 1.8501,
+      "step": 7631
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023693167772782878,
+      "loss": 0.2923,
+      "step": 7632
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023691620809999993,
+      "loss": 0.6055,
+      "step": 7633
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023691620809999993,
+      "loss": 1.0909,
+      "step": 7634
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023690073708033837,
+      "loss": 0.3101,
+      "step": 7635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023688526466909195,
+      "loss": 1.785,
+      "step": 7636
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023686979086650842,
+      "loss": 0.5883,
+      "step": 7637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023685431567283557,
+      "loss": 0.1128,
+      "step": 7638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002368388390883212,
+      "loss": 0.7366,
+      "step": 7639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023682336111321314,
+      "loss": 1.0468,
+      "step": 7640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023680788174775926,
+      "loss": 0.2956,
+      "step": 7641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023679240099220742,
+      "loss": 0.3053,
+      "step": 7642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023677691884680557,
+      "loss": 0.9651,
+      "step": 7643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023676143531180157,
+      "loss": 0.4367,
+      "step": 7644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002367459503874434,
+      "loss": 1.0155,
+      "step": 7645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000236730464073979,
+      "loss": 0.2126,
+      "step": 7646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023671497637165644,
+      "loss": 1.2277,
+      "step": 7647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023669948728072363,
+      "loss": 0.456,
+      "step": 7648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023668399680142865,
+      "loss": 0.4807,
+      "step": 7649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023666850493401955,
+      "loss": 0.0134,
+      "step": 7650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023665301167874442,
+      "loss": 0.2415,
+      "step": 7651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023663751703585138,
+      "loss": 0.2048,
+      "step": 7652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023662202100558852,
+      "loss": 0.3441,
+      "step": 7653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000236606523588204,
+      "loss": 0.2812,
+      "step": 7654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023659102478394594,
+      "loss": 0.2055,
+      "step": 7655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023657552459306264,
+      "loss": 0.4024,
+      "step": 7656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023656002301580218,
+      "loss": 0.3531,
+      "step": 7657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023654452005241285,
+      "loss": 0.3804,
+      "step": 7658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023652901570314302,
+      "loss": 0.0663,
+      "step": 7659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002365135099682408,
+      "loss": 0.1609,
+      "step": 7660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023649800284795454,
+      "loss": 0.897,
+      "step": 7661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002364824943425326,
+      "loss": 0.4787,
+      "step": 7662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002364669844522233,
+      "loss": 0.2247,
+      "step": 7663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023645147317727502,
+      "loss": 0.0204,
+      "step": 7664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023643596051793614,
+      "loss": 0.0708,
+      "step": 7665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023642044647445505,
+      "loss": 0.0082,
+      "step": 7666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023640493104708026,
+      "loss": 0.1325,
+      "step": 7667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023638941423606013,
+      "loss": 0.5926,
+      "step": 7668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002363738960416432,
+      "loss": 0.0382,
+      "step": 7669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023635837646407797,
+      "loss": 4.9678,
+      "step": 7670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023634285550361292,
+      "loss": 0.0136,
+      "step": 7671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023632733316049663,
+      "loss": 0.2609,
+      "step": 7672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023631180943497766,
+      "loss": 0.2398,
+      "step": 7673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002362962843273046,
+      "loss": 0.6406,
+      "step": 7674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023628075783772604,
+      "loss": 0.3611,
+      "step": 7675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002362652299664907,
+      "loss": 0.5287,
+      "step": 7676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002362497007138471,
+      "loss": 0.4862,
+      "step": 7677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023623417008004398,
+      "loss": 0.3209,
+      "step": 7678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002362186380653301,
+      "loss": 0.3434,
+      "step": 7679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023620310466995406,
+      "loss": 0.7303,
+      "step": 7680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023618756989416472,
+      "loss": 0.2127,
+      "step": 7681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023617203373821078,
+      "loss": 0.2098,
+      "step": 7682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023615649620234108,
+      "loss": 0.3163,
+      "step": 7683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023614095728680435,
+      "loss": 0.3524,
+      "step": 7684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002361254169918495,
+      "loss": 0.4196,
+      "step": 7685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002361098753177253,
+      "loss": 0.5007,
+      "step": 7686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023609433226468076,
+      "loss": 0.2489,
+      "step": 7687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002360787878329646,
+      "loss": 0.7034,
+      "step": 7688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023606324202282597,
+      "loss": 1.2387,
+      "step": 7689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023604769483451357,
+      "loss": 0.5309,
+      "step": 7690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023603214626827652,
+      "loss": 0.7718,
+      "step": 7691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023601659632436375,
+      "loss": 0.3333,
+      "step": 7692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023600104500302432,
+      "loss": 0.65,
+      "step": 7693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023598549230450716,
+      "loss": 0.4506,
+      "step": 7694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002359699382290614,
+      "loss": 0.9338,
+      "step": 7695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023595438277693615,
+      "loss": 0.4285,
+      "step": 7696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023593882594838045,
+      "loss": 0.0068,
+      "step": 7697
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023592326774364342,
+      "loss": 0.3989,
+      "step": 7698
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002359077081629742,
+      "loss": 0.598,
+      "step": 7699
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023589214720662194,
+      "loss": 0.6105,
+      "step": 7700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002358765848748359,
+      "loss": 0.2842,
+      "step": 7701
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023586102116786517,
+      "loss": 1.6985,
+      "step": 7702
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023584545608595908,
+      "loss": 0.3299,
+      "step": 7703
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002358298896293669,
+      "loss": 0.4204,
+      "step": 7704
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002358143217983378,
+      "loss": 0.8234,
+      "step": 7705
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023579875259312113,
+      "loss": 0.4153,
+      "step": 7706
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023578318201396617,
+      "loss": 0.3945,
+      "step": 7707
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023576761006112231,
+      "loss": 1.1338,
+      "step": 7708
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023575203673483895,
+      "loss": 0.2372,
+      "step": 7709
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023573646203536532,
+      "loss": 1.065,
+      "step": 7710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000235720885962951,
+      "loss": 0.4438,
+      "step": 7711
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023570530851784534,
+      "loss": 0.6774,
+      "step": 7712
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023568972970029777,
+      "loss": 0.2739,
+      "step": 7713
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023567414951055778,
+      "loss": 0.6877,
+      "step": 7714
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023565856794887486,
+      "loss": 0.5252,
+      "step": 7715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023564298501549852,
+      "loss": 0.7533,
+      "step": 7716
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023562740071067832,
+      "loss": 0.5141,
+      "step": 7717
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023561181503466377,
+      "loss": 0.3394,
+      "step": 7718
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023559622798770456,
+      "loss": 0.5429,
+      "step": 7719
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002355806395700502,
+      "loss": 0.8061,
+      "step": 7720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023556504978195032,
+      "loss": 0.496,
+      "step": 7721
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023554945862365456,
+      "loss": 0.3299,
+      "step": 7722
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023553386609541263,
+      "loss": 0.2558,
+      "step": 7723
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023551827219747424,
+      "loss": 0.7946,
+      "step": 7724
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000235502676930089,
+      "loss": 0.3981,
+      "step": 7725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023548708029350678,
+      "loss": 0.0031,
+      "step": 7726
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002354714822879772,
+      "loss": 0.2986,
+      "step": 7727
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023545588291375017,
+      "loss": 0.7299,
+      "step": 7728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023544028217107536,
+      "loss": 0.4322,
+      "step": 7729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023542468006020268,
+      "loss": 0.7465,
+      "step": 7730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000235409076581382,
+      "loss": 0.3179,
+      "step": 7731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002353934717348631,
+      "loss": 0.3525,
+      "step": 7732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023537786552089591,
+      "loss": 0.837,
+      "step": 7733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023536225793973037,
+      "loss": 0.9162,
+      "step": 7734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023534664899161633,
+      "loss": 0.6503,
+      "step": 7735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002353310386768038,
+      "loss": 0.4194,
+      "step": 7736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023531542699554276,
+      "loss": 0.8075,
+      "step": 7737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023529981394808323,
+      "loss": 0.2568,
+      "step": 7738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023528419953467515,
+      "loss": 0.4549,
+      "step": 7739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023526858375556866,
+      "loss": 0.1999,
+      "step": 7740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002352529666110137,
+      "loss": 0.0053,
+      "step": 7741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023523734810126045,
+      "loss": 0.0081,
+      "step": 7742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023522172822655905,
+      "loss": 0.9512,
+      "step": 7743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023520610698715949,
+      "loss": 0.4819,
+      "step": 7744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023519048438331206,
+      "loss": 0.0367,
+      "step": 7745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023517486041526686,
+      "loss": 0.3327,
+      "step": 7746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002351592350832741,
+      "loss": 0.8725,
+      "step": 7747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023514360838758399,
+      "loss": 0.2306,
+      "step": 7748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002351279803284468,
+      "loss": 0.441,
+      "step": 7749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002351123509061127,
+      "loss": 0.6864,
+      "step": 7750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023509672012083215,
+      "loss": 0.375,
+      "step": 7751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002350810879728552,
+      "loss": 1.8397,
+      "step": 7752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023506545446243238,
+      "loss": 0.3072,
+      "step": 7753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023504981958981402,
+      "loss": 0.3751,
+      "step": 7754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023503418335525043,
+      "loss": 0.2204,
+      "step": 7755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023501854575899193,
+      "loss": 0.9357,
+      "step": 7756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023500290680128908,
+      "loss": 0.524,
+      "step": 7757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002349872664823922,
+      "loss": 0.2864,
+      "step": 7758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023497162480255184,
+      "loss": 0.7406,
+      "step": 7759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023495598176201843,
+      "loss": 0.5459,
+      "step": 7760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023494033736104247,
+      "loss": 2.3047,
+      "step": 7761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023492469159987447,
+      "loss": 0.7821,
+      "step": 7762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023490904447876501,
+      "loss": 0.3717,
+      "step": 7763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023489339599796463,
+      "loss": 0.0032,
+      "step": 7764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023487774615772389,
+      "loss": 0.3187,
+      "step": 7765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023486209495829347,
+      "loss": 0.2391,
+      "step": 7766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023484644239992394,
+      "loss": 0.384,
+      "step": 7767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023483078848286597,
+      "loss": 0.0272,
+      "step": 7768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023481513320737024,
+      "loss": 0.3317,
+      "step": 7769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023479947657368747,
+      "loss": 0.4565,
+      "step": 7770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023478381858206828,
+      "loss": 0.3565,
+      "step": 7771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023476815923276353,
+      "loss": 0.3743,
+      "step": 7772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023475249852602392,
+      "loss": 0.2033,
+      "step": 7773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023473683646210022,
+      "loss": 1.0538,
+      "step": 7774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023472117304124324,
+      "loss": 0.2401,
+      "step": 7775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023470550826370387,
+      "loss": 0.7846,
+      "step": 7776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023468984212973292,
+      "loss": 0.7935,
+      "step": 7777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002346741746395812,
+      "loss": 0.0077,
+      "step": 7778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002346585057934997,
+      "loss": 0.6067,
+      "step": 7779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023464283559173927,
+      "loss": 0.5806,
+      "step": 7780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023462716403455084,
+      "loss": 0.0229,
+      "step": 7781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023461149112218537,
+      "loss": 0.477,
+      "step": 7782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023459581685489388,
+      "loss": 0.6163,
+      "step": 7783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023458014123292734,
+      "loss": 0.2025,
+      "step": 7784
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023456446425653685,
+      "loss": 0.9318,
+      "step": 7785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023454878592597326,
+      "loss": 0.0023,
+      "step": 7786
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023453310624148781,
+      "loss": 0.6564,
+      "step": 7787
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023451742520333153,
+      "loss": 1.4454,
+      "step": 7788
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023450174281175552,
+      "loss": 0.2335,
+      "step": 7789
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002344860590670109,
+      "loss": 0.8064,
+      "step": 7790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002344703739693488,
+      "loss": 0.5763,
+      "step": 7791
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023445468751902054,
+      "loss": 0.0031,
+      "step": 7792
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023443899971627716,
+      "loss": 0.3259,
+      "step": 7793
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002344233105613699,
+      "loss": 0.555,
+      "step": 7794
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023440762005455008,
+      "loss": 0.3423,
+      "step": 7795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002343919281960689,
+      "loss": 0.5474,
+      "step": 7796
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023437623498617758,
+      "loss": 0.2077,
+      "step": 7797
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002343605404251275,
+      "loss": 0.1996,
+      "step": 7798
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023434484451317,
+      "loss": 2.3302,
+      "step": 7799
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002343291472505564,
+      "loss": 0.5741,
+      "step": 7800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023431344863753803,
+      "loss": 0.5579,
+      "step": 7801
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023429774867436638,
+      "loss": 0.5672,
+      "step": 7802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002342820473612928,
+      "loss": 0.4267,
+      "step": 7803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023426634469856866,
+      "loss": 0.3916,
+      "step": 7804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023425064068644547,
+      "loss": 0.6572,
+      "step": 7805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023423493532517475,
+      "loss": 0.4014,
+      "step": 7806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023421922861500795,
+      "loss": 0.2069,
+      "step": 7807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002342035205561966,
+      "loss": 0.3528,
+      "step": 7808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023418781114899225,
+      "loss": 0.3909,
+      "step": 7809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023417210039364643,
+      "loss": 0.2846,
+      "step": 7810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023415638829041077,
+      "loss": 0.3549,
+      "step": 7811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023414067483953682,
+      "loss": 0.0064,
+      "step": 7812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023412496004127626,
+      "loss": 0.3028,
+      "step": 7813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023410924389588076,
+      "loss": 0.0031,
+      "step": 7814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002340935264036019,
+      "loss": 0.4889,
+      "step": 7815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023407780756469143,
+      "loss": 0.7892,
+      "step": 7816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023406208737940105,
+      "loss": 0.7342,
+      "step": 7817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023404636584798254,
+      "loss": 0.0017,
+      "step": 7818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002340306429706876,
+      "loss": 0.4654,
+      "step": 7819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023401491874776802,
+      "loss": 0.7408,
+      "step": 7820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023399919317947557,
+      "loss": 0.4494,
+      "step": 7821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023398346626606214,
+      "loss": 0.3647,
+      "step": 7822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023396773800777953,
+      "loss": 0.41,
+      "step": 7823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002339520084048796,
+      "loss": 0.238,
+      "step": 7824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023393627745761425,
+      "loss": 2.6368,
+      "step": 7825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002339205451662354,
+      "loss": 0.7405,
+      "step": 7826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023390481153099497,
+      "loss": 0.8017,
+      "step": 7827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023388907655214492,
+      "loss": 0.5932,
+      "step": 7828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002338733402299372,
+      "loss": 1.7862,
+      "step": 7829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023385760256462382,
+      "loss": 0.5356,
+      "step": 7830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023384186355645674,
+      "loss": 0.0054,
+      "step": 7831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002338261232056881,
+      "loss": 0.4376,
+      "step": 7832
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023381038151256987,
+      "loss": 0.0043,
+      "step": 7833
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337946384773542,
+      "loss": 0.4203,
+      "step": 7834
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337788941002931,
+      "loss": 0.3647,
+      "step": 7835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023376314838163877,
+      "loss": 1.5787,
+      "step": 7836
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337474013216433,
+      "loss": 0.8582,
+      "step": 7837
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023373165292055897,
+      "loss": 0.0075,
+      "step": 7838
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002337159031786378,
+      "loss": 0.3231,
+      "step": 7839
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023370015209613212,
+      "loss": 0.6742,
+      "step": 7840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023368439967329412,
+      "loss": 0.0137,
+      "step": 7841
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000233668645910376,
+      "loss": 0.7334,
+      "step": 7842
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023365289080763013,
+      "loss": 0.2147,
+      "step": 7843
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023363713436530876,
+      "loss": 2.7736,
+      "step": 7844
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023362137658366423,
+      "loss": 0.2986,
+      "step": 7845
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002336056174629488,
+      "loss": 0.4038,
+      "step": 7846
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002335898570034149,
+      "loss": 0.6761,
+      "step": 7847
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002335740952053149,
+      "loss": 0.317,
+      "step": 7848
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002335583320689012,
+      "loss": 0.7167,
+      "step": 7849
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023354256759442618,
+      "loss": 0.3028,
+      "step": 7850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00023352680178214235,
+      "loss": 0.6488,
+      "step": 7851
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023351103463230215,
+      "loss": 0.0349,
+      "step": 7852
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023349526614515806,
+      "loss": 0.5471,
+      "step": 7853
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023347949632096258,
+      "loss": 0.3552,
+      "step": 7854
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002334637251599682,
+      "loss": 0.4068,
+      "step": 7855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002334479526624276,
+      "loss": 0.6091,
+      "step": 7856
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023343217882859323,
+      "loss": 0.4604,
+      "step": 7857
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023341640365871777,
+      "loss": 0.4952,
+      "step": 7858
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023340062715305378,
+      "loss": 0.5021,
+      "step": 7859
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023338484931185392,
+      "loss": 0.1859,
+      "step": 7860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023336907013537083,
+      "loss": 0.2335,
+      "step": 7861
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023335328962385718,
+      "loss": 0.0788,
+      "step": 7862
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023333750777756573,
+      "loss": 0.0329,
+      "step": 7863
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023332172459674918,
+      "loss": 0.0201,
+      "step": 7864
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023330594008166023,
+      "loss": 0.3086,
+      "step": 7865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023330594008166023,
+      "loss": 0.7881,
+      "step": 7866
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023329015423255165,
+      "loss": 0.9456,
+      "step": 7867
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023327436704967627,
+      "loss": 0.0063,
+      "step": 7868
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023325857853328686,
+      "loss": 0.5269,
+      "step": 7869
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002332427886836363,
+      "loss": 0.3371,
+      "step": 7870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023322699750097739,
+      "loss": 3.0603,
+      "step": 7871
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023321120498556307,
+      "loss": 0.2778,
+      "step": 7872
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023319541113764615,
+      "loss": 0.8711,
+      "step": 7873
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023317961595747957,
+      "loss": 0.2694,
+      "step": 7874
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002331638194453163,
+      "loss": 0.3206,
+      "step": 7875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023314802160140922,
+      "loss": 0.8477,
+      "step": 7876
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023313222242601138,
+      "loss": 0.2279,
+      "step": 7877
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023311642191937583,
+      "loss": 0.1992,
+      "step": 7878
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023310062008175547,
+      "loss": 0.5106,
+      "step": 7879
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023308481691340337,
+      "loss": 0.3873,
+      "step": 7880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023306901241457265,
+      "loss": 0.8626,
+      "step": 7881
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023305320658551637,
+      "loss": 0.4442,
+      "step": 7882
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023303739942648764,
+      "loss": 0.5291,
+      "step": 7883
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023302159093773958,
+      "loss": 0.6908,
+      "step": 7884
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023300578111952529,
+      "loss": 1.0708,
+      "step": 7885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232989969972098,
+      "loss": 1.8012,
+      "step": 7886
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023297415749571094,
+      "loss": 0.037,
+      "step": 7887
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023295834369061727,
+      "loss": 0.32,
+      "step": 7888
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002329425285570702,
+      "loss": 0.0787,
+      "step": 7889
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023292671209532303,
+      "loss": 0.0176,
+      "step": 7890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232910894305629,
+      "loss": 0.3443,
+      "step": 7891
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023289507518824146,
+      "loss": 0.163,
+      "step": 7892
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023287925474341366,
+      "loss": 0.4588,
+      "step": 7893
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000232863432971399,
+      "loss": 0.2093,
+      "step": 7894
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023284760987245083,
+      "loss": 0.3915,
+      "step": 7895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002328317854468225,
+      "loss": 0.0116,
+      "step": 7896
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023281595969476746,
+      "loss": 0.9757,
+      "step": 7897
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023280013261653914,
+      "loss": 0.2825,
+      "step": 7898
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023278430421239097,
+      "loss": 2.2528,
+      "step": 7899
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023276847448257634,
+      "loss": 0.0307,
+      "step": 7900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023275264342734883,
+      "loss": 0.6015,
+      "step": 7901
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023273681104696195,
+      "loss": 0.8403,
+      "step": 7902
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023272097734166922,
+      "loss": 0.2333,
+      "step": 7903
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023270514231172416,
+      "loss": 0.3299,
+      "step": 7904
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326893059573804,
+      "loss": 0.9014,
+      "step": 7905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326734682788915,
+      "loss": 1.134,
+      "step": 7906
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023265762927651105,
+      "loss": 0.3536,
+      "step": 7907
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326417889504927,
+      "loss": 0.6289,
+      "step": 7908
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023262594730109014,
+      "loss": 0.6983,
+      "step": 7909
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002326101043285571,
+      "loss": 0.4723,
+      "step": 7910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023259426003314717,
+      "loss": 0.5532,
+      "step": 7911
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023257841441511412,
+      "loss": 0.006,
+      "step": 7912
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002325625674747117,
+      "loss": 0.6182,
+      "step": 7913
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002325467192121937,
+      "loss": 1.6756,
+      "step": 7914
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023253086962781385,
+      "loss": 0.2043,
+      "step": 7915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023251501872182602,
+      "loss": 0.2337,
+      "step": 7916
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023249916649448394,
+      "loss": 0.4206,
+      "step": 7917
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023248331294604155,
+      "loss": 0.0133,
+      "step": 7918
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002324674580767527,
+      "loss": 0.0019,
+      "step": 7919
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023245160188687126,
+      "loss": 0.9314,
+      "step": 7920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023243574437665119,
+      "loss": 0.7386,
+      "step": 7921
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023241988554634641,
+      "loss": 0.2608,
+      "step": 7922
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002324040253962108,
+      "loss": 0.4291,
+      "step": 7923
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002323881639264984,
+      "loss": 0.0057,
+      "step": 7924
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023237230113746322,
+      "loss": 1.2833,
+      "step": 7925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023235643702935923,
+      "loss": 0.1263,
+      "step": 7926
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023234057160244055,
+      "loss": 0.0053,
+      "step": 7927
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002323247048569612,
+      "loss": 0.0469,
+      "step": 7928
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023230883679317522,
+      "loss": 0.3204,
+      "step": 7929
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023229296741133674,
+      "loss": 0.5613,
+      "step": 7930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002322770967116999,
+      "loss": 0.0039,
+      "step": 7931
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023226122469451884,
+      "loss": 0.9054,
+      "step": 7932
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023224535136004773,
+      "loss": 1.5791,
+      "step": 7933
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023222947670854075,
+      "loss": 0.4615,
+      "step": 7934
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023221360074025208,
+      "loss": 1.7045,
+      "step": 7935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023219772345543603,
+      "loss": 0.9077,
+      "step": 7936
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023218184485434678,
+      "loss": 1.1407,
+      "step": 7937
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002321659649372386,
+      "loss": 0.0076,
+      "step": 7938
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023215008370436582,
+      "loss": 0.63,
+      "step": 7939
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023213420115598276,
+      "loss": 0.0075,
+      "step": 7940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002321183172923437,
+      "loss": 0.5192,
+      "step": 7941
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023210243211370304,
+      "loss": 2.6271,
+      "step": 7942
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023208654562031518,
+      "loss": 0.0043,
+      "step": 7943
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023207065781243445,
+      "loss": 0.4213,
+      "step": 7944
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002320547686903153,
+      "loss": 0.0381,
+      "step": 7945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023203887825421222,
+      "loss": 0.2594,
+      "step": 7946
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002320229865043796,
+      "loss": 1.3711,
+      "step": 7947
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023200709344107194,
+      "loss": 1.2018,
+      "step": 7948
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023199119906454376,
+      "loss": 0.4849,
+      "step": 7949
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002319753033750496,
+      "loss": 0.1663,
+      "step": 7950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000231959406372844,
+      "loss": 3.2451,
+      "step": 7951
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023194350805818146,
+      "loss": 0.3193,
+      "step": 7952
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023192760843131663,
+      "loss": 0.0105,
+      "step": 7953
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002319117074925041,
+      "loss": 0.3962,
+      "step": 7954
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023189580524199854,
+      "loss": 0.0375,
+      "step": 7955
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023187990168005456,
+      "loss": 0.2834,
+      "step": 7956
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002318639968069268,
+      "loss": 0.2779,
+      "step": 7957
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023184809062287001,
+      "loss": 1.668,
+      "step": 7958
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002318321831281389,
+      "loss": 0.179,
+      "step": 7959
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023181627432298818,
+      "loss": 0.6479,
+      "step": 7960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023180036420767264,
+      "loss": 0.392,
+      "step": 7961
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023178445278244702,
+      "loss": 0.0061,
+      "step": 7962
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002317685400475661,
+      "loss": 0.5788,
+      "step": 7963
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023175262600328475,
+      "loss": 0.0128,
+      "step": 7964
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023173671064985783,
+      "loss": 1.6807,
+      "step": 7965
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023172079398754011,
+      "loss": 0.2647,
+      "step": 7966
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023170487601658653,
+      "loss": 0.1967,
+      "step": 7967
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023168895673725197,
+      "loss": 1.0022,
+      "step": 7968
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002316730361497914,
+      "loss": 0.254,
+      "step": 7969
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023165711425445973,
+      "loss": 0.0125,
+      "step": 7970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023164119105151186,
+      "loss": 1.3149,
+      "step": 7971
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023162526654120288,
+      "loss": 0.0155,
+      "step": 7972
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023160934072378775,
+      "loss": 0.0935,
+      "step": 7973
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023159341359952156,
+      "loss": 0.3731,
+      "step": 7974
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023157748516865924,
+      "loss": 0.2497,
+      "step": 7975
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023156155543145596,
+      "loss": 0.466,
+      "step": 7976
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002315456243881668,
+      "loss": 0.3005,
+      "step": 7977
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002315296920390468,
+      "loss": 0.4746,
+      "step": 7978
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023151375838435114,
+      "loss": 0.1716,
+      "step": 7979
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000231497823424335,
+      "loss": 0.3821,
+      "step": 7980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023148188715925352,
+      "loss": 0.0972,
+      "step": 7981
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002314659495893619,
+      "loss": 0.6158,
+      "step": 7982
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023145001071491535,
+      "loss": 0.3769,
+      "step": 7983
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023143407053616914,
+      "loss": 0.5215,
+      "step": 7984
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002314181290533785,
+      "loss": 0.5605,
+      "step": 7985
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023140218626679874,
+      "loss": 0.0029,
+      "step": 7986
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002313862421766851,
+      "loss": 1.249,
+      "step": 7987
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023137029678329296,
+      "loss": 0.4497,
+      "step": 7988
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023135435008687759,
+      "loss": 0.4881,
+      "step": 7989
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023133840208769445,
+      "loss": 0.5678,
+      "step": 7990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023132245278599886,
+      "loss": 0.3002,
+      "step": 7991
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023130650218204622,
+      "loss": 0.4953,
+      "step": 7992
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000231290550276092,
+      "loss": 0.6557,
+      "step": 7993
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023127459706839156,
+      "loss": 0.0048,
+      "step": 7994
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023125864255920046,
+      "loss": 0.0043,
+      "step": 7995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023124268674877416,
+      "loss": 0.0076,
+      "step": 7996
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002312267296373682,
+      "loss": 0.6705,
+      "step": 7997
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023121077122523797,
+      "loss": 0.6016,
+      "step": 7998
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002311948115126392,
+      "loss": 0.1583,
+      "step": 7999
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023117885049982732,
+      "loss": 1.2713,
+      "step": 8000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023116288818705806,
+      "loss": 0.4147,
+      "step": 8001
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023114692457458687,
+      "loss": 0.8815,
+      "step": 8002
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002311309596626695,
+      "loss": 0.1502,
+      "step": 8003
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002311149934515616,
+      "loss": 0.4656,
+      "step": 8004
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023109902594151878,
+      "loss": 0.4908,
+      "step": 8005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023108305713279677,
+      "loss": 0.3918,
+      "step": 8006
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002310670870256513,
+      "loss": 1.158,
+      "step": 8007
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002310511156203381,
+      "loss": 0.0757,
+      "step": 8008
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023103514291711292,
+      "loss": 0.4513,
+      "step": 8009
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023101916891623157,
+      "loss": 0.0064,
+      "step": 8010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023100319361794982,
+      "loss": 0.5257,
+      "step": 8011
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023098721702252345,
+      "loss": 0.5287,
+      "step": 8012
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023097123913020837,
+      "loss": 0.4415,
+      "step": 8013
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002309552599412604,
+      "loss": 0.2616,
+      "step": 8014
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023093927945593544,
+      "loss": 0.387,
+      "step": 8015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023092329767448943,
+      "loss": 0.3309,
+      "step": 8016
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023090731459717823,
+      "loss": 0.1674,
+      "step": 8017
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023089133022425782,
+      "loss": 0.4729,
+      "step": 8018
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023087534455598415,
+      "loss": 0.6591,
+      "step": 8019
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002308593575926132,
+      "loss": 0.4868,
+      "step": 8020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023084336933440095,
+      "loss": 0.0115,
+      "step": 8021
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023082737978160356,
+      "loss": 0.6212,
+      "step": 8022
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023081138893447697,
+      "loss": 0.4365,
+      "step": 8023
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023079539679327725,
+      "loss": 0.7832,
+      "step": 8024
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307794033582605,
+      "loss": 0.4735,
+      "step": 8025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307634086296828,
+      "loss": 0.2249,
+      "step": 8026
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307474126078004,
+      "loss": 0.8291,
+      "step": 8027
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023073141529286932,
+      "loss": 0.7453,
+      "step": 8028
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002307154166851458,
+      "loss": 0.0047,
+      "step": 8029
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023069941678488602,
+      "loss": 2.0215,
+      "step": 8030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002306834155923462,
+      "loss": 0.0888,
+      "step": 8031
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002306674131077825,
+      "loss": 4.7136,
+      "step": 8032
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023065140933145137,
+      "loss": 0.0337,
+      "step": 8033
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023063540426360892,
+      "loss": 0.6726,
+      "step": 8034
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023061939790451142,
+      "loss": 0.2983,
+      "step": 8035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002306033902544153,
+      "loss": 0.7096,
+      "step": 8036
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023058738131357686,
+      "loss": 0.6242,
+      "step": 8037
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023057137108225244,
+      "loss": 0.6193,
+      "step": 8038
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023055535956069843,
+      "loss": 0.3548,
+      "step": 8039
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023053934674917124,
+      "loss": 0.2577,
+      "step": 8040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023052333264792727,
+      "loss": 0.2002,
+      "step": 8041
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023050731725722304,
+      "loss": 0.6309,
+      "step": 8042
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023049130057731484,
+      "loss": 0.0048,
+      "step": 8043
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023047528260845932,
+      "loss": 0.4908,
+      "step": 8044
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000230459263350913,
+      "loss": 0.2844,
+      "step": 8045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023044324280493224,
+      "loss": 0.1815,
+      "step": 8046
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023042722097077364,
+      "loss": 0.6087,
+      "step": 8047
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002304111978486939,
+      "loss": 0.547,
+      "step": 8048
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023039517343894942,
+      "loss": 0.4853,
+      "step": 8049
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023037914774179694,
+      "loss": 0.5914,
+      "step": 8050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000230363120757493,
+      "loss": 0.8468,
+      "step": 8051
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023034709248629433,
+      "loss": 0.5012,
+      "step": 8052
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023033106292845757,
+      "loss": 0.1789,
+      "step": 8053
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023031503208423943,
+      "loss": 0.8467,
+      "step": 8054
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023029899995389647,
+      "loss": 0.2088,
+      "step": 8055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002302829665376856,
+      "loss": 0.3905,
+      "step": 8056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023026693183586354,
+      "loss": 0.0022,
+      "step": 8057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023025089584868697,
+      "loss": 0.3804,
+      "step": 8058
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002302348585764128,
+      "loss": 0.3116,
+      "step": 8059
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023021882001929777,
+      "loss": 0.6545,
+      "step": 8060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002302027801775987,
+      "loss": 0.0046,
+      "step": 8061
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002301867390515725,
+      "loss": 0.3614,
+      "step": 8062
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023017069664147598,
+      "loss": 0.0047,
+      "step": 8063
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023015465294756612,
+      "loss": 2.2529,
+      "step": 8064
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023013860797009979,
+      "loss": 0.2838,
+      "step": 8065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023012256170933386,
+      "loss": 0.7769,
+      "step": 8066
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023012256170933386,
+      "loss": 2.9376,
+      "step": 8067
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023010651416552542,
+      "loss": 0.8598,
+      "step": 8068
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023009046533893138,
+      "loss": 0.0567,
+      "step": 8069
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023007441522980871,
+      "loss": 0.4985,
+      "step": 8070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023005836383841442,
+      "loss": 0.1695,
+      "step": 8071
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023004231116500567,
+      "loss": 3.2652,
+      "step": 8072
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023002625720983935,
+      "loss": 0.4181,
+      "step": 8073
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00023001020197317262,
+      "loss": 0.5507,
+      "step": 8074
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022999414545526266,
+      "loss": 0.2148,
+      "step": 8075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022997808765636647,
+      "loss": 0.1808,
+      "step": 8076
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022996202857674127,
+      "loss": 0.4143,
+      "step": 8077
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022994596821664413,
+      "loss": 0.0274,
+      "step": 8078
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002299299065763323,
+      "loss": 0.349,
+      "step": 8079
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000229913843656063,
+      "loss": 0.4298,
+      "step": 8080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022989777945609349,
+      "loss": 0.1367,
+      "step": 8081
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022988171397668081,
+      "loss": 0.8581,
+      "step": 8082
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022986564721808248,
+      "loss": 1.3616,
+      "step": 8083
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002298495791805556,
+      "loss": 0.3377,
+      "step": 8084
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002298335098643576,
+      "loss": 0.0221,
+      "step": 8085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022981743926974574,
+      "loss": 0.4434,
+      "step": 8086
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022980136739697736,
+      "loss": 0.4765,
+      "step": 8087
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022978529424630986,
+      "loss": 0.0082,
+      "step": 8088
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022976921981800064,
+      "loss": 0.8037,
+      "step": 8089
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022975314411230704,
+      "loss": 0.8975,
+      "step": 8090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022973706712948655,
+      "loss": 1.1475,
+      "step": 8091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002297209888697966,
+      "loss": 1.3627,
+      "step": 8092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022970490933349463,
+      "loss": 0.2742,
+      "step": 8093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022968882852083817,
+      "loss": 0.7156,
+      "step": 8094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022967274643208475,
+      "loss": 0.0889,
+      "step": 8095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022965666306749182,
+      "loss": 0.5415,
+      "step": 8096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022964057842731698,
+      "loss": 0.3956,
+      "step": 8097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022962449251181786,
+      "loss": 0.0129,
+      "step": 8098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022962449251181786,
+      "loss": 0.5752,
+      "step": 8099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00022960840532125195,
+      "loss": 0.624,
+      "step": 8100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022959231685587688,
+      "loss": 0.7648,
+      "step": 8101
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022957622711595037,
+      "loss": 0.0415,
+      "step": 8102
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022956013610172997,
+      "loss": 0.4063,
+      "step": 8103
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002295440438134734,
+      "loss": 0.4597,
+      "step": 8104
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022952795025143837,
+      "loss": 0.1874,
+      "step": 8105
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022951185541588247,
+      "loss": 0.3829,
+      "step": 8106
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022949575930706365,
+      "loss": 0.5823,
+      "step": 8107
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022947966192523954,
+      "loss": 0.2673,
+      "step": 8108
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022946356327066788,
+      "loss": 1.0564,
+      "step": 8109
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022944746334360657,
+      "loss": 0.1057,
+      "step": 8110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002294313621443133,
+      "loss": 0.2694,
+      "step": 8111
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000229415259673046,
+      "loss": 2.6759,
+      "step": 8112
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002293991559300625,
+      "loss": 0.619,
+      "step": 8113
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002293830509156207,
+      "loss": 0.1164,
+      "step": 8114
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022936694462997843,
+      "loss": 0.8767,
+      "step": 8115
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002293508370733937,
+      "loss": 0.2279,
+      "step": 8116
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002293347282461244,
+      "loss": 0.7005,
+      "step": 8117
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022931861814842847,
+      "loss": 0.4098,
+      "step": 8118
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022930250678056396,
+      "loss": 0.2264,
+      "step": 8119
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002292863941427888,
+      "loss": 0.2973,
+      "step": 8120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022927028023536096,
+      "loss": 1.2573,
+      "step": 8121
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022925416505853864,
+      "loss": 0.5366,
+      "step": 8122
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002292380486125798,
+      "loss": 0.252,
+      "step": 8123
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022922193089774248,
+      "loss": 0.1055,
+      "step": 8124
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022920581191428487,
+      "loss": 0.3118,
+      "step": 8125
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022918969166246502,
+      "loss": 0.0162,
+      "step": 8126
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002291735701425411,
+      "loss": 3.1779,
+      "step": 8127
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022915744735477132,
+      "loss": 0.1358,
+      "step": 8128
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022914132329941383,
+      "loss": 0.6493,
+      "step": 8129
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022912519797672678,
+      "loss": 0.4387,
+      "step": 8130
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022910907138696848,
+      "loss": 0.0158,
+      "step": 8131
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002290929435303971,
+      "loss": 0.308,
+      "step": 8132
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022907681440727088,
+      "loss": 0.431,
+      "step": 8133
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022906068401784823,
+      "loss": 0.9619,
+      "step": 8134
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022904455236238738,
+      "loss": 0.4237,
+      "step": 8135
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002290284194411466,
+      "loss": 1.2997,
+      "step": 8136
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022901228525438433,
+      "loss": 0.9286,
+      "step": 8137
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022899614980235886,
+      "loss": 0.348,
+      "step": 8138
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022898001308532863,
+      "loss": 0.1763,
+      "step": 8139
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000228963875103552,
+      "loss": 0.5453,
+      "step": 8140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022894773585728742,
+      "loss": 0.8753,
+      "step": 8141
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022893159534679338,
+      "loss": 0.3542,
+      "step": 8142
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022891545357232827,
+      "loss": 0.283,
+      "step": 8143
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002288993105341506,
+      "loss": 0.6846,
+      "step": 8144
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022888316623251892,
+      "loss": 0.0219,
+      "step": 8145
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022886702066769168,
+      "loss": 0.6596,
+      "step": 8146
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022885087383992745,
+      "loss": 0.0111,
+      "step": 8147
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002288347257494849,
+      "loss": 0.321,
+      "step": 8148
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022881857639662245,
+      "loss": 0.1339,
+      "step": 8149
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022880242578159882,
+      "loss": 0.0097,
+      "step": 8150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022878627390467257,
+      "loss": 0.2885,
+      "step": 8151
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002287701207661024,
+      "loss": 0.8259,
+      "step": 8152
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022875396636614699,
+      "loss": 0.3255,
+      "step": 8153
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022873781070506503,
+      "loss": 0.6498,
+      "step": 8154
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022872165378311508,
+      "loss": 0.0983,
+      "step": 8155
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002287054956005561,
+      "loss": 0.4694,
+      "step": 8156
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022868933615764671,
+      "loss": 1.3575,
+      "step": 8157
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022867317545464567,
+      "loss": 0.627,
+      "step": 8158
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022865701349181184,
+      "loss": 0.4974,
+      "step": 8159
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022864085026940394,
+      "loss": 0.1758,
+      "step": 8160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022862468578768085,
+      "loss": 0.4333,
+      "step": 8161
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022860852004690145,
+      "loss": 0.0938,
+      "step": 8162
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022859235304732457,
+      "loss": 0.1669,
+      "step": 8163
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022857618478920908,
+      "loss": 0.1461,
+      "step": 8164
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022856001527281393,
+      "loss": 2.9603,
+      "step": 8165
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022854384449839802,
+      "loss": 0.342,
+      "step": 8166
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022852767246622034,
+      "loss": 0.536,
+      "step": 8167
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022851149917653985,
+      "loss": 2.2305,
+      "step": 8168
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002284953246296155,
+      "loss": 0.241,
+      "step": 8169
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022847914882570634,
+      "loss": 0.245,
+      "step": 8170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022846297176507143,
+      "loss": 0.567,
+      "step": 8171
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022844679344796972,
+      "loss": 0.4776,
+      "step": 8172
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022843061387466037,
+      "loss": 0.0365,
+      "step": 8173
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022841443304540246,
+      "loss": 0.5153,
+      "step": 8174
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022839825096045507,
+      "loss": 0.551,
+      "step": 8175
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022838206762007735,
+      "loss": 0.2012,
+      "step": 8176
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022836588302452848,
+      "loss": 0.5615,
+      "step": 8177
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022834969717406761,
+      "loss": 2.1484,
+      "step": 8178
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002283335100689539,
+      "loss": 0.5051,
+      "step": 8179
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022831732170944662,
+      "loss": 0.4393,
+      "step": 8180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022830113209580492,
+      "loss": 0.0632,
+      "step": 8181
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022828494122828808,
+      "loss": 1.7911,
+      "step": 8182
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022826874910715552,
+      "loss": 0.218,
+      "step": 8183
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022825255573266632,
+      "loss": 0.6315,
+      "step": 8184
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022823636110507983,
+      "loss": 0.3232,
+      "step": 8185
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022822016522465552,
+      "loss": 0.3466,
+      "step": 8186
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022820396809165262,
+      "loss": 0.2463,
+      "step": 8187
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022818776970633055,
+      "loss": 0.5721,
+      "step": 8188
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002281715700689487,
+      "loss": 0.4684,
+      "step": 8189
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022815536917976647,
+      "loss": 0.8092,
+      "step": 8190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002281391670390433,
+      "loss": 1.2417,
+      "step": 8191
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002281229636470386,
+      "loss": 1.1328,
+      "step": 8192
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022810675900401193,
+      "loss": 0.1725,
+      "step": 8193
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022809055311022274,
+      "loss": 0.5785,
+      "step": 8194
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002280743459659305,
+      "loss": 0.2857,
+      "step": 8195
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022805813757139476,
+      "loss": 0.4374,
+      "step": 8196
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022804192792687518,
+      "loss": 0.665,
+      "step": 8197
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002280257170326312,
+      "loss": 0.0222,
+      "step": 8198
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022800950488892252,
+      "loss": 0.0231,
+      "step": 8199
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002279932914960086,
+      "loss": 0.3295,
+      "step": 8200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022797707685414921,
+      "loss": 0.5841,
+      "step": 8201
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000227960860963604,
+      "loss": 0.0113,
+      "step": 8202
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022794464382463255,
+      "loss": 0.9371,
+      "step": 8203
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002279284254374946,
+      "loss": 0.2064,
+      "step": 8204
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022791220580244993,
+      "loss": 0.7778,
+      "step": 8205
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022789598491975818,
+      "loss": 0.0896,
+      "step": 8206
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022787976278967914,
+      "loss": 1.3944,
+      "step": 8207
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022786353941247258,
+      "loss": 0.4054,
+      "step": 8208
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002278473147883983,
+      "loss": 0.3591,
+      "step": 8209
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002278310889177161,
+      "loss": 0.8179,
+      "step": 8210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022781486180068578,
+      "loss": 0.4179,
+      "step": 8211
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002277986334375673,
+      "loss": 0.566,
+      "step": 8212
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022778240382862044,
+      "loss": 2.4319,
+      "step": 8213
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022776617297410508,
+      "loss": 0.0111,
+      "step": 8214
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022774994087428123,
+      "loss": 0.7219,
+      "step": 8215
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022773370752940871,
+      "loss": 0.4527,
+      "step": 8216
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022771747293974756,
+      "loss": 0.532,
+      "step": 8217
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022770123710555776,
+      "loss": 0.5432,
+      "step": 8218
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022768500002709919,
+      "loss": 0.0362,
+      "step": 8219
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022766876170463195,
+      "loss": 0.6962,
+      "step": 8220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022765252213841606,
+      "loss": 0.8875,
+      "step": 8221
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022763628132871152,
+      "loss": 0.506,
+      "step": 8222
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022762003927577853,
+      "loss": 0.2842,
+      "step": 8223
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022760379597987706,
+      "loss": 2.237,
+      "step": 8224
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022758755144126728,
+      "loss": 0.005,
+      "step": 8225
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002275713056602093,
+      "loss": 1.1768,
+      "step": 8226
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022755505863696329,
+      "loss": 0.0047,
+      "step": 8227
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002275388103717894,
+      "loss": 0.6771,
+      "step": 8228
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022752256086494788,
+      "loss": 0.5765,
+      "step": 8229
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002275063101166988,
+      "loss": 0.4892,
+      "step": 8230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274900581273025,
+      "loss": 0.3037,
+      "step": 8231
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274738048970193,
+      "loss": 0.1683,
+      "step": 8232
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022745755042610931,
+      "loss": 0.6786,
+      "step": 8233
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022744129471483285,
+      "loss": 0.2885,
+      "step": 8234
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274250377634504,
+      "loss": 0.2211,
+      "step": 8235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002274087795722221,
+      "loss": 0.3625,
+      "step": 8236
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022739252014140833,
+      "loss": 0.8017,
+      "step": 8237
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022737625947126964,
+      "loss": 0.3341,
+      "step": 8238
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022735999756206616,
+      "loss": 0.3047,
+      "step": 8239
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022734373441405845,
+      "loss": 0.7293,
+      "step": 8240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022732747002750692,
+      "loss": 1.2222,
+      "step": 8241
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022731120440267199,
+      "loss": 0.0212,
+      "step": 8242
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022729493753981416,
+      "loss": 0.4959,
+      "step": 8243
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022727866943919394,
+      "loss": 0.1933,
+      "step": 8244
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022726240010107176,
+      "loss": 0.2662,
+      "step": 8245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022724612952570823,
+      "loss": 0.4346,
+      "step": 8246
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022722985771336386,
+      "loss": 0.8405,
+      "step": 8247
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002272135846642992,
+      "loss": 0.6859,
+      "step": 8248
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022719731037877487,
+      "loss": 0.7272,
+      "step": 8249
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022718103485705148,
+      "loss": 0.2338,
+      "step": 8250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022716475809938966,
+      "loss": 0.0026,
+      "step": 8251
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022714848010605004,
+      "loss": 0.4619,
+      "step": 8252
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002271322008772933,
+      "loss": 0.3761,
+      "step": 8253
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002271159204133801,
+      "loss": 0.9796,
+      "step": 8254
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002270996387145712,
+      "loss": 0.3778,
+      "step": 8255
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022708335578112726,
+      "loss": 0.0132,
+      "step": 8256
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022706707161330907,
+      "loss": 0.0131,
+      "step": 8257
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022705078621137743,
+      "loss": 0.084,
+      "step": 8258
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002270344995755931,
+      "loss": 0.6443,
+      "step": 8259
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022701821170621677,
+      "loss": 0.288,
+      "step": 8260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022700192260350946,
+      "loss": 0.3199,
+      "step": 8261
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002269856322677319,
+      "loss": 0.5678,
+      "step": 8262
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022696934069914496,
+      "loss": 0.3081,
+      "step": 8263
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022695304789800954,
+      "loss": 0.1836,
+      "step": 8264
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022693675386458657,
+      "loss": 0.6103,
+      "step": 8265
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022692045859913694,
+      "loss": 0.0025,
+      "step": 8266
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022690416210192162,
+      "loss": 0.2223,
+      "step": 8267
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022688786437320154,
+      "loss": 0.3882,
+      "step": 8268
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002268715654132377,
+      "loss": 1.116,
+      "step": 8269
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022685526522229113,
+      "loss": 1.1182,
+      "step": 8270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022683896380062282,
+      "loss": 0.6209,
+      "step": 8271
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022682266114849387,
+      "loss": 0.8998,
+      "step": 8272
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022680635726616524,
+      "loss": 0.5881,
+      "step": 8273
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022679005215389808,
+      "loss": 0.325,
+      "step": 8274
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022677374581195347,
+      "loss": 0.8704,
+      "step": 8275
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002267574382405926,
+      "loss": 0.7984,
+      "step": 8276
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002267411294400765,
+      "loss": 0.2424,
+      "step": 8277
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022672481941066636,
+      "loss": 0.2984,
+      "step": 8278
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002267085081526234,
+      "loss": 0.4074,
+      "step": 8279
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022669219566620884,
+      "loss": 0.302,
+      "step": 8280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002266758819516839,
+      "loss": 0.0058,
+      "step": 8281
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022665956700930972,
+      "loss": 0.0107,
+      "step": 8282
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022664325083934762,
+      "loss": 0.5092,
+      "step": 8283
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022662693344205893,
+      "loss": 0.5312,
+      "step": 8284
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022661061481770489,
+      "loss": 0.9081,
+      "step": 8285
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022659429496654676,
+      "loss": 1.2432,
+      "step": 8286
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022657797388884603,
+      "loss": 0.5364,
+      "step": 8287
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022656165158486393,
+      "loss": 2.379,
+      "step": 8288
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022654532805486193,
+      "loss": 0.3409,
+      "step": 8289
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022652900329910133,
+      "loss": 0.6104,
+      "step": 8290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022651267731784363,
+      "loss": 0.326,
+      "step": 8291
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002264963501113502,
+      "loss": 0.2083,
+      "step": 8292
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022648002167988257,
+      "loss": 3.417,
+      "step": 8293
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002264636920237021,
+      "loss": 0.2194,
+      "step": 8294
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022644736114307045,
+      "loss": 1.0635,
+      "step": 8295
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022643102903824905,
+      "loss": 0.8968,
+      "step": 8296
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022641469570949932,
+      "loss": 0.2488,
+      "step": 8297
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000226398361157083,
+      "loss": 0.5761,
+      "step": 8298
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022638202538126153,
+      "loss": 0.7806,
+      "step": 8299
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022636568838229663,
+      "loss": 0.341,
+      "step": 8300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002263493501604498,
+      "loss": 0.3419,
+      "step": 8301
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022633301071598268,
+      "loss": 0.5054,
+      "step": 8302
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022631667004915698,
+      "loss": 0.1123,
+      "step": 8303
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002263003281602344,
+      "loss": 1.1137,
+      "step": 8304
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022628398504947648,
+      "loss": 0.775,
+      "step": 8305
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022626764071714505,
+      "loss": 0.0239,
+      "step": 8306
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002262512951635019,
+      "loss": 0.1598,
+      "step": 8307
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002262349483888086,
+      "loss": 0.7802,
+      "step": 8308
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022621860039332702,
+      "loss": 0.0754,
+      "step": 8309
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022620225117731902,
+      "loss": 0.0683,
+      "step": 8310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022618590074104623,
+      "loss": 0.8892,
+      "step": 8311
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022616954908477062,
+      "loss": 0.5877,
+      "step": 8312
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000226153196208754,
+      "loss": 0.2703,
+      "step": 8313
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022613684211325827,
+      "loss": 0.0707,
+      "step": 8314
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022612048679854522,
+      "loss": 0.351,
+      "step": 8315
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022610413026487688,
+      "loss": 0.095,
+      "step": 8316
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022608777251251504,
+      "loss": 0.4292,
+      "step": 8317
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022607141354172172,
+      "loss": 1.8748,
+      "step": 8318
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022605505335275894,
+      "loss": 0.478,
+      "step": 8319
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022603869194588857,
+      "loss": 0.0036,
+      "step": 8320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022602232932137265,
+      "loss": 0.9056,
+      "step": 8321
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022600596547947327,
+      "loss": 0.3001,
+      "step": 8322
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022598960042045243,
+      "loss": 0.0333,
+      "step": 8323
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022597323414457217,
+      "loss": 0.2961,
+      "step": 8324
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022595686665209455,
+      "loss": 0.2897,
+      "step": 8325
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022594049794328176,
+      "loss": 0.5942,
+      "step": 8326
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022592412801839578,
+      "loss": 0.1688,
+      "step": 8327
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002259077568776989,
+      "loss": 0.5017,
+      "step": 8328
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022589138452145323,
+      "loss": 0.6402,
+      "step": 8329
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022587501094992093,
+      "loss": 0.1794,
+      "step": 8330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022585863616336418,
+      "loss": 0.0133,
+      "step": 8331
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022584226016204517,
+      "loss": 0.2524,
+      "step": 8332
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022582588294622625,
+      "loss": 1.0031,
+      "step": 8333
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022580950451616964,
+      "loss": 1.5944,
+      "step": 8334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022579312487213752,
+      "loss": 0.2871,
+      "step": 8335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002257767440143923,
+      "loss": 0.3276,
+      "step": 8336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022576036194319623,
+      "loss": 0.0488,
+      "step": 8337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002257439786588117,
+      "loss": 0.7216,
+      "step": 8338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022572759416150095,
+      "loss": 0.2291,
+      "step": 8339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002257112084515265,
+      "loss": 0.4775,
+      "step": 8340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022569482152915064,
+      "loss": 4.6282,
+      "step": 8341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022567843339463587,
+      "loss": 0.4318,
+      "step": 8342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022566204404824447,
+      "loss": 0.0174,
+      "step": 8343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022564565349023907,
+      "loss": 0.0064,
+      "step": 8344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000225629261720882,
+      "loss": 0.5026,
+      "step": 8345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022561286874043585,
+      "loss": 1.5622,
+      "step": 8346
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022559647454916304,
+      "loss": 0.1373,
+      "step": 8347
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022558007914732622,
+      "loss": 0.6722,
+      "step": 8348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022556368253518781,
+      "loss": 0.4274,
+      "step": 8349
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00022554728471301042,
+      "loss": 0.3487,
+      "step": 8350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022553088568105667,
+      "loss": 1.0663,
+      "step": 8351
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022551448543958914,
+      "loss": 0.0121,
+      "step": 8352
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022549808398887044,
+      "loss": 0.3284,
+      "step": 8353
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022548168132916326,
+      "loss": 0.1342,
+      "step": 8354
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022546527746073022,
+      "loss": 0.2962,
+      "step": 8355
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022544887238383404,
+      "loss": 0.0332,
+      "step": 8356
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022543246609873734,
+      "loss": 0.5122,
+      "step": 8357
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022541605860570293,
+      "loss": 0.371,
+      "step": 8358
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022539964990499355,
+      "loss": 0.7154,
+      "step": 8359
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022538323999687194,
+      "loss": 0.0236,
+      "step": 8360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022536682888160087,
+      "loss": 0.7173,
+      "step": 8361
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022535041655944314,
+      "loss": 0.312,
+      "step": 8362
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022533400303066155,
+      "loss": 0.9727,
+      "step": 8363
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000225317588295519,
+      "loss": 0.0157,
+      "step": 8364
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002253011723542783,
+      "loss": 1.1089,
+      "step": 8365
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002252847552072023,
+      "loss": 0.4488,
+      "step": 8366
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022526833685455395,
+      "loss": 0.1595,
+      "step": 8367
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022525191729659617,
+      "loss": 0.2178,
+      "step": 8368
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022523549653359184,
+      "loss": 0.4394,
+      "step": 8369
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022521907456580394,
+      "loss": 0.3636,
+      "step": 8370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022520265139349552,
+      "loss": 0.5689,
+      "step": 8371
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022518622701692945,
+      "loss": 0.0771,
+      "step": 8372
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022516980143636882,
+      "loss": 0.4296,
+      "step": 8373
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002251533746520766,
+      "loss": 0.6157,
+      "step": 8374
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002251369466643159,
+      "loss": 0.3743,
+      "step": 8375
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022512051747334975,
+      "loss": 0.8456,
+      "step": 8376
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002251040870794412,
+      "loss": 0.843,
+      "step": 8377
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022508765548285347,
+      "loss": 1.7217,
+      "step": 8378
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022507122268384963,
+      "loss": 0.94,
+      "step": 8379
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022505478868269283,
+      "loss": 0.3808,
+      "step": 8380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022503835347964618,
+      "loss": 0.1911,
+      "step": 8381
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000225021917074973,
+      "loss": 0.3331,
+      "step": 8382
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022500547946893636,
+      "loss": 0.1291,
+      "step": 8383
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002249890406617995,
+      "loss": 0.3341,
+      "step": 8384
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022497260065382576,
+      "loss": 0.5862,
+      "step": 8385
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002249561594452783,
+      "loss": 0.2757,
+      "step": 8386
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022493971703642046,
+      "loss": 0.6178,
+      "step": 8387
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002249232734275155,
+      "loss": 0.2776,
+      "step": 8388
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002249068286188268,
+      "loss": 0.486,
+      "step": 8389
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002248903826106176,
+      "loss": 0.5368,
+      "step": 8390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002248739354031514,
+      "loss": 0.4487,
+      "step": 8391
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022485748699669142,
+      "loss": 0.2867,
+      "step": 8392
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022484103739150112,
+      "loss": 0.4129,
+      "step": 8393
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000224824586587844,
+      "loss": 0.0028,
+      "step": 8394
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022480813458598337,
+      "loss": 0.4261,
+      "step": 8395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022479168138618273,
+      "loss": 0.4608,
+      "step": 8396
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002247752269887056,
+      "loss": 0.3801,
+      "step": 8397
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022475877139381542,
+      "loss": 0.2191,
+      "step": 8398
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002247423146017757,
+      "loss": 0.2041,
+      "step": 8399
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022472585661284996,
+      "loss": 2.2702,
+      "step": 8400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002247093974273018,
+      "loss": 0.3579,
+      "step": 8401
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022469293704539478,
+      "loss": 0.7816,
+      "step": 8402
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022467647546739244,
+      "loss": 0.3357,
+      "step": 8403
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022466001269355845,
+      "loss": 0.4364,
+      "step": 8404
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022464354872415637,
+      "loss": 0.0021,
+      "step": 8405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022462708355944997,
+      "loss": 0.0041,
+      "step": 8406
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002246106171997027,
+      "loss": 0.6824,
+      "step": 8407
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022459414964517844,
+      "loss": 0.0615,
+      "step": 8408
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022457768089614082,
+      "loss": 0.2604,
+      "step": 8409
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022456121095285357,
+      "loss": 0.0013,
+      "step": 8410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022454473981558038,
+      "loss": 1.1085,
+      "step": 8411
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002245282674845851,
+      "loss": 1.0031,
+      "step": 8412
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002245117939601315,
+      "loss": 0.5749,
+      "step": 8413
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022449531924248332,
+      "loss": 0.8032,
+      "step": 8414
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022447884333190432,
+      "loss": 0.0088,
+      "step": 8415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022446236622865852,
+      "loss": 0.7959,
+      "step": 8416
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022444588793300965,
+      "loss": 0.387,
+      "step": 8417
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002244294084452216,
+      "loss": 0.3502,
+      "step": 8418
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002244129277655583,
+      "loss": 0.5992,
+      "step": 8419
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022439644589428365,
+      "loss": 0.1507,
+      "step": 8420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022437996283166154,
+      "loss": 0.4296,
+      "step": 8421
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022436347857795593,
+      "loss": 0.2942,
+      "step": 8422
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022434699313343088,
+      "loss": 2.1213,
+      "step": 8423
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022433050649835028,
+      "loss": 0.4303,
+      "step": 8424
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022431401867297818,
+      "loss": 0.1737,
+      "step": 8425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002242975296575786,
+      "loss": 0.7613,
+      "step": 8426
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002242810394524156,
+      "loss": 0.1616,
+      "step": 8427
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002242645480577532,
+      "loss": 0.9463,
+      "step": 8428
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022424805547385557,
+      "loss": 0.239,
+      "step": 8429
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002242315617009867,
+      "loss": 0.0216,
+      "step": 8430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022421506673941079,
+      "loss": 0.5002,
+      "step": 8431
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022419857058939203,
+      "loss": 0.3459,
+      "step": 8432
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022418207325119443,
+      "loss": 0.0089,
+      "step": 8433
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022416557472508232,
+      "loss": 0.8542,
+      "step": 8434
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002241490750113198,
+      "loss": 0.5155,
+      "step": 8435
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022413257411017115,
+      "loss": 0.253,
+      "step": 8436
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022411607202190055,
+      "loss": 0.0006,
+      "step": 8437
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022409956874677234,
+      "loss": 0.0039,
+      "step": 8438
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022408306428505073,
+      "loss": 0.9001,
+      "step": 8439
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022406655863700005,
+      "loss": 0.2478,
+      "step": 8440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002240500518028845,
+      "loss": 1.5464,
+      "step": 8441
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022403354378296855,
+      "loss": 0.4493,
+      "step": 8442
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002240170345775166,
+      "loss": 0.2694,
+      "step": 8443
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002240005241867928,
+      "loss": 0.2769,
+      "step": 8444
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022398401261106168,
+      "loss": 0.5093,
+      "step": 8445
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002239674998505877,
+      "loss": 0.9106,
+      "step": 8446
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022395098590563515,
+      "loss": 0.3439,
+      "step": 8447
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002239344707764686,
+      "loss": 0.4817,
+      "step": 8448
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022391795446335242,
+      "loss": 0.0969,
+      "step": 8449
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022390143696655114,
+      "loss": 0.7231,
+      "step": 8450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022388491828632931,
+      "loss": 1.0555,
+      "step": 8451
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022386839842295132,
+      "loss": 0.4746,
+      "step": 8452
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022385187737668181,
+      "loss": 0.7461,
+      "step": 8453
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022383535514778537,
+      "loss": 1.161,
+      "step": 8454
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022381883173652653,
+      "loss": 0.0184,
+      "step": 8455
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022380230714316983,
+      "loss": 0.7398,
+      "step": 8456
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022378578136797998,
+      "loss": 0.6886,
+      "step": 8457
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022376925441122157,
+      "loss": 0.4809,
+      "step": 8458
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022375272627315926,
+      "loss": 2.9492,
+      "step": 8459
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022373619695405768,
+      "loss": 0.5613,
+      "step": 8460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022371966645418163,
+      "loss": 0.7093,
+      "step": 8461
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022370313477379573,
+      "loss": 0.312,
+      "step": 8462
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022368660191316475,
+      "loss": 0.4926,
+      "step": 8463
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022367006787255345,
+      "loss": 0.0191,
+      "step": 8464
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022365353265222653,
+      "loss": 0.3213,
+      "step": 8465
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022363699625244888,
+      "loss": 1.0947,
+      "step": 8466
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022362045867348518,
+      "loss": 0.489,
+      "step": 8467
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022360391991560038,
+      "loss": 0.6292,
+      "step": 8468
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022358737997905926,
+      "loss": 0.2414,
+      "step": 8469
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022357083886412666,
+      "loss": 0.4484,
+      "step": 8470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022355429657106748,
+      "loss": 0.0056,
+      "step": 8471
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002235377531001467,
+      "loss": 0.4446,
+      "step": 8472
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002235212084516291,
+      "loss": 0.0004,
+      "step": 8473
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002235046626257797,
+      "loss": 0.002,
+      "step": 8474
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022348811562286348,
+      "loss": 1.2784,
+      "step": 8475
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022347156744314536,
+      "loss": 0.4509,
+      "step": 8476
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022345501808689034,
+      "loss": 0.0079,
+      "step": 8477
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022343846755436346,
+      "loss": 0.0038,
+      "step": 8478
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002234219158458297,
+      "loss": 0.3577,
+      "step": 8479
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022340536296155416,
+      "loss": 0.1863,
+      "step": 8480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022338880890180194,
+      "loss": 0.8362,
+      "step": 8481
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022337225366683805,
+      "loss": 0.4398,
+      "step": 8482
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022335569725692766,
+      "loss": 0.5465,
+      "step": 8483
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022333913967233584,
+      "loss": 0.0034,
+      "step": 8484
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022332258091332783,
+      "loss": 0.3299,
+      "step": 8485
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022330602098016863,
+      "loss": 0.4515,
+      "step": 8486
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002232894598731236,
+      "loss": 0.3426,
+      "step": 8487
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002232728975924578,
+      "loss": 0.0568,
+      "step": 8488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022325633413843656,
+      "loss": 0.4477,
+      "step": 8489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022323976951132507,
+      "loss": 0.4705,
+      "step": 8490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022322320371138857,
+      "loss": 0.3596,
+      "step": 8491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022320663673889237,
+      "loss": 0.0013,
+      "step": 8492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022319006859410174,
+      "loss": 0.0186,
+      "step": 8493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000223173499277282,
+      "loss": 0.6151,
+      "step": 8494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002231569287886985,
+      "loss": 2.2647,
+      "step": 8495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022314035712861656,
+      "loss": 0.3678,
+      "step": 8496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022312378429730157,
+      "loss": 0.3718,
+      "step": 8497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022310721029501892,
+      "loss": 0.1739,
+      "step": 8498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022309063512203402,
+      "loss": 0.6113,
+      "step": 8499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022307405877861232,
+      "loss": 1.065,
+      "step": 8500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002230574812650192,
+      "loss": 0.2482,
+      "step": 8501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002230409025815202,
+      "loss": 0.0022,
+      "step": 8502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022302432272838072,
+      "loss": 2.0723,
+      "step": 8503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022300774170586635,
+      "loss": 0.3693,
+      "step": 8504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002229911595142425,
+      "loss": 0.3078,
+      "step": 8505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022297457615377483,
+      "loss": 0.1379,
+      "step": 8506
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022295799162472884,
+      "loss": 0.0022,
+      "step": 8507
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002229414059273701,
+      "loss": 0.5237,
+      "step": 8508
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002229248190619642,
+      "loss": 0.1726,
+      "step": 8509
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022290823102877677,
+      "loss": 0.3022,
+      "step": 8510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022289164182807345,
+      "loss": 0.2743,
+      "step": 8511
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022287505146011987,
+      "loss": 0.3797,
+      "step": 8512
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022285845992518172,
+      "loss": 0.746,
+      "step": 8513
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022284186722352467,
+      "loss": 0.2947,
+      "step": 8514
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022282527335541446,
+      "loss": 0.8549,
+      "step": 8515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022280867832111676,
+      "loss": 1.0635,
+      "step": 8516
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022279208212089738,
+      "loss": 0.4261,
+      "step": 8517
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022277548475502205,
+      "loss": 0.0096,
+      "step": 8518
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022275888622375653,
+      "loss": 0.6491,
+      "step": 8519
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022274228652736661,
+      "loss": 0.6269,
+      "step": 8520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002227256856661182,
+      "loss": 0.3261,
+      "step": 8521
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002227090836402771,
+      "loss": 0.2577,
+      "step": 8522
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002226924804501091,
+      "loss": 0.6193,
+      "step": 8523
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022267587609588012,
+      "loss": 1.1055,
+      "step": 8524
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002226592705778561,
+      "loss": 0.3768,
+      "step": 8525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022264266389630284,
+      "loss": 0.0045,
+      "step": 8526
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002226260560514864,
+      "loss": 0.6156,
+      "step": 8527
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002226094470436726,
+      "loss": 0.0155,
+      "step": 8528
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022259283687312752,
+      "loss": 0.9287,
+      "step": 8529
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022257622554011715,
+      "loss": 1.0075,
+      "step": 8530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022255961304490734,
+      "loss": 0.6903,
+      "step": 8531
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002225429993877643,
+      "loss": 0.971,
+      "step": 8532
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022252638456895404,
+      "loss": 0.6342,
+      "step": 8533
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002225097685887425,
+      "loss": 0.0088,
+      "step": 8534
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022249315144739587,
+      "loss": 0.4427,
+      "step": 8535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002224765331451802,
+      "loss": 0.2513,
+      "step": 8536
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022245991368236161,
+      "loss": 1.331,
+      "step": 8537
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022244329305920628,
+      "loss": 0.4004,
+      "step": 8538
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022242667127598032,
+      "loss": 0.2837,
+      "step": 8539
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002224100483329499,
+      "loss": 0.0059,
+      "step": 8540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022239342423038128,
+      "loss": 1.2305,
+      "step": 8541
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002223767989685406,
+      "loss": 1.8194,
+      "step": 8542
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022236017254769407,
+      "loss": 0.2008,
+      "step": 8543
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000222343544968108,
+      "loss": 0.4602,
+      "step": 8544
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022232691623004866,
+      "loss": 1.8731,
+      "step": 8545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022231028633378226,
+      "loss": 0.5607,
+      "step": 8546
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022229365527957516,
+      "loss": 0.6926,
+      "step": 8547
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022227702306769364,
+      "loss": 0.2218,
+      "step": 8548
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022226038969840408,
+      "loss": 0.2936,
+      "step": 8549
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022224375517197287,
+      "loss": 0.7998,
+      "step": 8550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002222271194886663,
+      "loss": 0.0002,
+      "step": 8551
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022221048264875083,
+      "loss": 0.7073,
+      "step": 8552
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022219384465249286,
+      "loss": 3.5508,
+      "step": 8553
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002221772055001588,
+      "loss": 0.5191,
+      "step": 8554
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022216056519201512,
+      "loss": 0.0006,
+      "step": 8555
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022214392372832832,
+      "loss": 0.0039,
+      "step": 8556
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022212728110936479,
+      "loss": 1.1046,
+      "step": 8557
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002221106373353911,
+      "loss": 0.0073,
+      "step": 8558
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022209399240667382,
+      "loss": 0.2758,
+      "step": 8559
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022207734632347944,
+      "loss": 0.8091,
+      "step": 8560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022206069908607454,
+      "loss": 0.0023,
+      "step": 8561
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002220440506947257,
+      "loss": 0.3643,
+      "step": 8562
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022202740114969947,
+      "loss": 0.2454,
+      "step": 8563
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022201075045126254,
+      "loss": 1.0036,
+      "step": 8564
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002219940985996815,
+      "loss": 0.283,
+      "step": 8565
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221977445595223,
+      "loss": 3.5047,
+      "step": 8566
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002219607914381538,
+      "loss": 0.1272,
+      "step": 8567
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022194413612874048,
+      "loss": 0.3959,
+      "step": 8568
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022192747966724976,
+      "loss": 0.0545,
+      "step": 8569
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022191082205394848,
+      "loss": 0.3236,
+      "step": 8570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022189416328910323,
+      "loss": 0.428,
+      "step": 8571
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022187750337298087,
+      "loss": 0.4686,
+      "step": 8572
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002218608423058482,
+      "loss": 0.3532,
+      "step": 8573
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022184418008797195,
+      "loss": 0.6764,
+      "step": 8574
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221827516719619,
+      "loss": 0.0027,
+      "step": 8575
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002218108522010562,
+      "loss": 0.5252,
+      "step": 8576
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002217941865325503,
+      "loss": 0.001,
+      "step": 8577
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002217775197143683,
+      "loss": 0.4286,
+      "step": 8578
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022176085174677705,
+      "loss": 0.3665,
+      "step": 8579
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022174418263004342,
+      "loss": 0.7016,
+      "step": 8580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002217275123644344,
+      "loss": 0.6655,
+      "step": 8581
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022171084095021695,
+      "loss": 0.449,
+      "step": 8582
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022169416838765797,
+      "loss": 0.001,
+      "step": 8583
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022167749467702448,
+      "loss": 6.2202,
+      "step": 8584
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002216608198185835,
+      "loss": 0.3454,
+      "step": 8585
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221644143812602,
+      "loss": 0.3937,
+      "step": 8586
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002216274666593471,
+      "loss": 2.238,
+      "step": 8587
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002216107883590858,
+      "loss": 0.3427,
+      "step": 8588
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002215941089120852,
+      "loss": 0.5727,
+      "step": 8589
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002215774283186124,
+      "loss": 4.7527,
+      "step": 8590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022156074657893454,
+      "loss": 0.4382,
+      "step": 8591
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022154406369331866,
+      "loss": 0.7356,
+      "step": 8592
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022152737966203202,
+      "loss": 0.9355,
+      "step": 8593
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022151069448534172,
+      "loss": 0.759,
+      "step": 8594
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022149400816351498,
+      "loss": 0.3712,
+      "step": 8595
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000221477320696819,
+      "loss": 0.1523,
+      "step": 8596
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022146063208552095,
+      "loss": 1.0869,
+      "step": 8597
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00022144394232988818,
+      "loss": 0.2291,
+      "step": 8598
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002214272514301879,
+      "loss": 0.0769,
+      "step": 8599
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022141055938668734,
+      "loss": 0.4531,
+      "step": 8600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022139386619965387,
+      "loss": 0.241,
+      "step": 8601
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002213771718693548,
+      "loss": 3.2272,
+      "step": 8602
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022136047639605743,
+      "loss": 1.2544,
+      "step": 8603
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022134377978002908,
+      "loss": 1.0365,
+      "step": 8604
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022132708202153727,
+      "loss": 0.0521,
+      "step": 8605
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022131038312084923,
+      "loss": 0.0147,
+      "step": 8606
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002212936830782324,
+      "loss": 0.3607,
+      "step": 8607
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002212769818939543,
+      "loss": 0.2033,
+      "step": 8608
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022126027956828224,
+      "loss": 0.0003,
+      "step": 8609
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022124357610148377,
+      "loss": 0.017,
+      "step": 8610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022122687149382634,
+      "loss": 0.4037,
+      "step": 8611
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022121016574557745,
+      "loss": 0.5847,
+      "step": 8612
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022119345885700467,
+      "loss": 0.4934,
+      "step": 8613
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022117675082837548,
+      "loss": 0.0082,
+      "step": 8614
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022116004165995744,
+      "loss": 3.5371,
+      "step": 8615
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022114333135201812,
+      "loss": 0.4355,
+      "step": 8616
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022112661990482518,
+      "loss": 0.0013,
+      "step": 8617
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002211099073186461,
+      "loss": 0.2367,
+      "step": 8618
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002210931935937486,
+      "loss": 0.3131,
+      "step": 8619
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002210764787304003,
+      "loss": 0.3583,
+      "step": 8620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022105976272886885,
+      "loss": 0.8522,
+      "step": 8621
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022104304558942194,
+      "loss": 0.2203,
+      "step": 8622
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002210263273123273,
+      "loss": 0.0052,
+      "step": 8623
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002210096078978526,
+      "loss": 0.1248,
+      "step": 8624
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022099288734626563,
+      "loss": 1.1251,
+      "step": 8625
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022097616565783408,
+      "loss": 0.4189,
+      "step": 8626
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022095944283282576,
+      "loss": 0.0047,
+      "step": 8627
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022094271887150853,
+      "loss": 0.7627,
+      "step": 8628
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022092599377415004,
+      "loss": 0.1832,
+      "step": 8629
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022090926754101824,
+      "loss": 0.8068,
+      "step": 8630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022089254017238093,
+      "loss": 0.2928,
+      "step": 8631
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000220875811668506,
+      "loss": 0.8651,
+      "step": 8632
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000220875811668506,
+      "loss": 0.9307,
+      "step": 8633
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022085908202966128,
+      "loss": 1.01,
+      "step": 8634
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022084235125611475,
+      "loss": 0.607,
+      "step": 8635
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022082561934813426,
+      "loss": 0.0008,
+      "step": 8636
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022080888630598777,
+      "loss": 0.6828,
+      "step": 8637
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022079215212994323,
+      "loss": 0.0026,
+      "step": 8638
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022077541682026863,
+      "loss": 1.2218,
+      "step": 8639
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022075868037723195,
+      "loss": 0.8746,
+      "step": 8640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022074194280110123,
+      "loss": 0.0443,
+      "step": 8641
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002207252040921444,
+      "loss": 0.6831,
+      "step": 8642
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002207084642506296,
+      "loss": 0.5846,
+      "step": 8643
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002206917232768249,
+      "loss": 0.0024,
+      "step": 8644
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002206749811709983,
+      "loss": 1.669,
+      "step": 8645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022065823793341794,
+      "loss": 0.4099,
+      "step": 8646
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022064149356435198,
+      "loss": 0.0023,
+      "step": 8647
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022062474806406847,
+      "loss": 0.4506,
+      "step": 8648
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022060800143283563,
+      "loss": 0.3121,
+      "step": 8649
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022059125367092163,
+      "loss": 0.9235,
+      "step": 8650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022057450477859465,
+      "loss": 0.1637,
+      "step": 8651
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022055775475612288,
+      "loss": 0.2242,
+      "step": 8652
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022054100360377455,
+      "loss": 0.3471,
+      "step": 8653
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022052425132181792,
+      "loss": 0.5369,
+      "step": 8654
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022050749791052127,
+      "loss": 0.3118,
+      "step": 8655
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022049074337015284,
+      "loss": 0.3271,
+      "step": 8656
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022047398770098093,
+      "loss": 4.5941,
+      "step": 8657
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022045723090327393,
+      "loss": 4.4245,
+      "step": 8658
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022044047297730006,
+      "loss": 0.4845,
+      "step": 8659
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022042371392332776,
+      "loss": 0.4597,
+      "step": 8660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022040695374162537,
+      "loss": 0.5926,
+      "step": 8661
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022039019243246126,
+      "loss": 1.0066,
+      "step": 8662
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022037342999610388,
+      "loss": 0.237,
+      "step": 8663
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022035666643282164,
+      "loss": 1.4674,
+      "step": 8664
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022033990174288295,
+      "loss": 0.8776,
+      "step": 8665
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022032313592655633,
+      "loss": 0.3582,
+      "step": 8666
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002203063689841102,
+      "loss": 0.0019,
+      "step": 8667
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002202896009158131,
+      "loss": 0.2968,
+      "step": 8668
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002202728317219335,
+      "loss": 0.4145,
+      "step": 8669
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022025606140274004,
+      "loss": 0.5611,
+      "step": 8670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022023928995850114,
+      "loss": 0.3518,
+      "step": 8671
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022022251738948546,
+      "loss": 0.0004,
+      "step": 8672
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002202057436959615,
+      "loss": 1.3623,
+      "step": 8673
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022018896887819798,
+      "loss": 0.2495,
+      "step": 8674
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002201721929364634,
+      "loss": 0.9903,
+      "step": 8675
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002201554158710265,
+      "loss": 0.2304,
+      "step": 8676
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022013863768215593,
+      "loss": 0.1785,
+      "step": 8677
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002201218583701203,
+      "loss": 0.1229,
+      "step": 8678
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002201050779351883,
+      "loss": 0.0051,
+      "step": 8679
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022008829637762878,
+      "loss": 0.1697,
+      "step": 8680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022007151369771036,
+      "loss": 0.7345,
+      "step": 8681
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022005472989570184,
+      "loss": 0.1829,
+      "step": 8682
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022003794497187185,
+      "loss": 0.2765,
+      "step": 8683
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00022002115892648936,
+      "loss": 0.4769,
+      "step": 8684
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002200043717598231,
+      "loss": 0.0015,
+      "step": 8685
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021998758347214188,
+      "loss": 0.2227,
+      "step": 8686
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021997079406371455,
+      "loss": 0.6967,
+      "step": 8687
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021995400353480994,
+      "loss": 0.5009,
+      "step": 8688
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021993721188569697,
+      "loss": 0.5382,
+      "step": 8689
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021992041911664454,
+      "loss": 4.1333,
+      "step": 8690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021990362522792146,
+      "loss": 3.004,
+      "step": 8691
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002198868302197968,
+      "loss": 0.473,
+      "step": 8692
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021987003409253944,
+      "loss": 0.5361,
+      "step": 8693
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021985323684641832,
+      "loss": 0.5987,
+      "step": 8694
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002198364384817024,
+      "loss": 0.1023,
+      "step": 8695
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002198196389986608,
+      "loss": 0.5614,
+      "step": 8696
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021980283839756246,
+      "loss": 0.2642,
+      "step": 8697
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002197860366786764,
+      "loss": 0.2508,
+      "step": 8698
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002197692338422717,
+      "loss": 0.0028,
+      "step": 8699
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002197524298886174,
+      "loss": 0.2894,
+      "step": 8700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021973562481798265,
+      "loss": 0.5811,
+      "step": 8701
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021971881863063652,
+      "loss": 0.7871,
+      "step": 8702
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021970201132684818,
+      "loss": 0.3349,
+      "step": 8703
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002196852029068867,
+      "loss": 0.2077,
+      "step": 8704
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002196683933710213,
+      "loss": 0.7759,
+      "step": 8705
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002196515827195211,
+      "loss": 0.3862,
+      "step": 8706
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021963477095265536,
+      "loss": 0.7271,
+      "step": 8707
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021961795807069328,
+      "loss": 0.4517,
+      "step": 8708
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021960114407390405,
+      "loss": 0.0465,
+      "step": 8709
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021958432896255697,
+      "loss": 0.0053,
+      "step": 8710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021956751273692127,
+      "loss": 0.7042,
+      "step": 8711
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021955069539726632,
+      "loss": 0.3268,
+      "step": 8712
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021953387694386133,
+      "loss": 0.3316,
+      "step": 8713
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021951705737697564,
+      "loss": 0.3373,
+      "step": 8714
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021950023669687862,
+      "loss": 1.6725,
+      "step": 8715
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021948341490383964,
+      "loss": 0.5886,
+      "step": 8716
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000219466591998128,
+      "loss": 0.0623,
+      "step": 8717
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021944976798001317,
+      "loss": 0.2652,
+      "step": 8718
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021943294284976458,
+      "loss": 3.9247,
+      "step": 8719
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021941611660765157,
+      "loss": 0.0016,
+      "step": 8720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002193992892539436,
+      "loss": 0.8008,
+      "step": 8721
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002193824607889102,
+      "loss": 0.4626,
+      "step": 8722
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021936563121282085,
+      "loss": 0.0047,
+      "step": 8723
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000219348800525945,
+      "loss": 2.7266,
+      "step": 8724
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021933196872855215,
+      "loss": 0.6241,
+      "step": 8725
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002193151358209119,
+      "loss": 0.5346,
+      "step": 8726
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021929830180329374,
+      "loss": 0.5356,
+      "step": 8727
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021928146667596736,
+      "loss": 0.6409,
+      "step": 8728
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021926463043920222,
+      "loss": 0.4471,
+      "step": 8729
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021924779309326798,
+      "loss": 0.6351,
+      "step": 8730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021923095463843425,
+      "loss": 1.2669,
+      "step": 8731
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021921411507497068,
+      "loss": 0.7464,
+      "step": 8732
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021919727440314698,
+      "loss": 0.7743,
+      "step": 8733
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021918043262323273,
+      "loss": 1.2121,
+      "step": 8734
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002191635897354977,
+      "loss": 0.294,
+      "step": 8735
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021914674574021156,
+      "loss": 0.1488,
+      "step": 8736
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002191299006376441,
+      "loss": 0.2684,
+      "step": 8737
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021911305442806497,
+      "loss": 0.9114,
+      "step": 8738
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021909620711174407,
+      "loss": 0.6119,
+      "step": 8739
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021907935868895103,
+      "loss": 0.3226,
+      "step": 8740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021906250915995577,
+      "loss": 0.5663,
+      "step": 8741
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021904565852502811,
+      "loss": 0.1303,
+      "step": 8742
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002190288067844378,
+      "loss": 0.4891,
+      "step": 8743
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021901195393845478,
+      "loss": 0.1841,
+      "step": 8744
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002189950999873489,
+      "loss": 0.49,
+      "step": 8745
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021897824493139005,
+      "loss": 0.0593,
+      "step": 8746
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021896138877084808,
+      "loss": 0.1767,
+      "step": 8747
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021894453150599297,
+      "loss": 2.6641,
+      "step": 8748
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002189276731370947,
+      "loss": 0.3046,
+      "step": 8749
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021891081366442317,
+      "loss": 0.0885,
+      "step": 8750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021889395308824836,
+      "loss": 0.787,
+      "step": 8751
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002188770914088403,
+      "loss": 0.1924,
+      "step": 8752
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021886022862646897,
+      "loss": 0.2731,
+      "step": 8753
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021884336474140448,
+      "loss": 1.9686,
+      "step": 8754
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021882649975391675,
+      "loss": 1.9004,
+      "step": 8755
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000218809633664276,
+      "loss": 0.2148,
+      "step": 8756
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002187927664727522,
+      "loss": 0.4486,
+      "step": 8757
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021877589817961547,
+      "loss": 2.2168,
+      "step": 8758
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021875902878513596,
+      "loss": 0.1676,
+      "step": 8759
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021874215828958382,
+      "loss": 0.3999,
+      "step": 8760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021872528669322915,
+      "loss": 0.3261,
+      "step": 8761
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021870841399634217,
+      "loss": 0.8384,
+      "step": 8762
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021869154019919307,
+      "loss": 0.6857,
+      "step": 8763
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021867466530205203,
+      "loss": 0.2105,
+      "step": 8764
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002186577893051893,
+      "loss": 0.0018,
+      "step": 8765
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002186409122088751,
+      "loss": 0.4115,
+      "step": 8766
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021862403401337968,
+      "loss": 0.5456,
+      "step": 8767
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021860715471897342,
+      "loss": 0.4421,
+      "step": 8768
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021859027432592654,
+      "loss": 0.4596,
+      "step": 8769
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021857339283450926,
+      "loss": 0.2352,
+      "step": 8770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021855651024499208,
+      "loss": 1.3201,
+      "step": 8771
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021853962655764527,
+      "loss": 0.8194,
+      "step": 8772
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002185227417727392,
+      "loss": 0.3445,
+      "step": 8773
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021850585589054422,
+      "loss": 0.4853,
+      "step": 8774
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021848896891133086,
+      "loss": 0.0436,
+      "step": 8775
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021847208083536937,
+      "loss": 0.6966,
+      "step": 8776
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021845519166293029,
+      "loss": 0.8579,
+      "step": 8777
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002184383013942841,
+      "loss": 0.3759,
+      "step": 8778
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021842141002970114,
+      "loss": 0.8152,
+      "step": 8779
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021840451756945206,
+      "loss": 0.6419,
+      "step": 8780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021838762401380724,
+      "loss": 0.011,
+      "step": 8781
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002183707293630373,
+      "loss": 0.002,
+      "step": 8782
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002183538336174127,
+      "loss": 0.0452,
+      "step": 8783
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021833693677720406,
+      "loss": 0.1787,
+      "step": 8784
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021832003884268193,
+      "loss": 0.186,
+      "step": 8785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021830313981411692,
+      "loss": 0.2287,
+      "step": 8786
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021828623969177963,
+      "loss": 0.0235,
+      "step": 8787
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021826933847594074,
+      "loss": 0.3764,
+      "step": 8788
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021825243616687077,
+      "loss": 0.324,
+      "step": 8789
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021823553276484046,
+      "loss": 0.7625,
+      "step": 8790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021821862827012058,
+      "loss": 0.4861,
+      "step": 8791
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021820172268298172,
+      "loss": 0.0133,
+      "step": 8792
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021818481600369458,
+      "loss": 0.6518,
+      "step": 8793
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021816790823253002,
+      "loss": 0.2534,
+      "step": 8794
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021815099936975868,
+      "loss": 0.6811,
+      "step": 8795
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021813408941565132,
+      "loss": 0.2067,
+      "step": 8796
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021811717837047883,
+      "loss": 0.3014,
+      "step": 8797
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000218100266234512,
+      "loss": 0.2895,
+      "step": 8798
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021808335300802152,
+      "loss": 0.5243,
+      "step": 8799
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021806643869127837,
+      "loss": 1.4847,
+      "step": 8800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002180495232845533,
+      "loss": 0.0038,
+      "step": 8801
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002180326067881173,
+      "loss": 0.3624,
+      "step": 8802
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021801568920224124,
+      "loss": 1.219,
+      "step": 8803
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021799877052719588,
+      "loss": 0.339,
+      "step": 8804
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021798185076325237,
+      "loss": 0.4835,
+      "step": 8805
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021796492991068151,
+      "loss": 0.1312,
+      "step": 8806
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021794800796975426,
+      "loss": 0.0247,
+      "step": 8807
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021793108494074168,
+      "loss": 0.3711,
+      "step": 8808
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021791416082391472,
+      "loss": 0.4936,
+      "step": 8809
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021789723561954437,
+      "loss": 0.5875,
+      "step": 8810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021788030932790172,
+      "loss": 0.5939,
+      "step": 8811
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002178633819492578,
+      "loss": 0.2771,
+      "step": 8812
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021784645348388366,
+      "loss": 0.1685,
+      "step": 8813
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021782952393205042,
+      "loss": 0.2139,
+      "step": 8814
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021781259329402907,
+      "loss": 0.5387,
+      "step": 8815
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021779566157009085,
+      "loss": 0.5607,
+      "step": 8816
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177787287605069,
+      "loss": 0.0012,
+      "step": 8817
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177617948655483,
+      "loss": 0.1029,
+      "step": 8818
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177448598854862,
+      "loss": 0.6446,
+      "step": 8819
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002177279238205919,
+      "loss": 0.0366,
+      "step": 8820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021771098667113653,
+      "loss": 0.4341,
+      "step": 8821
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021769404843739135,
+      "loss": 0.4939,
+      "step": 8822
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021767710911962753,
+      "loss": 0.4859,
+      "step": 8823
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002176601687181164,
+      "loss": 1.713,
+      "step": 8824
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021764322723312923,
+      "loss": 0.3344,
+      "step": 8825
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002176262846649373,
+      "loss": 0.9678,
+      "step": 8826
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021760934101381187,
+      "loss": 0.1818,
+      "step": 8827
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021759239628002435,
+      "loss": 0.4437,
+      "step": 8828
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021757545046384606,
+      "loss": 0.0013,
+      "step": 8829
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021755850356554825,
+      "loss": 0.0369,
+      "step": 8830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002175415555854025,
+      "loss": 0.0222,
+      "step": 8831
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021752460652368008,
+      "loss": 0.259,
+      "step": 8832
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021750765638065242,
+      "loss": 0.0355,
+      "step": 8833
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021749070515659092,
+      "loss": 0.4716,
+      "step": 8834
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021747375285176713,
+      "loss": 0.6401,
+      "step": 8835
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002174567994664524,
+      "loss": 0.3464,
+      "step": 8836
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021743984500091827,
+      "loss": 0.4707,
+      "step": 8837
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021742288945543627,
+      "loss": 0.3783,
+      "step": 8838
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021740593283027786,
+      "loss": 0.0338,
+      "step": 8839
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002173889751257146,
+      "loss": 0.3263,
+      "step": 8840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021737201634201806,
+      "loss": 0.2198,
+      "step": 8841
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021735505647945978,
+      "loss": 0.2699,
+      "step": 8842
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021733809553831134,
+      "loss": 0.8909,
+      "step": 8843
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021732113351884434,
+      "loss": 0.0567,
+      "step": 8844
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021730417042133043,
+      "loss": 1.5946,
+      "step": 8845
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021728720624604128,
+      "loss": 0.0012,
+      "step": 8846
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021727024099324854,
+      "loss": 0.2405,
+      "step": 8847
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00021725327466322377,
+      "loss": 0.1895,
+      "step": 8848
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021723630725623875,
+      "loss": 0.6809,
+      "step": 8849
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002172193387725652,
+      "loss": 0.4959,
+      "step": 8850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021720236921247477,
+      "loss": 0.7347,
+      "step": 8851
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021718539857623934,
+      "loss": 0.3306,
+      "step": 8852
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002171684268641305,
+      "loss": 0.0016,
+      "step": 8853
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002171514540764202,
+      "loss": 1.7875,
+      "step": 8854
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021713448021338006,
+      "loss": 3.408,
+      "step": 8855
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021711750527528205,
+      "loss": 0.0271,
+      "step": 8856
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021710052926239787,
+      "loss": 0.3248,
+      "step": 8857
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002170835521749994,
+      "loss": 0.0031,
+      "step": 8858
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021706657401335854,
+      "loss": 1.0043,
+      "step": 8859
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021704959477774718,
+      "loss": 0.5513,
+      "step": 8860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021703261446843713,
+      "loss": 0.0016,
+      "step": 8861
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021701563308570044,
+      "loss": 1.5444,
+      "step": 8862
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021699865062980888,
+      "loss": 0.4013,
+      "step": 8863
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021698166710103453,
+      "loss": 0.0248,
+      "step": 8864
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002169646824996493,
+      "loss": 0.0143,
+      "step": 8865
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002169476968259252,
+      "loss": 0.5314,
+      "step": 8866
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021693071008013418,
+      "loss": 0.3135,
+      "step": 8867
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021691372226254827,
+      "loss": 0.0084,
+      "step": 8868
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021689673337343956,
+      "loss": 0.0548,
+      "step": 8869
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021687974341308002,
+      "loss": 0.6688,
+      "step": 8870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002168627523817418,
+      "loss": 0.0711,
+      "step": 8871
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002168457602796969,
+      "loss": 0.5319,
+      "step": 8872
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021682876710721753,
+      "loss": 0.62,
+      "step": 8873
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021681177286457572,
+      "loss": 3.2071,
+      "step": 8874
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021679477755204364,
+      "loss": 0.2888,
+      "step": 8875
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021677778116989345,
+      "loss": 0.037,
+      "step": 8876
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002167607837183973,
+      "loss": 0.9983,
+      "step": 8877
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021674378519782745,
+      "loss": 0.3645,
+      "step": 8878
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021672678560845596,
+      "loss": 0.2312,
+      "step": 8879
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021670978495055517,
+      "loss": 0.0016,
+      "step": 8880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002166927832243973,
+      "loss": 0.8856,
+      "step": 8881
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021667578043025462,
+      "loss": 0.7018,
+      "step": 8882
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021665877656839937,
+      "loss": 0.4788,
+      "step": 8883
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021664177163910385,
+      "loss": 0.4066,
+      "step": 8884
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021662476564264038,
+      "loss": 0.0032,
+      "step": 8885
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021660775857928127,
+      "loss": 1.0489,
+      "step": 8886
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021659075044929888,
+      "loss": 0.693,
+      "step": 8887
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021657374125296554,
+      "loss": 0.6306,
+      "step": 8888
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021655673099055364,
+      "loss": 0.5645,
+      "step": 8889
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021653971966233563,
+      "loss": 0.027,
+      "step": 8890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021652270726858384,
+      "loss": 0.0703,
+      "step": 8891
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021650569380957075,
+      "loss": 0.0022,
+      "step": 8892
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021648867928556874,
+      "loss": 0.3016,
+      "step": 8893
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002164716636968503,
+      "loss": 1.2921,
+      "step": 8894
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000216454647043688,
+      "loss": 0.362,
+      "step": 8895
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021643762932635423,
+      "loss": 0.4816,
+      "step": 8896
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021642061054512153,
+      "loss": 0.8981,
+      "step": 8897
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021640359070026246,
+      "loss": 0.426,
+      "step": 8898
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002163865697920495,
+      "loss": 2.1494,
+      "step": 8899
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021636954782075529,
+      "loss": 1.7334,
+      "step": 8900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021635252478665237,
+      "loss": 0.7385,
+      "step": 8901
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021633550069001336,
+      "loss": 2.4922,
+      "step": 8902
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002163184755311108,
+      "loss": 0.2596,
+      "step": 8903
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002163014493102175,
+      "loss": 0.884,
+      "step": 8904
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002162844220276059,
+      "loss": 0.1098,
+      "step": 8905
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021626739368354875,
+      "loss": 0.883,
+      "step": 8906
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021625036427831882,
+      "loss": 0.6871,
+      "step": 8907
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002162333338121887,
+      "loss": 0.815,
+      "step": 8908
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021621630228543112,
+      "loss": 0.2252,
+      "step": 8909
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021619926969831886,
+      "loss": 0.2396,
+      "step": 8910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021618223605112462,
+      "loss": 0.4883,
+      "step": 8911
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002161652013441212,
+      "loss": 2.4012,
+      "step": 8912
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002161481655775814,
+      "loss": 0.5144,
+      "step": 8913
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000216131128751778,
+      "loss": 1.0138,
+      "step": 8914
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021611409086698386,
+      "loss": 0.0075,
+      "step": 8915
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021609705192347174,
+      "loss": 0.3058,
+      "step": 8916
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021608001192151448,
+      "loss": 0.5885,
+      "step": 8917
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021606297086138507,
+      "loss": 0.3098,
+      "step": 8918
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021604592874335628,
+      "loss": 1.9968,
+      "step": 8919
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021602888556770112,
+      "loss": 0.3168,
+      "step": 8920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021601184133469243,
+      "loss": 0.035,
+      "step": 8921
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021599479604460313,
+      "loss": 0.4557,
+      "step": 8922
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021597774969770627,
+      "loss": 0.8574,
+      "step": 8923
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021596070229427475,
+      "loss": 0.8847,
+      "step": 8924
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021594365383458154,
+      "loss": 2.4142,
+      "step": 8925
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002159266043188997,
+      "loss": 0.0624,
+      "step": 8926
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002159095537475023,
+      "loss": 0.4962,
+      "step": 8927
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021589250212066221,
+      "loss": 0.8207,
+      "step": 8928
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021587544943865264,
+      "loss": 0.7511,
+      "step": 8929
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021585839570174666,
+      "loss": 1.1817,
+      "step": 8930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021584134091021725,
+      "loss": 0.8309,
+      "step": 8931
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021582428506433754,
+      "loss": 1.3637,
+      "step": 8932
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002158072281643808,
+      "loss": 0.4771,
+      "step": 8933
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021579017021062,
+      "loss": 0.4775,
+      "step": 8934
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021577311120332836,
+      "loss": 0.3964,
+      "step": 8935
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021575605114277908,
+      "loss": 0.5249,
+      "step": 8936
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002157389900292453,
+      "loss": 1.3635,
+      "step": 8937
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002157219278630003,
+      "loss": 0.3457,
+      "step": 8938
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021570486464431724,
+      "loss": 0.0032,
+      "step": 8939
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021568780037346935,
+      "loss": 0.3031,
+      "step": 8940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021567073505072995,
+      "loss": 0.8308,
+      "step": 8941
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002156536686763723,
+      "loss": 0.4194,
+      "step": 8942
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002156366012506696,
+      "loss": 0.3766,
+      "step": 8943
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021561953277389535,
+      "loss": 0.0022,
+      "step": 8944
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002156024632463227,
+      "loss": 0.8364,
+      "step": 8945
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021558539266822505,
+      "loss": 0.0265,
+      "step": 8946
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002155683210398758,
+      "loss": 0.2467,
+      "step": 8947
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021555124836154826,
+      "loss": 0.3188,
+      "step": 8948
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021553417463351587,
+      "loss": 0.1051,
+      "step": 8949
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021551709985605203,
+      "loss": 3.7021,
+      "step": 8950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021550002402943015,
+      "loss": 0.0177,
+      "step": 8951
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002154829471539237,
+      "loss": 0.6594,
+      "step": 8952
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021546586922980613,
+      "loss": 1.9882,
+      "step": 8953
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002154487902573509,
+      "loss": 0.5961,
+      "step": 8954
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002154317102368315,
+      "loss": 0.7456,
+      "step": 8955
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021541462916852154,
+      "loss": 0.9529,
+      "step": 8956
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002153975470526944,
+      "loss": 0.2378,
+      "step": 8957
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002153804638896237,
+      "loss": 0.5849,
+      "step": 8958
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000215363379679583,
+      "loss": 0.5953,
+      "step": 8959
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021534629442284585,
+      "loss": 0.4211,
+      "step": 8960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021532920811968593,
+      "loss": 0.4184,
+      "step": 8961
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021531212077037672,
+      "loss": 0.3962,
+      "step": 8962
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021529503237519194,
+      "loss": 0.671,
+      "step": 8963
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002152779429344052,
+      "loss": 1.4483,
+      "step": 8964
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002152608524482902,
+      "loss": 0.4339,
+      "step": 8965
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021524376091712054,
+      "loss": 2.2735,
+      "step": 8966
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021522666834117003,
+      "loss": 0.5837,
+      "step": 8967
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002152095747207123,
+      "loss": 0.3568,
+      "step": 8968
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021519248005602107,
+      "loss": 0.0032,
+      "step": 8969
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021517538434737016,
+      "loss": 0.2561,
+      "step": 8970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021515828759503325,
+      "loss": 0.3216,
+      "step": 8971
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021514118979928414,
+      "loss": 0.4556,
+      "step": 8972
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021512409096039665,
+      "loss": 0.8539,
+      "step": 8973
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002151069910786446,
+      "loss": 0.0055,
+      "step": 8974
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021508989015430176,
+      "loss": 1.046,
+      "step": 8975
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021507278818764207,
+      "loss": 0.2066,
+      "step": 8976
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021505568517893927,
+      "loss": 0.435,
+      "step": 8977
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021503858112846734,
+      "loss": 0.8406,
+      "step": 8978
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021502147603650018,
+      "loss": 0.4208,
+      "step": 8979
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021500436990331163,
+      "loss": 0.6046,
+      "step": 8980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002149872627291756,
+      "loss": 0.1415,
+      "step": 8981
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021497015451436618,
+      "loss": 0.4458,
+      "step": 8982
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002149530452591572,
+      "loss": 0.3192,
+      "step": 8983
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002149359349638227,
+      "loss": 0.4329,
+      "step": 8984
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021491882362863665,
+      "loss": 0.3902,
+      "step": 8985
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021490171125387305,
+      "loss": 0.1105,
+      "step": 8986
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000214884597839806,
+      "loss": 0.4591,
+      "step": 8987
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021486748338670945,
+      "loss": 0.9493,
+      "step": 8988
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021485036789485747,
+      "loss": 0.2979,
+      "step": 8989
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021483325136452421,
+      "loss": 0.8238,
+      "step": 8990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021481613379598376,
+      "loss": 0.0041,
+      "step": 8991
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021479901518951016,
+      "loss": 0.5041,
+      "step": 8992
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021478189554537764,
+      "loss": 0.8948,
+      "step": 8993
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021476477486386024,
+      "loss": 0.2814,
+      "step": 8994
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002147476531452322,
+      "loss": 0.0033,
+      "step": 8995
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021473053038976761,
+      "loss": 0.3757,
+      "step": 8996
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021471340659774075,
+      "loss": 0.8917,
+      "step": 8997
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002146962817694258,
+      "loss": 0.9264,
+      "step": 8998
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000214679155905097,
+      "loss": 0.1641,
+      "step": 8999
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021466202900502862,
+      "loss": 0.3644,
+      "step": 9000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021464490106949484,
+      "loss": 1.1357,
+      "step": 9001
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021462777209877007,
+      "loss": 0.3453,
+      "step": 9002
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021461064209312843,
+      "loss": 0.6496,
+      "step": 9003
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002145935110528443,
+      "loss": 0.9006,
+      "step": 9004
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021457637897819213,
+      "loss": 0.0896,
+      "step": 9005
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021455924586944613,
+      "loss": 0.333,
+      "step": 9006
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021454211172688067,
+      "loss": 0.0968,
+      "step": 9007
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021452497655077022,
+      "loss": 0.0003,
+      "step": 9008
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021450784034138907,
+      "loss": 0.7602,
+      "step": 9009
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021449070309901162,
+      "loss": 0.0616,
+      "step": 9010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002144735648239124,
+      "loss": 0.2598,
+      "step": 9011
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002144564255163658,
+      "loss": 0.0002,
+      "step": 9012
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021443928517664627,
+      "loss": 0.5736,
+      "step": 9013
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021442214380502832,
+      "loss": 0.0066,
+      "step": 9014
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021440500140178638,
+      "loss": 1.4947,
+      "step": 9015
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021438785796719503,
+      "loss": 0.3512,
+      "step": 9016
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021437071350152877,
+      "loss": 0.3533,
+      "step": 9017
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021435356800506212,
+      "loss": 0.057,
+      "step": 9018
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021433642147806968,
+      "loss": 0.2083,
+      "step": 9019
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021431927392082597,
+      "loss": 0.0092,
+      "step": 9020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021430212533360567,
+      "loss": 0.1803,
+      "step": 9021
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002142849757166833,
+      "loss": 0.8858,
+      "step": 9022
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021426782507033353,
+      "loss": 0.659,
+      "step": 9023
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000214250673394831,
+      "loss": 0.6197,
+      "step": 9024
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021423352069045037,
+      "loss": 1.8838,
+      "step": 9025
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002142163669574663,
+      "loss": 0.8262,
+      "step": 9026
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021419921219615346,
+      "loss": 0.4514,
+      "step": 9027
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021418205640678662,
+      "loss": 0.7931,
+      "step": 9028
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021416489958964047,
+      "loss": 1.2194,
+      "step": 9029
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002141477417449897,
+      "loss": 0.6198,
+      "step": 9030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002141305828731092,
+      "loss": 0.2123,
+      "step": 9031
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002141134229742736,
+      "loss": 0.0029,
+      "step": 9032
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002140962620487578,
+      "loss": 1.4451,
+      "step": 9033
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021407910009683652,
+      "loss": 0.5844,
+      "step": 9034
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021406193711878463,
+      "loss": 0.4032,
+      "step": 9035
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021404477311487695,
+      "loss": 0.9313,
+      "step": 9036
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002140276080853884,
+      "loss": 0.2678,
+      "step": 9037
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021401044203059373,
+      "loss": 0.9572,
+      "step": 9038
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021399327495076793,
+      "loss": 1.2918,
+      "step": 9039
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002139761068461859,
+      "loss": 1.9935,
+      "step": 9040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021395893771712247,
+      "loss": 0.4647,
+      "step": 9041
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021394176756385266,
+      "loss": 1.1199,
+      "step": 9042
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021392459638665146,
+      "loss": 2.4825,
+      "step": 9043
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021390742418579375,
+      "loss": 0.3454,
+      "step": 9044
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021389025096155454,
+      "loss": 0.697,
+      "step": 9045
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021387307671420886,
+      "loss": 0.1293,
+      "step": 9046
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002138559014440317,
+      "loss": 0.376,
+      "step": 9047
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002138387251512981,
+      "loss": 0.5472,
+      "step": 9048
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021382154783628314,
+      "loss": 0.7305,
+      "step": 9049
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002138043694992619,
+      "loss": 0.6377,
+      "step": 9050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021378719014050943,
+      "loss": 2.4215,
+      "step": 9051
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021377000976030087,
+      "loss": 0.6503,
+      "step": 9052
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021375282835891125,
+      "loss": 0.2292,
+      "step": 9053
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021373564593661577,
+      "loss": 0.4012,
+      "step": 9054
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002137184624936896,
+      "loss": 0.4121,
+      "step": 9055
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021370127803040784,
+      "loss": 0.3092,
+      "step": 9056
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021368409254704577,
+      "loss": 0.3129,
+      "step": 9057
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021366690604387854,
+      "loss": 0.856,
+      "step": 9058
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002136497185211813,
+      "loss": 0.0059,
+      "step": 9059
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021363252997922937,
+      "loss": 1.2601,
+      "step": 9060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021361534041829794,
+      "loss": 0.1896,
+      "step": 9061
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021359814983866238,
+      "loss": 0.3544,
+      "step": 9062
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021358095824059787,
+      "loss": 0.6087,
+      "step": 9063
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021356376562437968,
+      "loss": 0.441,
+      "step": 9064
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021354657199028327,
+      "loss": 0.559,
+      "step": 9065
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002135293773385838,
+      "loss": 0.889,
+      "step": 9066
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021351218166955677,
+      "loss": 0.0031,
+      "step": 9067
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021349498498347743,
+      "loss": 0.2293,
+      "step": 9068
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002134777872806212,
+      "loss": 0.7364,
+      "step": 9069
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021346058856126345,
+      "loss": 0.9582,
+      "step": 9070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021344338882567962,
+      "loss": 0.2162,
+      "step": 9071
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021342618807414516,
+      "loss": 1.2114,
+      "step": 9072
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021340898630693546,
+      "loss": 1.027,
+      "step": 9073
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021339178352432602,
+      "loss": 0.0396,
+      "step": 9074
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021337457972659231,
+      "loss": 3.4205,
+      "step": 9075
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021335737491400981,
+      "loss": 0.7665,
+      "step": 9076
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002133401690868541,
+      "loss": 0.3617,
+      "step": 9077
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021332296224540058,
+      "loss": 0.5602,
+      "step": 9078
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021330575438992483,
+      "loss": 0.4802,
+      "step": 9079
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021328854552070246,
+      "loss": 0.1281,
+      "step": 9080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021327133563800904,
+      "loss": 0.8357,
+      "step": 9081
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002132541247421201,
+      "loss": 2.4047,
+      "step": 9082
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002132369128333113,
+      "loss": 0.3213,
+      "step": 9083
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021321969991185824,
+      "loss": 0.5001,
+      "step": 9084
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021320248597803658,
+      "loss": 0.7632,
+      "step": 9085
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021318527103212197,
+      "loss": 0.5498,
+      "step": 9086
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021316805507439002,
+      "loss": 0.561,
+      "step": 9087
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002131508381051165,
+      "loss": 0.4074,
+      "step": 9088
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021313362012457713,
+      "loss": 0.9874,
+      "step": 9089
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002131164011330475,
+      "loss": 0.2726,
+      "step": 9090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021309918113080343,
+      "loss": 0.4103,
+      "step": 9091
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021308196011812078,
+      "loss": 0.0124,
+      "step": 9092
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021306473809527514,
+      "loss": 0.6493,
+      "step": 9093
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021304751506254238,
+      "loss": 0.3836,
+      "step": 9094
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021303029102019827,
+      "loss": 0.3147,
+      "step": 9095
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021301306596851864,
+      "loss": 1.0122,
+      "step": 9096
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00021299583990777934,
+      "loss": 0.0163,
+      "step": 9097
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002129786128382562,
+      "loss": 0.781,
+      "step": 9098
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002129613847602251,
+      "loss": 0.3264,
+      "step": 9099
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021294415567396193,
+      "loss": 0.0902,
+      "step": 9100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021292692557974257,
+      "loss": 0.7039,
+      "step": 9101
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002129096944778429,
+      "loss": 0.3091,
+      "step": 9102
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021289246236853894,
+      "loss": 0.3412,
+      "step": 9103
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021287522925210654,
+      "loss": 0.3556,
+      "step": 9104
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002128579951288217,
+      "loss": 2.1956,
+      "step": 9105
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021284075999896044,
+      "loss": 0.4556,
+      "step": 9106
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002128235238627987,
+      "loss": 0.2194,
+      "step": 9107
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021280628672061254,
+      "loss": 0.0016,
+      "step": 9108
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021278904857267793,
+      "loss": 0.0099,
+      "step": 9109
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021277180941927095,
+      "loss": 0.3515,
+      "step": 9110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021275456926066763,
+      "loss": 0.0306,
+      "step": 9111
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002127373280971441,
+      "loss": 0.2177,
+      "step": 9112
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021272008592897637,
+      "loss": 0.9241,
+      "step": 9113
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002127028427564406,
+      "loss": 1.3506,
+      "step": 9114
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000212685598579813,
+      "loss": 1.2843,
+      "step": 9115
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021266835339936953,
+      "loss": 0.2953,
+      "step": 9116
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002126511072153864,
+      "loss": 1.213,
+      "step": 9117
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021263386002813987,
+      "loss": 0.2124,
+      "step": 9118
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021261661183790605,
+      "loss": 0.1859,
+      "step": 9119
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002125993626449612,
+      "loss": 0.0065,
+      "step": 9120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021258211244958147,
+      "loss": 0.8204,
+      "step": 9121
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021256486125204317,
+      "loss": 0.3842,
+      "step": 9122
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021254760905262246,
+      "loss": 0.6784,
+      "step": 9123
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021253035585159567,
+      "loss": 0.0071,
+      "step": 9124
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021251310164923913,
+      "loss": 0.0027,
+      "step": 9125
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021249584644582903,
+      "loss": 1.0836,
+      "step": 9126
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002124785902416418,
+      "loss": 0.3084,
+      "step": 9127
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021246133303695364,
+      "loss": 0.3485,
+      "step": 9128
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000212444074832041,
+      "loss": 0.0027,
+      "step": 9129
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021242681562718026,
+      "loss": 0.0044,
+      "step": 9130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021240955542264773,
+      "loss": 0.2899,
+      "step": 9131
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021239229421871983,
+      "loss": 0.0096,
+      "step": 9132
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000212375032015673,
+      "loss": 0.7339,
+      "step": 9133
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002123577688137836,
+      "loss": 0.0021,
+      "step": 9134
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021234050461332816,
+      "loss": 0.3097,
+      "step": 9135
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021232323941458301,
+      "loss": 0.7287,
+      "step": 9136
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002123059732178248,
+      "loss": 0.0077,
+      "step": 9137
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021228870602332998,
+      "loss": 0.011,
+      "step": 9138
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021227143783137488,
+      "loss": 0.8697,
+      "step": 9139
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021225416864223626,
+      "loss": 0.2438,
+      "step": 9140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021223689845619053,
+      "loss": 0.4059,
+      "step": 9141
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021221962727351428,
+      "loss": 0.0131,
+      "step": 9142
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021220235509448405,
+      "loss": 0.5369,
+      "step": 9143
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021218508191937652,
+      "loss": 0.8985,
+      "step": 9144
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021216780774846819,
+      "loss": 0.0004,
+      "step": 9145
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021215053258203572,
+      "loss": 0.6833,
+      "step": 9146
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021213325642035574,
+      "loss": 0.0549,
+      "step": 9147
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002121159792637049,
+      "loss": 0.0028,
+      "step": 9148
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002120987011123599,
+      "loss": 0.614,
+      "step": 9149
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002120814219665974,
+      "loss": 0.4839,
+      "step": 9150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021206414182669406,
+      "loss": 0.5064,
+      "step": 9151
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021204686069292665,
+      "loss": 0.5543,
+      "step": 9152
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021202957856557192,
+      "loss": 0.8864,
+      "step": 9153
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021201229544490652,
+      "loss": 0.6914,
+      "step": 9154
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021199501133120735,
+      "loss": 0.0007,
+      "step": 9155
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021197772622475106,
+      "loss": 0.5917,
+      "step": 9156
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021196044012581453,
+      "loss": 0.2317,
+      "step": 9157
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021194315303467453,
+      "loss": 0.2549,
+      "step": 9158
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021192586495160788,
+      "loss": 0.2528,
+      "step": 9159
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002119085758768915,
+      "loss": 1.14,
+      "step": 9160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021189128581080216,
+      "loss": 0.6738,
+      "step": 9161
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021187399475361673,
+      "loss": 0.9824,
+      "step": 9162
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021185670270561214,
+      "loss": 0.6082,
+      "step": 9163
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021183940966706534,
+      "loss": 0.7769,
+      "step": 9164
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021182211563825314,
+      "loss": 0.5359,
+      "step": 9165
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021180482061945255,
+      "loss": 1.2738,
+      "step": 9166
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021178752461094058,
+      "loss": 0.509,
+      "step": 9167
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021177022761299406,
+      "loss": 1.0427,
+      "step": 9168
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021175292962589005,
+      "loss": 0.4413,
+      "step": 9169
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021173563064990557,
+      "loss": 0.0852,
+      "step": 9170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021171833068531762,
+      "loss": 0.1263,
+      "step": 9171
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021170102973240326,
+      "loss": 0.8219,
+      "step": 9172
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021168372779143945,
+      "loss": 0.137,
+      "step": 9173
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021166642486270334,
+      "loss": 0.2536,
+      "step": 9174
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021164912094647203,
+      "loss": 0.0221,
+      "step": 9175
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002116318160430225,
+      "loss": 0.358,
+      "step": 9176
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021161451015263197,
+      "loss": 0.7368,
+      "step": 9177
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021159720327557755,
+      "loss": 0.3421,
+      "step": 9178
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021157989541213632,
+      "loss": 0.2476,
+      "step": 9179
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021156258656258552,
+      "loss": 0.381,
+      "step": 9180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021154527672720226,
+      "loss": 0.5243,
+      "step": 9181
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021152796590626375,
+      "loss": 0.0013,
+      "step": 9182
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021151065410004722,
+      "loss": 0.4645,
+      "step": 9183
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002114933413088299,
+      "loss": 0.0006,
+      "step": 9184
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021147602753288894,
+      "loss": 0.0702,
+      "step": 9185
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021145871277250175,
+      "loss": 0.2238,
+      "step": 9186
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021144139702794547,
+      "loss": 0.6434,
+      "step": 9187
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021142408029949743,
+      "loss": 0.2152,
+      "step": 9188
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002114067625874349,
+      "loss": 0.9778,
+      "step": 9189
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002113894438920353,
+      "loss": 0.5711,
+      "step": 9190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021137212421357585,
+      "loss": 0.0516,
+      "step": 9191
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021135480355233388,
+      "loss": 0.4661,
+      "step": 9192
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002113374819085869,
+      "loss": 0.7426,
+      "step": 9193
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021132015928261217,
+      "loss": 0.7083,
+      "step": 9194
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002113028356746871,
+      "loss": 0.3507,
+      "step": 9195
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021128551108508912,
+      "loss": 0.4457,
+      "step": 9196
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021126818551409567,
+      "loss": 0.0008,
+      "step": 9197
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021125085896198418,
+      "loss": 0.8676,
+      "step": 9198
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021123353142903214,
+      "loss": 0.5687,
+      "step": 9199
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021121620291551693,
+      "loss": 0.0066,
+      "step": 9200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021119887342171612,
+      "loss": 0.002,
+      "step": 9201
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021118154294790724,
+      "loss": 0.4833,
+      "step": 9202
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002111642114943677,
+      "loss": 0.416,
+      "step": 9203
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021114687906137518,
+      "loss": 2.0938,
+      "step": 9204
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002111295456492071,
+      "loss": 0.0769,
+      "step": 9205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021111221125814117,
+      "loss": 1.4835,
+      "step": 9206
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021109487588845483,
+      "loss": 0.1313,
+      "step": 9207
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021107753954042576,
+      "loss": 1.6448,
+      "step": 9208
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021106020221433152,
+      "loss": 0.4019,
+      "step": 9209
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021104286391044984,
+      "loss": 0.4035,
+      "step": 9210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002110255246290583,
+      "loss": 0.3521,
+      "step": 9211
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021100818437043452,
+      "loss": 0.3545,
+      "step": 9212
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002109908431348563,
+      "loss": 0.245,
+      "step": 9213
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002109735009226012,
+      "loss": 0.4867,
+      "step": 9214
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021095615773394696,
+      "loss": 0.6104,
+      "step": 9215
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021093881356917146,
+      "loss": 0.4958,
+      "step": 9216
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021092146842855222,
+      "loss": 0.1723,
+      "step": 9217
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021090412231236714,
+      "loss": 0.6923,
+      "step": 9218
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021088677522089388,
+      "loss": 0.7958,
+      "step": 9219
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021086942715441034,
+      "loss": 0.3909,
+      "step": 9220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021085207811319428,
+      "loss": 0.3049,
+      "step": 9221
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002108347280975235,
+      "loss": 0.3806,
+      "step": 9222
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021081737710767587,
+      "loss": 0.3332,
+      "step": 9223
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021080002514392923,
+      "loss": 0.6267,
+      "step": 9224
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002107826722065614,
+      "loss": 0.0049,
+      "step": 9225
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021076531829585028,
+      "loss": 0.308,
+      "step": 9226
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002107479634120738,
+      "loss": 1.4031,
+      "step": 9227
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021073060755550993,
+      "loss": 0.7058,
+      "step": 9228
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021071325072643645,
+      "loss": 0.002,
+      "step": 9229
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002106958929251314,
+      "loss": 0.2248,
+      "step": 9230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021067853415187267,
+      "loss": 0.5887,
+      "step": 9231
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021066117440693834,
+      "loss": 0.5663,
+      "step": 9232
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021064381369060632,
+      "loss": 0.5073,
+      "step": 9233
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021062645200315466,
+      "loss": 0.6783,
+      "step": 9234
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002106090893448613,
+      "loss": 1.4884,
+      "step": 9235
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021059172571600437,
+      "loss": 0.8443,
+      "step": 9236
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021057436111686188,
+      "loss": 0.1837,
+      "step": 9237
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002105569955477119,
+      "loss": 0.9456,
+      "step": 9238
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021053962900883258,
+      "loss": 0.1517,
+      "step": 9239
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002105222615005019,
+      "loss": 0.306,
+      "step": 9240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021050489302299803,
+      "loss": 1.2798,
+      "step": 9241
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021048752357659913,
+      "loss": 0.5211,
+      "step": 9242
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021047015316158333,
+      "loss": 0.114,
+      "step": 9243
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021045278177822874,
+      "loss": 0.5166,
+      "step": 9244
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002104354094268136,
+      "loss": 0.3508,
+      "step": 9245
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021041803610761606,
+      "loss": 0.0027,
+      "step": 9246
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002104006618209144,
+      "loss": 0.5069,
+      "step": 9247
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021038328656698674,
+      "loss": 0.3698,
+      "step": 9248
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021036591034611136,
+      "loss": 0.2176,
+      "step": 9249
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021034853315856654,
+      "loss": 0.6715,
+      "step": 9250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021033115500463057,
+      "loss": 0.3047,
+      "step": 9251
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021031377588458163,
+      "loss": 0.2508,
+      "step": 9252
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021029639579869813,
+      "loss": 0.26,
+      "step": 9253
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021027901474725836,
+      "loss": 0.7236,
+      "step": 9254
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021026163273054057,
+      "loss": 0.5549,
+      "step": 9255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002102442497488232,
+      "loss": 0.3592,
+      "step": 9256
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002102268658023846,
+      "loss": 0.876,
+      "step": 9257
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021020948089150312,
+      "loss": 0.6061,
+      "step": 9258
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002101920950164572,
+      "loss": 0.5437,
+      "step": 9259
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021017470817752518,
+      "loss": 0.0024,
+      "step": 9260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002101573203749855,
+      "loss": 0.0006,
+      "step": 9261
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021013993160911667,
+      "loss": 0.5604,
+      "step": 9262
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021012254188019708,
+      "loss": 1.419,
+      "step": 9263
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021010515118850514,
+      "loss": 0.9877,
+      "step": 9264
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021008775953431952,
+      "loss": 0.0007,
+      "step": 9265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021007036691791855,
+      "loss": 0.2373,
+      "step": 9266
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021005297333958077,
+      "loss": 0.1784,
+      "step": 9267
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002100355787995848,
+      "loss": 0.202,
+      "step": 9268
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021001818329820912,
+      "loss": 0.0004,
+      "step": 9269
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021000078683573235,
+      "loss": 2.3468,
+      "step": 9270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020998338941243295,
+      "loss": 0.1374,
+      "step": 9271
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020996599102858968,
+      "loss": 0.0022,
+      "step": 9272
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020994859168448098,
+      "loss": 0.2586,
+      "step": 9273
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002099311913803856,
+      "loss": 0.7401,
+      "step": 9274
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020991379011658213,
+      "loss": 0.2294,
+      "step": 9275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002098963878933492,
+      "loss": 0.2065,
+      "step": 9276
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020987898471096556,
+      "loss": 1.0898,
+      "step": 9277
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020986158056970984,
+      "loss": 0.5324,
+      "step": 9278
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020984417546986067,
+      "loss": 0.5918,
+      "step": 9279
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002098267694116969,
+      "loss": 0.5467,
+      "step": 9280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020980936239549722,
+      "loss": 0.2269,
+      "step": 9281
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020979195442154033,
+      "loss": 0.3135,
+      "step": 9282
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020977454549010507,
+      "loss": 0.5431,
+      "step": 9283
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020975713560147013,
+      "loss": 0.0003,
+      "step": 9284
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020973972475591435,
+      "loss": 0.2228,
+      "step": 9285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020972231295371657,
+      "loss": 0.3433,
+      "step": 9286
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020970490019515554,
+      "loss": 0.22,
+      "step": 9287
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020968748648051019,
+      "loss": 0.1715,
+      "step": 9288
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020967007181005927,
+      "loss": 0.4868,
+      "step": 9289
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020965265618408168,
+      "loss": 0.5396,
+      "step": 9290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002096352396028564,
+      "loss": 0.1267,
+      "step": 9291
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002096178220666622,
+      "loss": 0.2261,
+      "step": 9292
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020960040357577807,
+      "loss": 0.5811,
+      "step": 9293
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020958298413048293,
+      "loss": 0.6331,
+      "step": 9294
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020956556373105572,
+      "loss": 0.0087,
+      "step": 9295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020954814237777542,
+      "loss": 0.013,
+      "step": 9296
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020953072007092098,
+      "loss": 0.0005,
+      "step": 9297
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020951329681077137,
+      "loss": 0.5883,
+      "step": 9298
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020949587259760564,
+      "loss": 0.0345,
+      "step": 9299
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020947844743170283,
+      "loss": 0.2885,
+      "step": 9300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020946102131334194,
+      "loss": 0.0005,
+      "step": 9301
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020944359424280199,
+      "loss": 0.2031,
+      "step": 9302
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020942616622036215,
+      "loss": 0.0865,
+      "step": 9303
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020940873724630143,
+      "loss": 0.6815,
+      "step": 9304
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020939130732089891,
+      "loss": 0.4656,
+      "step": 9305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020937387644443374,
+      "loss": 0.4232,
+      "step": 9306
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002093564446171851,
+      "loss": 1.0914,
+      "step": 9307
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020933901183943203,
+      "loss": 0.0103,
+      "step": 9308
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020932157811145375,
+      "loss": 0.5171,
+      "step": 9309
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020930414343352943,
+      "loss": 1.8729,
+      "step": 9310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020928670780593825,
+      "loss": 0.2456,
+      "step": 9311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020926927122895948,
+      "loss": 0.0006,
+      "step": 9312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002092518337028722,
+      "loss": 0.0041,
+      "step": 9313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002092343952279558,
+      "loss": 0.99,
+      "step": 9314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020921695580448937,
+      "loss": 0.0025,
+      "step": 9315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020919951543275233,
+      "loss": 1.0552,
+      "step": 9316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020918207411302385,
+      "loss": 0.3274,
+      "step": 9317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020916463184558333,
+      "loss": 0.0003,
+      "step": 9318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020914718863071,
+      "loss": 0.3032,
+      "step": 9319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002091297444686832,
+      "loss": 0.4354,
+      "step": 9320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020911229935978228,
+      "loss": 0.1673,
+      "step": 9321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020909485330428663,
+      "loss": 1.5785,
+      "step": 9322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020907740630247558,
+      "loss": 1.1118,
+      "step": 9323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020905995835462853,
+      "loss": 0.0003,
+      "step": 9324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020904250946102485,
+      "loss": 0.5645,
+      "step": 9325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020902505962194404,
+      "loss": 0.3682,
+      "step": 9326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020900760883766546,
+      "loss": 0.678,
+      "step": 9327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020899015710846859,
+      "loss": 0.1062,
+      "step": 9328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002089727044346329,
+      "loss": 0.5255,
+      "step": 9329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020895525081643783,
+      "loss": 2.8633,
+      "step": 9330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020893779625416291,
+      "loss": 0.0024,
+      "step": 9331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020892034074808762,
+      "loss": 0.3125,
+      "step": 9332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020890288429849154,
+      "loss": 0.3525,
+      "step": 9333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020888542690565413,
+      "loss": 1.2269,
+      "step": 9334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000208867968569855,
+      "loss": 0.3844,
+      "step": 9335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020885050929137368,
+      "loss": 0.2174,
+      "step": 9336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002088330490704898,
+      "loss": 0.8116,
+      "step": 9337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020881558790748292,
+      "loss": 1.2049,
+      "step": 9338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020879812580263266,
+      "loss": 1.0383,
+      "step": 9339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002087806627562187,
+      "loss": 0.9491,
+      "step": 9340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020876319876852062,
+      "loss": 0.226,
+      "step": 9341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020874573383981813,
+      "loss": 0.2743,
+      "step": 9342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020872826797039082,
+      "loss": 0.5968,
+      "step": 9343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020871080116051846,
+      "loss": 0.0274,
+      "step": 9344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020869333341048076,
+      "loss": 0.419,
+      "step": 9345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020867586472055738,
+      "loss": 0.0003,
+      "step": 9346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00020865839509102806,
+      "loss": 0.7989,
+      "step": 9347
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020864092452217264,
+      "loss": 1.9853,
+      "step": 9348
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002086234530142708,
+      "loss": 0.0773,
+      "step": 9349
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020860598056760235,
+      "loss": 0.9304,
+      "step": 9350
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020858850718244703,
+      "loss": 0.1812,
+      "step": 9351
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020857103285908476,
+      "loss": 0.7201,
+      "step": 9352
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002085535575977953,
+      "loss": 0.7057,
+      "step": 9353
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020853608139885838,
+      "loss": 0.2738,
+      "step": 9354
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020851860426255412,
+      "loss": 0.4395,
+      "step": 9355
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020850112618916213,
+      "loss": 0.9214,
+      "step": 9356
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020848364717896244,
+      "loss": 0.1088,
+      "step": 9357
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002084661672322349,
+      "loss": 0.6383,
+      "step": 9358
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020844868634925946,
+      "loss": 0.2791,
+      "step": 9359
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000208431204530316,
+      "loss": 0.0065,
+      "step": 9360
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020841372177568456,
+      "loss": 0.4655,
+      "step": 9361
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002083962380856449,
+      "loss": 0.0859,
+      "step": 9362
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020837875346047725,
+      "loss": 0.3594,
+      "step": 9363
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020836126790046142,
+      "loss": 0.4906,
+      "step": 9364
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002083437814058775,
+      "loss": 2.5137,
+      "step": 9365
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020832629397700547,
+      "loss": 0.2823,
+      "step": 9366
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002083088056141254,
+      "loss": 0.5471,
+      "step": 9367
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020829131631751728,
+      "loss": 0.74,
+      "step": 9368
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020827382608746122,
+      "loss": 0.2772,
+      "step": 9369
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002082563349242373,
+      "loss": 0.3198,
+      "step": 9370
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020823884282812562,
+      "loss": 0.5536,
+      "step": 9371
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002082213497994063,
+      "loss": 0.4709,
+      "step": 9372
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020820385583835934,
+      "loss": 0.5406,
+      "step": 9373
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020818636094526504,
+      "loss": 0.5871,
+      "step": 9374
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020816886512040354,
+      "loss": 1.0389,
+      "step": 9375
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002081513683640549,
+      "loss": 0.3455,
+      "step": 9376
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020813387067649937,
+      "loss": 0.0047,
+      "step": 9377
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002081163720580172,
+      "loss": 0.703,
+      "step": 9378
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002080988725088885,
+      "loss": 0.5464,
+      "step": 9379
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020808137202939358,
+      "loss": 0.4074,
+      "step": 9380
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020806387061981266,
+      "loss": 1.2854,
+      "step": 9381
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000208046368280426,
+      "loss": 0.3573,
+      "step": 9382
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020802886501151383,
+      "loss": 0.5062,
+      "step": 9383
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020801136081335649,
+      "loss": 0.4281,
+      "step": 9384
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020799385568623428,
+      "loss": 0.7632,
+      "step": 9385
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020797634963042746,
+      "loss": 0.2285,
+      "step": 9386
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002079588426462165,
+      "loss": 0.5334,
+      "step": 9387
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020794133473388157,
+      "loss": 0.4968,
+      "step": 9388
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020792382589370317,
+      "loss": 0.0004,
+      "step": 9389
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020790631612596164,
+      "loss": 0.0013,
+      "step": 9390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020788880543093737,
+      "loss": 0.3573,
+      "step": 9391
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002078712938089107,
+      "loss": 0.1322,
+      "step": 9392
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020785378126016214,
+      "loss": 0.2358,
+      "step": 9393
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002078362677849721,
+      "loss": 0.0009,
+      "step": 9394
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020781875338362108,
+      "loss": 2.1485,
+      "step": 9395
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020780123805638946,
+      "loss": 0.0014,
+      "step": 9396
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020778372180355775,
+      "loss": 0.1749,
+      "step": 9397
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020776620462540648,
+      "loss": 0.0905,
+      "step": 9398
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002077486865222161,
+      "loss": 0.4094,
+      "step": 9399
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002077311674942672,
+      "loss": 0.8222,
+      "step": 9400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002077136475418403,
+      "loss": 0.2269,
+      "step": 9401
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020769612666521594,
+      "loss": 0.0042,
+      "step": 9402
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020767860486467466,
+      "loss": 0.0004,
+      "step": 9403
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020766108214049717,
+      "loss": 0.5762,
+      "step": 9404
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020764355849296393,
+      "loss": 0.2873,
+      "step": 9405
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020762603392235563,
+      "loss": 0.6332,
+      "step": 9406
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020760850842895288,
+      "loss": 0.3726,
+      "step": 9407
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020759098201303633,
+      "loss": 0.326,
+      "step": 9408
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020757345467488662,
+      "loss": 0.6414,
+      "step": 9409
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020755592641478449,
+      "loss": 2.4278,
+      "step": 9410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002075383972330105,
+      "loss": 0.3233,
+      "step": 9411
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020752086712984543,
+      "loss": 1.0372,
+      "step": 9412
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002075033361055701,
+      "loss": 0.1715,
+      "step": 9413
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020748580416046506,
+      "loss": 0.0047,
+      "step": 9414
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020746827129481117,
+      "loss": 1.5037,
+      "step": 9415
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020745073750888914,
+      "loss": 0.1819,
+      "step": 9416
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002074332028029798,
+      "loss": 0.2751,
+      "step": 9417
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020741566717736388,
+      "loss": 0.6835,
+      "step": 9418
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020739813063232223,
+      "loss": 0.2816,
+      "step": 9419
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020738059316813566,
+      "loss": 0.5158,
+      "step": 9420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000207363054785085,
+      "loss": 0.3504,
+      "step": 9421
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020734551548345115,
+      "loss": 0.4541,
+      "step": 9422
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002073279752635149,
+      "loss": 0.0009,
+      "step": 9423
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020731043412555716,
+      "loss": 0.533,
+      "step": 9424
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020729289206985885,
+      "loss": 0.3113,
+      "step": 9425
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002072753490967008,
+      "loss": 0.6952,
+      "step": 9426
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020725780520636407,
+      "loss": 0.1473,
+      "step": 9427
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002072402603991295,
+      "loss": 0.0075,
+      "step": 9428
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020722271467527804,
+      "loss": 1.4014,
+      "step": 9429
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002072051680350907,
+      "loss": 1.1384,
+      "step": 9430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020718762047884843,
+      "loss": 0.0591,
+      "step": 9431
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020717007200683226,
+      "loss": 1.6963,
+      "step": 9432
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020715252261932317,
+      "loss": 0.409,
+      "step": 9433
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020713497231660222,
+      "loss": 0.832,
+      "step": 9434
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020711742109895043,
+      "loss": 0.4683,
+      "step": 9435
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070998689666489,
+      "loss": 3.3341,
+      "step": 9436
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070823159199786,
+      "loss": 0.2854,
+      "step": 9437
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070647619592207,
+      "loss": 0.0032,
+      "step": 9438
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020704720708465633,
+      "loss": 1.3369,
+      "step": 9439
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002070296512965665,
+      "loss": 0.3379,
+      "step": 9440
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020701209459523243,
+      "loss": 0.2734,
+      "step": 9441
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020699453698093517,
+      "loss": 0.4544,
+      "step": 9442
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000206976978453956,
+      "loss": 1.95,
+      "step": 9443
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000206959419014576,
+      "loss": 0.0136,
+      "step": 9444
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020694185866307636,
+      "loss": 0.0125,
+      "step": 9445
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020692429739973837,
+      "loss": 0.8176,
+      "step": 9446
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020690673522484317,
+      "loss": 0.1536,
+      "step": 9447
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020688917213867195,
+      "loss": 2.1677,
+      "step": 9448
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020687160814150608,
+      "loss": 0.2728,
+      "step": 9449
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002068540432336268,
+      "loss": 1.5293,
+      "step": 9450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020683647741531527,
+      "loss": 0.4409,
+      "step": 9451
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020681891068685283,
+      "loss": 1.1875,
+      "step": 9452
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020680134304852088,
+      "loss": 2.4179,
+      "step": 9453
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020678377450060066,
+      "loss": 0.0949,
+      "step": 9454
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020676620504337347,
+      "loss": 0.7979,
+      "step": 9455
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020674863467712073,
+      "loss": 0.0035,
+      "step": 9456
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020673106340212376,
+      "loss": 0.4891,
+      "step": 9457
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020671349121866397,
+      "loss": 0.2974,
+      "step": 9458
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020669591812702274,
+      "loss": 0.0012,
+      "step": 9459
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020667834412748144,
+      "loss": 0.0029,
+      "step": 9460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002066607692203215,
+      "loss": 0.3522,
+      "step": 9461
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020664319340582446,
+      "loss": 0.9249,
+      "step": 9462
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020662561668427162,
+      "loss": 0.2476,
+      "step": 9463
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020660803905594448,
+      "loss": 0.5954,
+      "step": 9464
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020659046052112464,
+      "loss": 0.0249,
+      "step": 9465
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020657288108009343,
+      "loss": 1.8975,
+      "step": 9466
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020655530073313252,
+      "loss": 0.3758,
+      "step": 9467
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020653771948052325,
+      "loss": 0.3039,
+      "step": 9468
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020652013732254732,
+      "loss": 0.0022,
+      "step": 9469
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020650255425948616,
+      "loss": 0.4241,
+      "step": 9470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020648497029162143,
+      "loss": 0.2885,
+      "step": 9471
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020646738541923467,
+      "loss": 1.0821,
+      "step": 9472
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002064497996426075,
+      "loss": 1.2149,
+      "step": 9473
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002064322129620215,
+      "loss": 0.248,
+      "step": 9474
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020641462537775824,
+      "loss": 0.0079,
+      "step": 9475
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020639703689009952,
+      "loss": 0.4876,
+      "step": 9476
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020637944749932682,
+      "loss": 0.4629,
+      "step": 9477
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020636185720572192,
+      "loss": 1.607,
+      "step": 9478
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020634426600956644,
+      "loss": 0.3027,
+      "step": 9479
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020632667391114212,
+      "loss": 0.4438,
+      "step": 9480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020630908091073062,
+      "loss": 0.9127,
+      "step": 9481
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020629148700861371,
+      "loss": 0.8299,
+      "step": 9482
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020627389220507312,
+      "loss": 0.9252,
+      "step": 9483
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002062562965003906,
+      "loss": 0.3658,
+      "step": 9484
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020623869989484797,
+      "loss": 0.3129,
+      "step": 9485
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002062211023887269,
+      "loss": 0.8663,
+      "step": 9486
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020620350398230927,
+      "loss": 0.3037,
+      "step": 9487
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002061859046758769,
+      "loss": 0.4116,
+      "step": 9488
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020616830446971155,
+      "loss": 0.0031,
+      "step": 9489
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002061507033640951,
+      "loss": 0.4472,
+      "step": 9490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020613310135930943,
+      "loss": 0.3487,
+      "step": 9491
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002061154984556364,
+      "loss": 1.0985,
+      "step": 9492
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020609789465335786,
+      "loss": 0.2972,
+      "step": 9493
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002060802899527557,
+      "loss": 2.1041,
+      "step": 9494
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020606268435411187,
+      "loss": 0.0879,
+      "step": 9495
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020604507785770831,
+      "loss": 0.0628,
+      "step": 9496
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020602747046382695,
+      "loss": 3.3086,
+      "step": 9497
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002060098621727497,
+      "loss": 0.0031,
+      "step": 9498
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002059922529847586,
+      "loss": 0.5881,
+      "step": 9499
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002059746429001356,
+      "loss": 0.6696,
+      "step": 9500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020595703191916267,
+      "loss": 0.4462,
+      "step": 9501
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002059394200421219,
+      "loss": 0.1248,
+      "step": 9502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020592180726929524,
+      "loss": 0.4036,
+      "step": 9503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020590419360096477,
+      "loss": 0.0021,
+      "step": 9504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020588657903741254,
+      "loss": 0.939,
+      "step": 9505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002058689635789206,
+      "loss": 2.0246,
+      "step": 9506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020585134722577106,
+      "loss": 0.009,
+      "step": 9507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020583372997824606,
+      "loss": 0.4551,
+      "step": 9508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020581611183662763,
+      "loss": 0.0018,
+      "step": 9509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020579849280119794,
+      "loss": 0.5921,
+      "step": 9510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020578087287223917,
+      "loss": 0.7241,
+      "step": 9511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020576325205003338,
+      "loss": 0.7244,
+      "step": 9512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020574563033486277,
+      "loss": 0.9465,
+      "step": 9513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020572800772700967,
+      "loss": 0.6222,
+      "step": 9514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020571038422675605,
+      "loss": 0.6144,
+      "step": 9515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020569275983438426,
+      "loss": 0.002,
+      "step": 9516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020567513455017655,
+      "loss": 0.399,
+      "step": 9517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002056575083744151,
+      "loss": 0.4679,
+      "step": 9518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002056398813073822,
+      "loss": 0.2285,
+      "step": 9519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020562225334936005,
+      "loss": 0.3041,
+      "step": 9520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020560462450063102,
+      "loss": 0.9227,
+      "step": 9521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020558699476147735,
+      "loss": 0.0319,
+      "step": 9522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020556936413218145,
+      "loss": 0.5452,
+      "step": 9523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002055517326130255,
+      "loss": 0.3989,
+      "step": 9524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020553410020429202,
+      "loss": 0.0128,
+      "step": 9525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020551646690626323,
+      "loss": 0.0282,
+      "step": 9526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002054988327192215,
+      "loss": 0.0303,
+      "step": 9527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020548119764344928,
+      "loss": 0.3569,
+      "step": 9528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020546356167922895,
+      "loss": 0.6366,
+      "step": 9529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002054459248268429,
+      "loss": 0.9035,
+      "step": 9530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002054282870865736,
+      "loss": 0.1822,
+      "step": 9531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020541064845870348,
+      "loss": 0.6328,
+      "step": 9532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020539300894351496,
+      "loss": 0.6594,
+      "step": 9533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020537536854129058,
+      "loss": 0.9429,
+      "step": 9534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020535772725231268,
+      "loss": 0.3737,
+      "step": 9535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020534008507686394,
+      "loss": 1.158,
+      "step": 9536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020532244201522677,
+      "loss": 0.2146,
+      "step": 9537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020530479806768375,
+      "loss": 0.2531,
+      "step": 9538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020528715323451735,
+      "loss": 0.5065,
+      "step": 9539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020526950751601018,
+      "loss": 0.5022,
+      "step": 9540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002052518609124448,
+      "loss": 0.5908,
+      "step": 9541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020523421342410377,
+      "loss": 0.1597,
+      "step": 9542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002052165650512697,
+      "loss": 1.5467,
+      "step": 9543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020519891579422523,
+      "loss": 0.2864,
+      "step": 9544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020518126565325298,
+      "loss": 0.1875,
+      "step": 9545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020516361462863554,
+      "loss": 0.8984,
+      "step": 9546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020514596272065564,
+      "loss": 0.1326,
+      "step": 9547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020512830992959588,
+      "loss": 0.081,
+      "step": 9548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020511065625573898,
+      "loss": 0.3972,
+      "step": 9549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002050930016993676,
+      "loss": 1.5226,
+      "step": 9550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020507534626076454,
+      "loss": 0.0082,
+      "step": 9551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020505768994021246,
+      "loss": 0.2671,
+      "step": 9552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020504003273799407,
+      "loss": 0.3426,
+      "step": 9553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020502237465439216,
+      "loss": 0.7035,
+      "step": 9554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020500471568968954,
+      "loss": 0.4371,
+      "step": 9555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002049870558441689,
+      "loss": 0.5076,
+      "step": 9556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020496939511811308,
+      "loss": 0.7673,
+      "step": 9557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020495173351180494,
+      "loss": 1.1882,
+      "step": 9558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020493407102552722,
+      "loss": 0.3218,
+      "step": 9559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020491640765956285,
+      "loss": 1.1925,
+      "step": 9560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020489874341419459,
+      "loss": 0.7035,
+      "step": 9561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002048810782897053,
+      "loss": 3.8065,
+      "step": 9562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000204863412286378,
+      "loss": 0.8936,
+      "step": 9563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020484574540449542,
+      "loss": 0.7915,
+      "step": 9564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020482807764434055,
+      "loss": 0.3567,
+      "step": 9565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020481040900619633,
+      "loss": 1.9805,
+      "step": 9566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020479273949034564,
+      "loss": 0.2852,
+      "step": 9567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020477506909707147,
+      "loss": 0.0017,
+      "step": 9568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002047573978266568,
+      "loss": 0.3399,
+      "step": 9569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020473972567938456,
+      "loss": 0.0024,
+      "step": 9570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020472205265553776,
+      "loss": 0.879,
+      "step": 9571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020470437875539946,
+      "loss": 0.4751,
+      "step": 9572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020468670397925256,
+      "loss": 0.0059,
+      "step": 9573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002046690283273802,
+      "loss": 0.3701,
+      "step": 9574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002046513518000654,
+      "loss": 0.5657,
+      "step": 9575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020463367439759126,
+      "loss": 0.6732,
+      "step": 9576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020461599612024077,
+      "loss": 0.4121,
+      "step": 9577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020459831696829705,
+      "loss": 0.2079,
+      "step": 9578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020458063694204327,
+      "loss": 0.3166,
+      "step": 9579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020456295604176248,
+      "loss": 0.2934,
+      "step": 9580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020454527426773785,
+      "loss": 0.6405,
+      "step": 9581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020452759162025252,
+      "loss": 0.2296,
+      "step": 9582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020450990809958963,
+      "loss": 0.411,
+      "step": 9583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020449222370603234,
+      "loss": 0.7859,
+      "step": 9584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002044745384398639,
+      "loss": 0.3637,
+      "step": 9585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020445685230136753,
+      "loss": 0.3804,
+      "step": 9586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020443916529082633,
+      "loss": 0.5501,
+      "step": 9587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020442147740852362,
+      "loss": 0.7571,
+      "step": 9588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020440378865474264,
+      "loss": 0.1559,
+      "step": 9589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020438609902976662,
+      "loss": 0.4119,
+      "step": 9590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020436840853387884,
+      "loss": 1.7938,
+      "step": 9591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020435071716736259,
+      "loss": 0.6835,
+      "step": 9592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020433302493050117,
+      "loss": 0.7096,
+      "step": 9593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020431533182357792,
+      "loss": 0.4274,
+      "step": 9594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020429763784687613,
+      "loss": 0.3384,
+      "step": 9595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00020427994300067917,
+      "loss": 0.8201,
+      "step": 9596
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020426224728527036,
+      "loss": 0.5158,
+      "step": 9597
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020424455070093314,
+      "loss": 0.3597,
+      "step": 9598
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020422685324795082,
+      "loss": 0.3935,
+      "step": 9599
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020420915492660684,
+      "loss": 0.6771,
+      "step": 9600
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020419145573718461,
+      "loss": 1.5822,
+      "step": 9601
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020417375567996753,
+      "loss": 0.288,
+      "step": 9602
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020415605475523907,
+      "loss": 0.0021,
+      "step": 9603
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020413835296328268,
+      "loss": 0.2935,
+      "step": 9604
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020412065030438176,
+      "loss": 0.1668,
+      "step": 9605
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002041029467788199,
+      "loss": 0.0066,
+      "step": 9606
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020408524238688051,
+      "loss": 0.6077,
+      "step": 9607
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020406753712884717,
+      "loss": 0.2899,
+      "step": 9608
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020404983100500337,
+      "loss": 0.0064,
+      "step": 9609
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002040321240156326,
+      "loss": 0.934,
+      "step": 9610
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020401441616101846,
+      "loss": 0.4455,
+      "step": 9611
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020399670744144455,
+      "loss": 0.0031,
+      "step": 9612
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002039789978571944,
+      "loss": 0.2317,
+      "step": 9613
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020396128740855156,
+      "loss": 0.5406,
+      "step": 9614
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020394357609579975,
+      "loss": 0.362,
+      "step": 9615
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002039258639192225,
+      "loss": 0.2659,
+      "step": 9616
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020390815087910347,
+      "loss": 0.3092,
+      "step": 9617
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002038904369757263,
+      "loss": 0.3004,
+      "step": 9618
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002038727222093747,
+      "loss": 0.3902,
+      "step": 9619
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020385500658033225,
+      "loss": 0.1848,
+      "step": 9620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020383729008888275,
+      "loss": 0.6251,
+      "step": 9621
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020381957273530978,
+      "loss": 0.5019,
+      "step": 9622
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020380185451989718,
+      "loss": 0.0005,
+      "step": 9623
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020378413544292861,
+      "loss": 0.3665,
+      "step": 9624
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002037664155046878,
+      "loss": 0.1158,
+      "step": 9625
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020374869470545857,
+      "loss": 0.9019,
+      "step": 9626
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002037309730455247,
+      "loss": 0.2016,
+      "step": 9627
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020371325052516987,
+      "loss": 0.0017,
+      "step": 9628
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020369552714467797,
+      "loss": 3.02,
+      "step": 9629
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020367780290433276,
+      "loss": 0.5912,
+      "step": 9630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020366007780441814,
+      "loss": 0.2401,
+      "step": 9631
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020364235184521786,
+      "loss": 0.2487,
+      "step": 9632
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020362462502701583,
+      "loss": 1.8651,
+      "step": 9633
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020360689735009595,
+      "loss": 0.9066,
+      "step": 9634
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020358916881474206,
+      "loss": 0.4509,
+      "step": 9635
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020357143942123805,
+      "loss": 1.1571,
+      "step": 9636
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002035537091698678,
+      "loss": 0.928,
+      "step": 9637
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002035359780609153,
+      "loss": 0.5673,
+      "step": 9638
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020351824609466445,
+      "loss": 0.915,
+      "step": 9639
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002035005132713992,
+      "loss": 0.848,
+      "step": 9640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020348277959140353,
+      "loss": 0.4863,
+      "step": 9641
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020346504505496142,
+      "loss": 0.4466,
+      "step": 9642
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020344730966235684,
+      "loss": 0.2029,
+      "step": 9643
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002034295734138738,
+      "loss": 0.0092,
+      "step": 9644
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002034118363097964,
+      "loss": 2.2321,
+      "step": 9645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020339409835040852,
+      "loss": 0.5428,
+      "step": 9646
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002033763595359943,
+      "loss": 1.038,
+      "step": 9647
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020335861986683782,
+      "loss": 0.2949,
+      "step": 9648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020334087934322306,
+      "loss": 0.5396,
+      "step": 9649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020332313796543426,
+      "loss": 0.3204,
+      "step": 9650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020330539573375537,
+      "loss": 0.002,
+      "step": 9651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020328765264847055,
+      "loss": 0.6824,
+      "step": 9652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020326990870986404,
+      "loss": 0.3423,
+      "step": 9653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020325216391821983,
+      "loss": 1.1583,
+      "step": 9654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020323441827382213,
+      "loss": 0.0493,
+      "step": 9655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020321667177695512,
+      "loss": 0.4908,
+      "step": 9656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020319892442790296,
+      "loss": 0.3817,
+      "step": 9657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002031811762269499,
+      "loss": 0.5061,
+      "step": 9658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020316342717438014,
+      "loss": 0.2192,
+      "step": 9659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002031456772704778,
+      "loss": 0.4983,
+      "step": 9660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020312792651552727,
+      "loss": 0.7167,
+      "step": 9661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002031101749098127,
+      "loss": 0.355,
+      "step": 9662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020309242245361835,
+      "loss": 0.9346,
+      "step": 9663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020307466914722862,
+      "loss": 0.0144,
+      "step": 9664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020305691499092766,
+      "loss": 0.0023,
+      "step": 9665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020303915998499985,
+      "loss": 0.7867,
+      "step": 9666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002030214041297295,
+      "loss": 0.8064,
+      "step": 9667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002030036474254009,
+      "loss": 0.1042,
+      "step": 9668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020298588987229847,
+      "loss": 0.4888,
+      "step": 9669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020296813147070654,
+      "loss": 0.8096,
+      "step": 9670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020295037222090945,
+      "loss": 0.4301,
+      "step": 9671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020293261212319159,
+      "loss": 0.6575,
+      "step": 9672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020291485117783744,
+      "loss": 0.6631,
+      "step": 9673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020289708938513136,
+      "loss": 0.3374,
+      "step": 9674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020287932674535776,
+      "loss": 0.0011,
+      "step": 9675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020286156325880113,
+      "loss": 0.3993,
+      "step": 9676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020284379892574587,
+      "loss": 0.3497,
+      "step": 9677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020282603374647648,
+      "loss": 0.4656,
+      "step": 9678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020280826772127745,
+      "loss": 0.5019,
+      "step": 9679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020279050085043329,
+      "loss": 0.3545,
+      "step": 9680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020277273313422844,
+      "loss": 0.6921,
+      "step": 9681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002027549645729475,
+      "loss": 0.6049,
+      "step": 9682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020273719516687494,
+      "loss": 0.0009,
+      "step": 9683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020271942491629537,
+      "loss": 0.1431,
+      "step": 9684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002027016538214934,
+      "loss": 1.4245,
+      "step": 9685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020268388188275343,
+      "loss": 0.1621,
+      "step": 9686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020266610910036023,
+      "loss": 0.0024,
+      "step": 9687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002026483354745983,
+      "loss": 1.0172,
+      "step": 9688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002026305610057523,
+      "loss": 0.1571,
+      "step": 9689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020261278569410683,
+      "loss": 0.6285,
+      "step": 9690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020259500953994657,
+      "loss": 0.887,
+      "step": 9691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020257723254355617,
+      "loss": 0.9507,
+      "step": 9692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020255945470522033,
+      "loss": 0.6352,
+      "step": 9693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020254167602522366,
+      "loss": 0.5706,
+      "step": 9694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020252389650385095,
+      "loss": 2.0595,
+      "step": 9695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020250611614138685,
+      "loss": 0.5788,
+      "step": 9696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020248833493811602,
+      "loss": 0.0095,
+      "step": 9697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020247055289432335,
+      "loss": 0.6956,
+      "step": 9698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020245277001029353,
+      "loss": 1.0176,
+      "step": 9699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002024349862863113,
+      "loss": 0.0016,
+      "step": 9700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020241720172266145,
+      "loss": 0.3427,
+      "step": 9701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020239941631962883,
+      "loss": 0.0076,
+      "step": 9702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002023816300774982,
+      "loss": 2.0218,
+      "step": 9703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020236384299655432,
+      "loss": 0.3653,
+      "step": 9704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020234605507708209,
+      "loss": 0.0009,
+      "step": 9705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020232826631936637,
+      "loss": 0.3963,
+      "step": 9706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020231047672369197,
+      "loss": 0.4234,
+      "step": 9707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020229268629034386,
+      "loss": 0.4555,
+      "step": 9708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020227489501960678,
+      "loss": 1.2217,
+      "step": 9709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002022571029117657,
+      "loss": 2.6856,
+      "step": 9710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020223930996710567,
+      "loss": 0.0021,
+      "step": 9711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020222151618591136,
+      "loss": 0.6054,
+      "step": 9712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002022037215684679,
+      "loss": 0.4114,
+      "step": 9713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020218592611506017,
+      "loss": 0.2984,
+      "step": 9714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020216812982597315,
+      "loss": 0.2079,
+      "step": 9715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020215033270149185,
+      "loss": 0.899,
+      "step": 9716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002021325347419012,
+      "loss": 0.0036,
+      "step": 9717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020211473594748628,
+      "loss": 0.8391,
+      "step": 9718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002020969363185321,
+      "loss": 0.0083,
+      "step": 9719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002020791358553236,
+      "loss": 0.5378,
+      "step": 9720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020206133455814593,
+      "loss": 3.511,
+      "step": 9721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020204353242728415,
+      "loss": 0.019,
+      "step": 9722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020202572946302326,
+      "loss": 0.1182,
+      "step": 9723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020200792566564839,
+      "loss": 0.6108,
+      "step": 9724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020199012103544472,
+      "loss": 0.3852,
+      "step": 9725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002019723155726972,
+      "loss": 0.0922,
+      "step": 9726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020195450927769108,
+      "loss": 0.0737,
+      "step": 9727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020193670215071146,
+      "loss": 0.2866,
+      "step": 9728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020191889419204353,
+      "loss": 0.5029,
+      "step": 9729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002019010854019724,
+      "loss": 0.2497,
+      "step": 9730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002018832757807833,
+      "loss": 0.3023,
+      "step": 9731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020186546532876136,
+      "loss": 0.3691,
+      "step": 9732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020184765404619186,
+      "loss": 0.0136,
+      "step": 9733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020182984193336002,
+      "loss": 0.5327,
+      "step": 9734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020181202899055098,
+      "loss": 0.1469,
+      "step": 9735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020179421521805013,
+      "loss": 0.0157,
+      "step": 9736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020177640061614264,
+      "loss": 0.4468,
+      "step": 9737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020175858518511376,
+      "loss": 0.3757,
+      "step": 9738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020174076892524886,
+      "loss": 0.0134,
+      "step": 9739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020172295183683314,
+      "loss": 0.2626,
+      "step": 9740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000201705133920152,
+      "loss": 0.1186,
+      "step": 9741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020168731517549078,
+      "loss": 0.4205,
+      "step": 9742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020166949560313475,
+      "loss": 0.5157,
+      "step": 9743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020165167520336925,
+      "loss": 0.9728,
+      "step": 9744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020163385397647978,
+      "loss": 0.4624,
+      "step": 9745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020161603192275155,
+      "loss": 2.6878,
+      "step": 9746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020159820904247004,
+      "loss": 0.0366,
+      "step": 9747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002015803853359207,
+      "loss": 0.4819,
+      "step": 9748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020156256080338888,
+      "loss": 0.5856,
+      "step": 9749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020154473544516,
+      "loss": 0.0299,
+      "step": 9750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002015269092615196,
+      "loss": 0.8761,
+      "step": 9751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020150908225275303,
+      "loss": 1.0416,
+      "step": 9752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020149125441914584,
+      "loss": 0.2223,
+      "step": 9753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020147342576098346,
+      "loss": 0.5033,
+      "step": 9754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020145559627855142,
+      "loss": 1.4302,
+      "step": 9755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020143776597213525,
+      "loss": 0.3054,
+      "step": 9756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020141993484202046,
+      "loss": 0.3785,
+      "step": 9757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020140210288849256,
+      "loss": 0.3437,
+      "step": 9758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020138427011183712,
+      "loss": 0.505,
+      "step": 9759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020136643651233978,
+      "loss": 3.0439,
+      "step": 9760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020134860209028595,
+      "loss": 1.3182,
+      "step": 9761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020133076684596138,
+      "loss": 0.3311,
+      "step": 9762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020131293077965164,
+      "loss": 0.4037,
+      "step": 9763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020129509389164225,
+      "loss": 0.0249,
+      "step": 9764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020127725618221894,
+      "loss": 0.154,
+      "step": 9765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020125941765166737,
+      "loss": 0.9674,
+      "step": 9766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002012415783002731,
+      "loss": 0.2952,
+      "step": 9767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020122373812832189,
+      "loss": 0.8531,
+      "step": 9768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020120589713609939,
+      "loss": 0.4695,
+      "step": 9769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002011880553238913,
+      "loss": 1.4276,
+      "step": 9770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020117021269198333,
+      "loss": 0.3581,
+      "step": 9771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002011523692406612,
+      "loss": 2.3089,
+      "step": 9772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002011345249702106,
+      "loss": 0.0295,
+      "step": 9773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020111667988091734,
+      "loss": 0.9375,
+      "step": 9774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002010988339730672,
+      "loss": 0.8648,
+      "step": 9775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002010809872469459,
+      "loss": 2.888,
+      "step": 9776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020106313970283927,
+      "loss": 0.2012,
+      "step": 9777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002010452913410331,
+      "loss": 1.3774,
+      "step": 9778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020102744216181316,
+      "loss": 0.644,
+      "step": 9779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002010095921654653,
+      "loss": 0.0057,
+      "step": 9780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020099174135227544,
+      "loss": 0.2152,
+      "step": 9781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002009738897225293,
+      "loss": 0.6021,
+      "step": 9782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020095603727651292,
+      "loss": 1.1058,
+      "step": 9783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000200938184014512,
+      "loss": 2.4063,
+      "step": 9784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002009203299368125,
+      "loss": 0.0225,
+      "step": 9785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020090247504370035,
+      "loss": 0.2434,
+      "step": 9786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002008846193354615,
+      "loss": 0.0392,
+      "step": 9787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020086676281238182,
+      "loss": 0.0133,
+      "step": 9788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002008489054747472,
+      "loss": 0.356,
+      "step": 9789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020083104732284375,
+      "loss": 0.3503,
+      "step": 9790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020081318835695735,
+      "loss": 0.4438,
+      "step": 9791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000200795328577374,
+      "loss": 0.3043,
+      "step": 9792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020077746798437967,
+      "loss": 0.6878,
+      "step": 9793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020075960657826045,
+      "loss": 0.304,
+      "step": 9794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020074174435930227,
+      "loss": 0.0153,
+      "step": 9795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002007238813277912,
+      "loss": 0.4025,
+      "step": 9796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020070601748401333,
+      "loss": 0.5693,
+      "step": 9797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020068815282825466,
+      "loss": 0.3944,
+      "step": 9798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002006702873608013,
+      "loss": 0.2714,
+      "step": 9799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020065242108193934,
+      "loss": 0.3276,
+      "step": 9800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020063455399195487,
+      "loss": 0.0097,
+      "step": 9801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020061668609113402,
+      "loss": 1.4459,
+      "step": 9802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020059881737976288,
+      "loss": 0.3207,
+      "step": 9803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020058094785812763,
+      "loss": 0.4585,
+      "step": 9804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020056307752651443,
+      "loss": 0.8398,
+      "step": 9805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020054520638520943,
+      "loss": 0.3228,
+      "step": 9806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020052733443449874,
+      "loss": 0.4755,
+      "step": 9807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002005094616746687,
+      "loss": 0.3319,
+      "step": 9808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020049158810600545,
+      "loss": 1.2403,
+      "step": 9809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020047371372879512,
+      "loss": 0.4789,
+      "step": 9810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020045583854332404,
+      "loss": 0.5114,
+      "step": 9811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020043796254987849,
+      "loss": 1.122,
+      "step": 9812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020042008574874457,
+      "loss": 0.7811,
+      "step": 9813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002004022081402087,
+      "loss": 0.4179,
+      "step": 9814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020038432972455708,
+      "loss": 0.6862,
+      "step": 9815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020036645050207605,
+      "loss": 0.2282,
+      "step": 9816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020034857047305192,
+      "loss": 0.0068,
+      "step": 9817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000200330689637771,
+      "loss": 0.0957,
+      "step": 9818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002003128079965196,
+      "loss": 0.0667,
+      "step": 9819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020029492554958408,
+      "loss": 0.6008,
+      "step": 9820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020027704229725086,
+      "loss": 0.3556,
+      "step": 9821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020025915823980618,
+      "loss": 0.7825,
+      "step": 9822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002002412733775366,
+      "loss": 0.6634,
+      "step": 9823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002002233877107284,
+      "loss": 0.5014,
+      "step": 9824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020020550123966795,
+      "loss": 0.2984,
+      "step": 9825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020018761396464182,
+      "loss": 0.3245,
+      "step": 9826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020016972588593636,
+      "loss": 0.5887,
+      "step": 9827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020015183700383804,
+      "loss": 0.1237,
+      "step": 9828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002001339473186333,
+      "loss": 1.3789,
+      "step": 9829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020011605683060865,
+      "loss": 0.0758,
+      "step": 9830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020009816554005056,
+      "loss": 0.4874,
+      "step": 9831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020008027344724555,
+      "loss": 0.1785,
+      "step": 9832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020006238055248008,
+      "loss": 0.1783,
+      "step": 9833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002000444868560407,
+      "loss": 1.5118,
+      "step": 9834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00020002659235821406,
+      "loss": 0.3388,
+      "step": 9835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002000086970592866,
+      "loss": 0.2633,
+      "step": 9836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019999080095954482,
+      "loss": 0.0049,
+      "step": 9837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019997290405927547,
+      "loss": 0.3227,
+      "step": 9838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019995500635876506,
+      "loss": 0.2799,
+      "step": 9839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019993710785830014,
+      "loss": 1.1631,
+      "step": 9840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019991920855816743,
+      "loss": 0.4442,
+      "step": 9841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001999013084586535,
+      "loss": 0.2438,
+      "step": 9842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00019988340756004502,
+      "loss": 0.6391,
+      "step": 9843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001998655058626286,
+      "loss": 0.4123,
+      "step": 9844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000199847603366691,
+      "loss": 0.1554,
+      "step": 9845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019982970007251878,
+      "loss": 0.5393,
+      "step": 9846
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019981179598039877,
+      "loss": 0.3167,
+      "step": 9847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997938910906175,
+      "loss": 0.4491,
+      "step": 9848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997759854034619,
+      "loss": 0.1897,
+      "step": 9849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997580789192185,
+      "loss": 0.1567,
+      "step": 9850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997401716381742,
+      "loss": 0.6685,
+      "step": 9851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001997222635606157,
+      "loss": 0.0213,
+      "step": 9852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019970435468682978,
+      "loss": 0.6315,
+      "step": 9853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001996864450171032,
+      "loss": 0.2057,
+      "step": 9854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019966853455172278,
+      "loss": 0.5811,
+      "step": 9855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019965062329097528,
+      "loss": 0.011,
+      "step": 9856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019963271123514756,
+      "loss": 0.0049,
+      "step": 9857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019961479838452657,
+      "loss": 0.3936,
+      "step": 9858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001995968847393989,
+      "loss": 0.3259,
+      "step": 9859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001995789703000516,
+      "loss": 0.8298,
+      "step": 9860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019956105506677157,
+      "loss": 0.8332,
+      "step": 9861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019954313903984558,
+      "loss": 0.4184,
+      "step": 9862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019952522221956055,
+      "loss": 0.3983,
+      "step": 9863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019950730460620343,
+      "loss": 0.1766,
+      "step": 9864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019948938620006115,
+      "loss": 0.0044,
+      "step": 9865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001994714670014206,
+      "loss": 0.6955,
+      "step": 9866
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019945354701056872,
+      "loss": 0.3146,
+      "step": 9867
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019943562622779255,
+      "loss": 0.5757,
+      "step": 9868
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000199417704653379,
+      "loss": 0.7708,
+      "step": 9869
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019939978228761512,
+      "loss": 0.2804,
+      "step": 9870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001993818591307878,
+      "loss": 0.798,
+      "step": 9871
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019936393518318416,
+      "loss": 0.389,
+      "step": 9872
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019934601044509117,
+      "loss": 0.275,
+      "step": 9873
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019932808491679582,
+      "loss": 0.9521,
+      "step": 9874
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019931015859858534,
+      "loss": 0.5906,
+      "step": 9875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001992922314907466,
+      "loss": 0.2854,
+      "step": 9876
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019927430359356675,
+      "loss": 0.5564,
+      "step": 9877
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001992563749073329,
+      "loss": 0.5128,
+      "step": 9878
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001992384454323321,
+      "loss": 0.0201,
+      "step": 9879
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019922051516885154,
+      "loss": 1.6018,
+      "step": 9880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019920258411717826,
+      "loss": 0.4943,
+      "step": 9881
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019918465227759946,
+      "loss": 0.1844,
+      "step": 9882
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019916671965040223,
+      "loss": 0.7172,
+      "step": 9883
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019914878623587384,
+      "loss": 0.3599,
+      "step": 9884
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019913085203430135,
+      "loss": 1.2205,
+      "step": 9885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000199112917045972,
+      "loss": 0.5094,
+      "step": 9886
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019909498127117304,
+      "loss": 0.7307,
+      "step": 9887
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001990770447101916,
+      "loss": 0.5243,
+      "step": 9888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001990591073633149,
+      "loss": 0.001,
+      "step": 9889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001990411692308303,
+      "loss": 0.0636,
+      "step": 9890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019902323031302494,
+      "loss": 0.3562,
+      "step": 9891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019900529061018613,
+      "loss": 0.4745,
+      "step": 9892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001989873501226011,
+      "loss": 0.3608,
+      "step": 9893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001989694088505572,
+      "loss": 0.3786,
+      "step": 9894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019895146679434173,
+      "loss": 0.984,
+      "step": 9895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019893352395424201,
+      "loss": 0.2767,
+      "step": 9896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019891558033054529,
+      "loss": 0.2519,
+      "step": 9897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000198897635923539,
+      "loss": 0.6781,
+      "step": 9898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019887969073351043,
+      "loss": 0.0094,
+      "step": 9899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000198861744760747,
+      "loss": 0.489,
+      "step": 9900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000198843798005536,
+      "loss": 1.4873,
+      "step": 9901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019882585046816494,
+      "loss": 0.3671,
+      "step": 9902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019880790214892111,
+      "loss": 0.8951,
+      "step": 9903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019878995304809202,
+      "loss": 0.2022,
+      "step": 9904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019877200316596505,
+      "loss": 0.6344,
+      "step": 9905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001987540525028276,
+      "loss": 0.4242,
+      "step": 9906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019873610105896724,
+      "loss": 0.3882,
+      "step": 9907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019871814883467128,
+      "loss": 0.7044,
+      "step": 9908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019870019583022734,
+      "loss": 0.0071,
+      "step": 9909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019868224204592288,
+      "loss": 0.2722,
+      "step": 9910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019866428748204534,
+      "loss": 0.3313,
+      "step": 9911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019864633213888223,
+      "loss": 0.4247,
+      "step": 9912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019862837601672116,
+      "loss": 0.0214,
+      "step": 9913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019861041911584966,
+      "loss": 0.8795,
+      "step": 9914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001985924614365552,
+      "loss": 0.011,
+      "step": 9915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019857450297912542,
+      "loss": 0.3256,
+      "step": 9916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001985565437438479,
+      "loss": 0.2161,
+      "step": 9917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019853858373101015,
+      "loss": 0.3745,
+      "step": 9918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019852062294089986,
+      "loss": 2.4294,
+      "step": 9919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019850266137380458,
+      "loss": 0.0314,
+      "step": 9920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000198484699030012,
+      "loss": 0.3743,
+      "step": 9921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019846673590980977,
+      "loss": 0.2962,
+      "step": 9922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019844877201348541,
+      "loss": 0.4672,
+      "step": 9923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019843080734132678,
+      "loss": 0.5733,
+      "step": 9924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019841284189362138,
+      "loss": 0.8232,
+      "step": 9925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019839487567065698,
+      "loss": 1.0552,
+      "step": 9926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001983769086727213,
+      "loss": 0.656,
+      "step": 9927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019835894090010204,
+      "loss": 0.3403,
+      "step": 9928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001983409723530869,
+      "loss": 0.9038,
+      "step": 9929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019832300303196363,
+      "loss": 0.2913,
+      "step": 9930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019830503293702,
+      "loss": 0.6101,
+      "step": 9931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019828706206854375,
+      "loss": 0.194,
+      "step": 9932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001982690904268227,
+      "loss": 0.458,
+      "step": 9933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019825111801214458,
+      "loss": 0.2225,
+      "step": 9934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019823314482479725,
+      "loss": 0.8595,
+      "step": 9935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019821517086506847,
+      "loss": 0.2806,
+      "step": 9936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001981971961332461,
+      "loss": 0.3034,
+      "step": 9937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019817922062961797,
+      "loss": 0.514,
+      "step": 9938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001981612443544719,
+      "loss": 0.0009,
+      "step": 9939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019814326730809581,
+      "loss": 0.3787,
+      "step": 9940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019812528949077752,
+      "loss": 0.001,
+      "step": 9941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019810731090280496,
+      "loss": 0.6375,
+      "step": 9942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019808933154446601,
+      "loss": 0.0082,
+      "step": 9943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019807135141604863,
+      "loss": 0.0042,
+      "step": 9944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019805337051784067,
+      "loss": 0.6967,
+      "step": 9945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019803538885013007,
+      "loss": 0.0058,
+      "step": 9946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019801740641320483,
+      "loss": 0.2935,
+      "step": 9947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019799942320735292,
+      "loss": 0.0035,
+      "step": 9948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019798143923286228,
+      "loss": 0.0479,
+      "step": 9949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019796345449002087,
+      "loss": 0.4144,
+      "step": 9950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019794546897911673,
+      "loss": 0.2431,
+      "step": 9951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019792748270043787,
+      "loss": 0.303,
+      "step": 9952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019790949565427228,
+      "loss": 0.3212,
+      "step": 9953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019789150784090803,
+      "loss": 0.3501,
+      "step": 9954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019787351926063316,
+      "loss": 0.1493,
+      "step": 9955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019785552991373573,
+      "loss": 1.1586,
+      "step": 9956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001978375398005038,
+      "loss": 0.1605,
+      "step": 9957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001978195489212255,
+      "loss": 0.7064,
+      "step": 9958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019780155727618884,
+      "loss": 1.4122,
+      "step": 9959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019778356486568202,
+      "loss": 3.2208,
+      "step": 9960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019776557168999306,
+      "loss": 0.3367,
+      "step": 9961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019774757774941026,
+      "loss": 1.3165,
+      "step": 9962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019772958304422158,
+      "loss": 0.0016,
+      "step": 9963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019771158757471527,
+      "loss": 0.2722,
+      "step": 9964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001976935913411795,
+      "loss": 1.0722,
+      "step": 9965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019767559434390246,
+      "loss": 0.0631,
+      "step": 9966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001976575965831723,
+      "loss": 1.0789,
+      "step": 9967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019763959805927728,
+      "loss": 0.8122,
+      "step": 9968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019762159877250555,
+      "loss": 0.3345,
+      "step": 9969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019760359872314541,
+      "loss": 0.123,
+      "step": 9970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019758559791148512,
+      "loss": 0.5866,
+      "step": 9971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019756759633781284,
+      "loss": 0.3425,
+      "step": 9972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019754959400241693,
+      "loss": 0.3251,
+      "step": 9973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019753159090558561,
+      "loss": 0.0608,
+      "step": 9974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001975135870476072,
+      "loss": 0.3775,
+      "step": 9975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019749558242877004,
+      "loss": 0.0114,
+      "step": 9976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019747757704936236,
+      "loss": 2.3771,
+      "step": 9977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019745957090967255,
+      "loss": 0.7653,
+      "step": 9978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019744156400998894,
+      "loss": 0.0092,
+      "step": 9979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019742355635059988,
+      "loss": 0.035,
+      "step": 9980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019740554793179374,
+      "loss": 0.0058,
+      "step": 9981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019738753875385892,
+      "loss": 0.3105,
+      "step": 9982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019736952881708374,
+      "loss": 1.1187,
+      "step": 9983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019735151812175667,
+      "loss": 0.808,
+      "step": 9984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001973335066681661,
+      "loss": 0.0036,
+      "step": 9985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019731549445660046,
+      "loss": 0.0079,
+      "step": 9986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001972974814873482,
+      "loss": 0.5507,
+      "step": 9987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019727946776069774,
+      "loss": 0.0109,
+      "step": 9988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019726145327693756,
+      "loss": 0.451,
+      "step": 9989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019724343803635615,
+      "loss": 0.0033,
+      "step": 9990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019722542203924197,
+      "loss": 0.0049,
+      "step": 9991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019720740528588355,
+      "loss": 1.9435,
+      "step": 9992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019718938777656933,
+      "loss": 0.8925,
+      "step": 9993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019717136951158794,
+      "loss": 0.3121,
+      "step": 9994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019715335049122785,
+      "loss": 0.0445,
+      "step": 9995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019713533071577757,
+      "loss": 1.906,
+      "step": 9996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001971173101855258,
+      "loss": 0.3503,
+      "step": 9997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019709928890076096,
+      "loss": 0.524,
+      "step": 9998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001970812668617717,
+      "loss": 0.4916,
+      "step": 9999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019706324406884666,
+      "loss": 0.278,
+      "step": 10000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019704522052227435,
+      "loss": 0.3822,
+      "step": 10001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019702719622234342,
+      "loss": 0.0188,
+      "step": 10002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019700917116934256,
+      "loss": 0.3078,
+      "step": 10003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019699114536356034,
+      "loss": 0.3775,
+      "step": 10004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019697311880528546,
+      "loss": 0.9693,
+      "step": 10005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019695509149480662,
+      "loss": 0.0072,
+      "step": 10006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001969370634324124,
+      "loss": 1.966,
+      "step": 10007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001969190346183916,
+      "loss": 0.0074,
+      "step": 10008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019690100505303287,
+      "loss": 0.4529,
+      "step": 10009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001968829747366249,
+      "loss": 0.3094,
+      "step": 10010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019686494366945648,
+      "loss": 0.2204,
+      "step": 10011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001968469118518163,
+      "loss": 0.5246,
+      "step": 10012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019682887928399316,
+      "loss": 0.2834,
+      "step": 10013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001968108459662758,
+      "loss": 0.8908,
+      "step": 10014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019679281189895298,
+      "loss": 0.2918,
+      "step": 10015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001967747770823135,
+      "loss": 0.3483,
+      "step": 10016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019675674151664618,
+      "loss": 0.0171,
+      "step": 10017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019673870520223974,
+      "loss": 0.133,
+      "step": 10018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019672066813938318,
+      "loss": 0.6055,
+      "step": 10019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019670263032836523,
+      "loss": 0.246,
+      "step": 10020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019668459176947468,
+      "loss": 0.2972,
+      "step": 10021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001966665524630005,
+      "loss": 0.464,
+      "step": 10022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019664851240923153,
+      "loss": 0.8457,
+      "step": 10023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019663047160845663,
+      "loss": 0.5884,
+      "step": 10024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019661243006096469,
+      "loss": 0.432,
+      "step": 10025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019659438776704468,
+      "loss": 0.3502,
+      "step": 10026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019657634472698544,
+      "loss": 0.5002,
+      "step": 10027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019655830094107596,
+      "loss": 0.0144,
+      "step": 10028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019654025640960512,
+      "loss": 1.2998,
+      "step": 10029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000196522211132862,
+      "loss": 0.7603,
+      "step": 10030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019650416511113544,
+      "loss": 0.4468,
+      "step": 10031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019648611834471448,
+      "loss": 0.3678,
+      "step": 10032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019646807083388806,
+      "loss": 4.8954,
+      "step": 10033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001964500225789453,
+      "loss": 0.4632,
+      "step": 10034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019643197358017511,
+      "loss": 0.1745,
+      "step": 10035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019641392383786655,
+      "loss": 0.8584,
+      "step": 10036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019639587335230864,
+      "loss": 0.0031,
+      "step": 10037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019637782212379045,
+      "loss": 0.2076,
+      "step": 10038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019635977015260107,
+      "loss": 0.4194,
+      "step": 10039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001963417174390295,
+      "loss": 0.0087,
+      "step": 10040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001963236639833649,
+      "loss": 0.4924,
+      "step": 10041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019630560978589636,
+      "loss": 0.0064,
+      "step": 10042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019628755484691298,
+      "loss": 0.3708,
+      "step": 10043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019626949916670382,
+      "loss": 0.1884,
+      "step": 10044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001962514427455581,
+      "loss": 2.7561,
+      "step": 10045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019623338558376495,
+      "loss": 0.8762,
+      "step": 10046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019621532768161352,
+      "loss": 0.8672,
+      "step": 10047
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019619726903939293,
+      "loss": 0.3106,
+      "step": 10048
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001961792096573925,
+      "loss": 0.0052,
+      "step": 10049
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019616114953590127,
+      "loss": 3.6192,
+      "step": 10050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001961430886752085,
+      "loss": 1.1958,
+      "step": 10051
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019612502707560343,
+      "loss": 0.5643,
+      "step": 10052
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019610696473737528,
+      "loss": 0.4997,
+      "step": 10053
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001960889016608133,
+      "loss": 0.5004,
+      "step": 10054
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019607083784620673,
+      "loss": 0.3352,
+      "step": 10055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001960527732938448,
+      "loss": 0.7331,
+      "step": 10056
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001960347080040168,
+      "loss": 0.1627,
+      "step": 10057
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019601664197701213,
+      "loss": 0.0241,
+      "step": 10058
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019599857521311994,
+      "loss": 1.0207,
+      "step": 10059
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019598050771262962,
+      "loss": 0.5612,
+      "step": 10060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019596243947583045,
+      "loss": 0.2697,
+      "step": 10061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019594437050301182,
+      "loss": 0.529,
+      "step": 10062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019592630079446306,
+      "loss": 0.3047,
+      "step": 10063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019590823035047346,
+      "loss": 0.0111,
+      "step": 10064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001958901591713325,
+      "loss": 0.6289,
+      "step": 10065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019587208725732955,
+      "loss": 0.7084,
+      "step": 10066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001958540146087539,
+      "loss": 0.0039,
+      "step": 10067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019583594122589502,
+      "loss": 0.2462,
+      "step": 10068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019581786710904238,
+      "loss": 0.8627,
+      "step": 10069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019579979225848527,
+      "loss": 3.3193,
+      "step": 10070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019578171667451332,
+      "loss": 1.7227,
+      "step": 10071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019576364035741586,
+      "loss": 0.5221,
+      "step": 10072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019574556330748236,
+      "loss": 1.9904,
+      "step": 10073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019572748552500231,
+      "loss": 0.5436,
+      "step": 10074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019570940701026527,
+      "loss": 0.2648,
+      "step": 10075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019569132776356066,
+      "loss": 0.0675,
+      "step": 10076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019567324778517795,
+      "loss": 0.0165,
+      "step": 10077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019565516707540677,
+      "loss": 0.7927,
+      "step": 10078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001956370856345366,
+      "loss": 0.5897,
+      "step": 10079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019561900346285694,
+      "loss": 0.0273,
+      "step": 10080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019560092056065753,
+      "loss": 0.6859,
+      "step": 10081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001955828369282277,
+      "loss": 0.9417,
+      "step": 10082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019556475256585717,
+      "loss": 0.8288,
+      "step": 10083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019554666747383554,
+      "loss": 0.1923,
+      "step": 10084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019552858165245237,
+      "loss": 0.0049,
+      "step": 10085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001955104951019973,
+      "loss": 0.5299,
+      "step": 10086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019549240782275996,
+      "loss": 0.4983,
+      "step": 10087
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019547431981502996,
+      "loss": 1.0788,
+      "step": 10088
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000195456231079097,
+      "loss": 0.013,
+      "step": 10089
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019543814161525068,
+      "loss": 0.0062,
+      "step": 10090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019542005142378076,
+      "loss": 0.3012,
+      "step": 10091
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019540196050497687,
+      "loss": 0.2622,
+      "step": 10092
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019538386885912872,
+      "loss": 0.6219,
+      "step": 10093
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019536577648652598,
+      "loss": 0.1924,
+      "step": 10094
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00019534768338745848,
+      "loss": 0.7321,
+      "step": 10095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001953295895622159,
+      "loss": 0.3262,
+      "step": 10096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019531149501108788,
+      "loss": 0.6851,
+      "step": 10097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019529339973436434,
+      "loss": 2.3533,
+      "step": 10098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019527530373233498,
+      "loss": 0.0104,
+      "step": 10099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019525720700528954,
+      "loss": 0.3167,
+      "step": 10100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001952391095535179,
+      "loss": 0.3894,
+      "step": 10101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019522101137730977,
+      "loss": 0.144,
+      "step": 10102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000195202912476955,
+      "loss": 0.757,
+      "step": 10103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019518481285274347,
+      "loss": 0.5734,
+      "step": 10104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019516671250496498,
+      "loss": 0.7229,
+      "step": 10105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019514861143390933,
+      "loss": 0.4366,
+      "step": 10106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019513050963986648,
+      "loss": 0.7839,
+      "step": 10107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001951124071231262,
+      "loss": 0.6552,
+      "step": 10108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019509430388397849,
+      "loss": 0.4587,
+      "step": 10109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001950761999227131,
+      "loss": 0.7134,
+      "step": 10110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019505809523962005,
+      "loss": 0.0557,
+      "step": 10111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019503998983498924,
+      "loss": 0.4992,
+      "step": 10112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019502188370911058,
+      "loss": 0.0132,
+      "step": 10113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019500377686227402,
+      "loss": 0.4228,
+      "step": 10114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019498566929476954,
+      "loss": 0.3147,
+      "step": 10115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019496756100688701,
+      "loss": 0.4338,
+      "step": 10116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019494945199891655,
+      "loss": 0.0085,
+      "step": 10117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019493134227114805,
+      "loss": 0.6633,
+      "step": 10118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019491323182387152,
+      "loss": 0.3447,
+      "step": 10119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194895120657377,
+      "loss": 0.2175,
+      "step": 10120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019487700877195456,
+      "loss": 0.6588,
+      "step": 10121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019485889616789408,
+      "loss": 0.004,
+      "step": 10122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001948407828454857,
+      "loss": 0.207,
+      "step": 10123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019482266880501954,
+      "loss": 4.3148,
+      "step": 10124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001948045540467856,
+      "loss": 0.0492,
+      "step": 10125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019478643857107398,
+      "loss": 0.2009,
+      "step": 10126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019476832237817472,
+      "loss": 1.2412,
+      "step": 10127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194750205468378,
+      "loss": 0.3836,
+      "step": 10128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019473208784197385,
+      "loss": 0.8216,
+      "step": 10129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019471396949925253,
+      "loss": 0.0453,
+      "step": 10130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194695850440504,
+      "loss": 0.1713,
+      "step": 10131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019467773066601855,
+      "loss": 0.0349,
+      "step": 10132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001946596101760863,
+      "loss": 1.5185,
+      "step": 10133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019464148897099744,
+      "loss": 0.4264,
+      "step": 10134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019462336705104212,
+      "loss": 0.5312,
+      "step": 10135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019460524441651053,
+      "loss": 0.2368,
+      "step": 10136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001945871210676929,
+      "loss": 0.302,
+      "step": 10137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019456899700487942,
+      "loss": 0.3762,
+      "step": 10138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019455087222836035,
+      "loss": 1.0077,
+      "step": 10139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019453274673842592,
+      "loss": 0.3641,
+      "step": 10140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001945146205353664,
+      "loss": 0.5722,
+      "step": 10141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019449649361947201,
+      "loss": 0.554,
+      "step": 10142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019447836599103312,
+      "loss": 0.8956,
+      "step": 10143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019446023765033992,
+      "loss": 0.2633,
+      "step": 10144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019444210859768273,
+      "loss": 0.2442,
+      "step": 10145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019442397883335185,
+      "loss": 0.5417,
+      "step": 10146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019440584835763767,
+      "loss": 0.6232,
+      "step": 10147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019438771717083042,
+      "loss": 0.7403,
+      "step": 10148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001943695852732205,
+      "loss": 0.5726,
+      "step": 10149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001943514526650983,
+      "loss": 0.2371,
+      "step": 10150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019433331934675413,
+      "loss": 0.5468,
+      "step": 10151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019431518531847836,
+      "loss": 0.2793,
+      "step": 10152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019429705058056145,
+      "loss": 0.626,
+      "step": 10153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019427891513329373,
+      "loss": 0.4377,
+      "step": 10154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019426077897696564,
+      "loss": 0.3889,
+      "step": 10155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019424264211186763,
+      "loss": 0.4417,
+      "step": 10156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019422450453829006,
+      "loss": 0.4295,
+      "step": 10157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019420636625652342,
+      "loss": 0.748,
+      "step": 10158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019418822726685822,
+      "loss": 6.0747,
+      "step": 10159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019417008756958482,
+      "loss": 0.0042,
+      "step": 10160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019415194716499378,
+      "loss": 0.5626,
+      "step": 10161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019413380605337558,
+      "loss": 0.4108,
+      "step": 10162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001941156642350207,
+      "loss": 0.6667,
+      "step": 10163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019409752171021968,
+      "loss": 0.5798,
+      "step": 10164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000194079378479263,
+      "loss": 0.3224,
+      "step": 10165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019406123454244128,
+      "loss": 0.7788,
+      "step": 10166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019404308990004495,
+      "loss": 0.6013,
+      "step": 10167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019402494455236468,
+      "loss": 0.3051,
+      "step": 10168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019400679849969097,
+      "loss": 0.0126,
+      "step": 10169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001939886517423145,
+      "loss": 0.1359,
+      "step": 10170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019397050428052572,
+      "loss": 0.2884,
+      "step": 10171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019395235611461528,
+      "loss": 0.003,
+      "step": 10172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001939342072448739,
+      "loss": 2.4266,
+      "step": 10173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019391605767159206,
+      "loss": 0.3629,
+      "step": 10174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001938979073950605,
+      "loss": 1.0506,
+      "step": 10175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001938797564155698,
+      "loss": 0.2989,
+      "step": 10176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019386160473341068,
+      "loss": 3.1247,
+      "step": 10177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001938434523488738,
+      "loss": 0.5048,
+      "step": 10178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019382529926224984,
+      "loss": 1.1619,
+      "step": 10179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019380714547382943,
+      "loss": 0.345,
+      "step": 10180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019378899098390336,
+      "loss": 0.4154,
+      "step": 10181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019377083579276232,
+      "loss": 0.0096,
+      "step": 10182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019375267990069705,
+      "loss": 0.5007,
+      "step": 10183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019373452330799818,
+      "loss": 0.5087,
+      "step": 10184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019371636601495668,
+      "loss": 1.5758,
+      "step": 10185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019369820802186315,
+      "loss": 0.5388,
+      "step": 10186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019368004932900838,
+      "loss": 0.5404,
+      "step": 10187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019366188993668319,
+      "loss": 0.2824,
+      "step": 10188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019364372984517834,
+      "loss": 0.2064,
+      "step": 10189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001936255690547847,
+      "loss": 1.049,
+      "step": 10190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000193607407565793,
+      "loss": 0.0019,
+      "step": 10191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019358924537849414,
+      "loss": 0.309,
+      "step": 10192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019357108249317892,
+      "loss": 0.6444,
+      "step": 10193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019355291891013824,
+      "loss": 0.2538,
+      "step": 10194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019353475462966283,
+      "loss": 0.0189,
+      "step": 10195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001935165896520438,
+      "loss": 0.0024,
+      "step": 10196
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019349842397757178,
+      "loss": 2.6485,
+      "step": 10197
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019348025760653782,
+      "loss": 0.0036,
+      "step": 10198
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001934620905392328,
+      "loss": 0.6729,
+      "step": 10199
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019344392277594762,
+      "loss": 0.4236,
+      "step": 10200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019342575431697325,
+      "loss": 0.2683,
+      "step": 10201
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019340758516260054,
+      "loss": 0.2444,
+      "step": 10202
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933894153131205,
+      "loss": 0.2788,
+      "step": 10203
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019337124476882415,
+      "loss": 0.1543,
+      "step": 10204
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933530735300024,
+      "loss": 0.1603,
+      "step": 10205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933349015969462,
+      "loss": 1.448,
+      "step": 10206
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001933167289699466,
+      "loss": 0.7955,
+      "step": 10207
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001932985556492946,
+      "loss": 0.8667,
+      "step": 10208
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019328038163528123,
+      "loss": 0.2928,
+      "step": 10209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019326220692819743,
+      "loss": 1.0118,
+      "step": 10210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019324403152833438,
+      "loss": 0.8623,
+      "step": 10211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019322585543598306,
+      "loss": 0.742,
+      "step": 10212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019320767865143454,
+      "loss": 0.205,
+      "step": 10213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019318950117497988,
+      "loss": 0.2952,
+      "step": 10214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019317132300691018,
+      "loss": 0.3517,
+      "step": 10215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019315314414751654,
+      "loss": 2.3771,
+      "step": 10216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019313496459709005,
+      "loss": 0.7316,
+      "step": 10217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001931167843559218,
+      "loss": 0.7733,
+      "step": 10218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019309860342430307,
+      "loss": 0.197,
+      "step": 10219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001930804218025248,
+      "loss": 0.3705,
+      "step": 10220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019306223949087822,
+      "loss": 0.7435,
+      "step": 10221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019304405648965455,
+      "loss": 0.2448,
+      "step": 10222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001930258727991449,
+      "loss": 0.5696,
+      "step": 10223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019300768841964044,
+      "loss": 0.0427,
+      "step": 10224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019298950335143244,
+      "loss": 0.2739,
+      "step": 10225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019297131759481203,
+      "loss": 0.0053,
+      "step": 10226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019295313115007048,
+      "loss": 0.2791,
+      "step": 10227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000192934944017499,
+      "loss": 0.345,
+      "step": 10228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001929167561973888,
+      "loss": 0.3998,
+      "step": 10229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019289856769003115,
+      "loss": 0.1719,
+      "step": 10230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019288037849571742,
+      "loss": 0.2546,
+      "step": 10231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001928621886147387,
+      "loss": 0.0103,
+      "step": 10232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019284399804738632,
+      "loss": 0.0592,
+      "step": 10233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019282580679395167,
+      "loss": 0.0404,
+      "step": 10234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019280761485472602,
+      "loss": 1.9289,
+      "step": 10235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019278942223000057,
+      "loss": 1.2985,
+      "step": 10236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019277122892006685,
+      "loss": 0.4888,
+      "step": 10237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000192753034925216,
+      "loss": 0.4216,
+      "step": 10238
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001927348402457395,
+      "loss": 0.738,
+      "step": 10239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019271664488192868,
+      "loss": 0.4823,
+      "step": 10240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019269844883407488,
+      "loss": 0.3533,
+      "step": 10241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019268025210246952,
+      "loss": 0.6814,
+      "step": 10242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019266205468740396,
+      "loss": 0.7545,
+      "step": 10243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019264385658916962,
+      "loss": 0.6435,
+      "step": 10244
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019262565780805795,
+      "loss": 0.6083,
+      "step": 10245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019260745834436034,
+      "loss": 0.241,
+      "step": 10246
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019258925819836816,
+      "loss": 0.3632,
+      "step": 10247
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000192571057370373,
+      "loss": 1.6632,
+      "step": 10248
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019255285586066621,
+      "loss": 0.5472,
+      "step": 10249
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001925346536695393,
+      "loss": 0.2271,
+      "step": 10250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019251645079728375,
+      "loss": 0.3805,
+      "step": 10251
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019249824724419103,
+      "loss": 0.4974,
+      "step": 10252
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001924800430105527,
+      "loss": 0.1372,
+      "step": 10253
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019246183809666024,
+      "loss": 0.3907,
+      "step": 10254
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001924436325028051,
+      "loss": 1.3297,
+      "step": 10255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019242542622927895,
+      "loss": 1.3975,
+      "step": 10256
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019240721927637328,
+      "loss": 0.2769,
+      "step": 10257
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001923890116443796,
+      "loss": 0.1732,
+      "step": 10258
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019237080333358949,
+      "loss": 0.0063,
+      "step": 10259
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019235259434429464,
+      "loss": 0.3486,
+      "step": 10260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001923343846767865,
+      "loss": 1.4199,
+      "step": 10261
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019231617433135672,
+      "loss": 0.4491,
+      "step": 10262
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001922979633082969,
+      "loss": 0.8063,
+      "step": 10263
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001922797516078987,
+      "loss": 0.7906,
+      "step": 10264
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019226153923045377,
+      "loss": 0.245,
+      "step": 10265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019224332617625366,
+      "loss": 3.5488,
+      "step": 10266
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019222511244559008,
+      "loss": 0.3121,
+      "step": 10267
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019220689803875474,
+      "loss": 0.0519,
+      "step": 10268
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019218868295603925,
+      "loss": 0.0128,
+      "step": 10269
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019217046719773528,
+      "loss": 0.4507,
+      "step": 10270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001921522507641346,
+      "loss": 0.6264,
+      "step": 10271
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019213403365552889,
+      "loss": 0.4215,
+      "step": 10272
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019211581587220988,
+      "loss": 0.3652,
+      "step": 10273
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019209759741446925,
+      "loss": 0.4353,
+      "step": 10274
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001920793782825988,
+      "loss": 0.7433,
+      "step": 10275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019206115847689024,
+      "loss": 0.8134,
+      "step": 10276
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019204293799763536,
+      "loss": 0.0059,
+      "step": 10277
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019202471684512593,
+      "loss": 0.6835,
+      "step": 10278
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019200649501965373,
+      "loss": 0.5911,
+      "step": 10279
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001919882725215106,
+      "loss": 0.6855,
+      "step": 10280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019197004935098824,
+      "loss": 0.6616,
+      "step": 10281
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019195182550837853,
+      "loss": 0.1954,
+      "step": 10282
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019193360099397334,
+      "loss": 0.1881,
+      "step": 10283
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001919153758080644,
+      "loss": 0.4153,
+      "step": 10284
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019189714995094365,
+      "loss": 0.3832,
+      "step": 10285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019187892342290296,
+      "loss": 1.3482,
+      "step": 10286
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001918606962242341,
+      "loss": 0.3983,
+      "step": 10287
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019184246835522904,
+      "loss": 0.4395,
+      "step": 10288
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019182423981617964,
+      "loss": 1.0403,
+      "step": 10289
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019180601060737786,
+      "loss": 0.597,
+      "step": 10290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001917877807291155,
+      "loss": 0.5666,
+      "step": 10291
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019176955018168458,
+      "loss": 0.3589,
+      "step": 10292
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019175131896537697,
+      "loss": 0.3908,
+      "step": 10293
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019173308708048467,
+      "loss": 1.0634,
+      "step": 10294
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019171485452729964,
+      "loss": 0.3354,
+      "step": 10295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019169662130611378,
+      "loss": 3.1251,
+      "step": 10296
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001916783874172191,
+      "loss": 0.6232,
+      "step": 10297
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019166015286090763,
+      "loss": 1.3241,
+      "step": 10298
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019164191763747132,
+      "loss": 0.0079,
+      "step": 10299
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019162368174720218,
+      "loss": 1.5039,
+      "step": 10300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019160544519039225,
+      "loss": 1.5077,
+      "step": 10301
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019158720796733357,
+      "loss": 0.0073,
+      "step": 10302
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019156897007831816,
+      "loss": 0.9971,
+      "step": 10303
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019155073152363807,
+      "loss": 0.0079,
+      "step": 10304
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019153249230358538,
+      "loss": 0.325,
+      "step": 10305
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001915142524184522,
+      "loss": 0.5952,
+      "step": 10306
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019149601186853053,
+      "loss": 0.4296,
+      "step": 10307
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019147777065411248,
+      "loss": 0.8874,
+      "step": 10308
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019145952877549027,
+      "loss": 0.4909,
+      "step": 10309
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019144128623295586,
+      "loss": 0.1195,
+      "step": 10310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019142304302680147,
+      "loss": 0.0073,
+      "step": 10311
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019140479915731918,
+      "loss": 1.9931,
+      "step": 10312
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019138655462480118,
+      "loss": 2.1944,
+      "step": 10313
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019136830942953964,
+      "loss": 0.7357,
+      "step": 10314
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001913500635718267,
+      "loss": 0.743,
+      "step": 10315
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019133181705195456,
+      "loss": 1.3232,
+      "step": 10316
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001913135698702154,
+      "loss": 0.3631,
+      "step": 10317
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019129532202690142,
+      "loss": 0.4376,
+      "step": 10318
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019127707352230479,
+      "loss": 0.007,
+      "step": 10319
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019125882435671785,
+      "loss": 1.2484,
+      "step": 10320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019124057453043274,
+      "loss": 0.5791,
+      "step": 10321
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001912223240437417,
+      "loss": 0.5616,
+      "step": 10322
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000191204072896937,
+      "loss": 1.0256,
+      "step": 10323
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019118582109031092,
+      "loss": 0.8885,
+      "step": 10324
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019116756862415572,
+      "loss": 0.0123,
+      "step": 10325
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001911493154987637,
+      "loss": 0.896,
+      "step": 10326
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019113106171442717,
+      "loss": 0.7306,
+      "step": 10327
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019111280727143837,
+      "loss": 0.1324,
+      "step": 10328
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019109455217008973,
+      "loss": 0.184,
+      "step": 10329
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019107629641067347,
+      "loss": 0.4833,
+      "step": 10330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019105803999348194,
+      "loss": 0.0127,
+      "step": 10331
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019103978291880758,
+      "loss": 0.3527,
+      "step": 10332
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019102152518694264,
+      "loss": 0.292,
+      "step": 10333
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019100326679817952,
+      "loss": 0.1172,
+      "step": 10334
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001909850077528107,
+      "loss": 1.0891,
+      "step": 10335
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001909667480511284,
+      "loss": 0.7626,
+      "step": 10336
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001909484876934252,
+      "loss": 0.0194,
+      "step": 10337
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019093022667999337,
+      "loss": 1.0266,
+      "step": 10338
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019091196501112541,
+      "loss": 0.6142,
+      "step": 10339
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001908937026871137,
+      "loss": 0.2335,
+      "step": 10340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019087543970825079,
+      "loss": 0.7383,
+      "step": 10341
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019085717607482898,
+      "loss": 0.5382,
+      "step": 10342
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019083891178714083,
+      "loss": 0.3411,
+      "step": 10343
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00019082064684547884,
+      "loss": 0.3067,
+      "step": 10344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001908023812501354,
+      "loss": 0.3637,
+      "step": 10345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019078411500140304,
+      "loss": 0.1537,
+      "step": 10346
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019076584809957437,
+      "loss": 0.7199,
+      "step": 10347
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019074758054494175,
+      "loss": 0.633,
+      "step": 10348
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019072931233779785,
+      "loss": 0.0806,
+      "step": 10349
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019071104347843508,
+      "loss": 0.372,
+      "step": 10350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019069277396714608,
+      "loss": 2.4063,
+      "step": 10351
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019067450380422333,
+      "loss": 0.0326,
+      "step": 10352
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001906562329899595,
+      "loss": 0.4156,
+      "step": 10353
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019063796152464708,
+      "loss": 1.8339,
+      "step": 10354
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019061968940857875,
+      "loss": 0.3084,
+      "step": 10355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019060141664204702,
+      "loss": 0.7064,
+      "step": 10356
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019058314322534448,
+      "loss": 1.1097,
+      "step": 10357
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001905648691587639,
+      "loss": 0.5593,
+      "step": 10358
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019054659444259775,
+      "loss": 0.501,
+      "step": 10359
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019052831907713882,
+      "loss": 0.387,
+      "step": 10360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001905100430626796,
+      "loss": 0.8787,
+      "step": 10361
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019049176639951288,
+      "loss": 0.0112,
+      "step": 10362
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001904734890879313,
+      "loss": 0.0081,
+      "step": 10363
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019045521112822752,
+      "loss": 0.3041,
+      "step": 10364
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019043693252069423,
+      "loss": 0.6035,
+      "step": 10365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019041865326562416,
+      "loss": 0.9494,
+      "step": 10366
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019040037336331007,
+      "loss": 0.2167,
+      "step": 10367
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019038209281404457,
+      "loss": 0.0099,
+      "step": 10368
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019036381161812046,
+      "loss": 0.7551,
+      "step": 10369
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019034552977583054,
+      "loss": 0.1885,
+      "step": 10370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019032724728746748,
+      "loss": 0.014,
+      "step": 10371
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019030896415332408,
+      "loss": 0.2345,
+      "step": 10372
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001902906803736931,
+      "loss": 0.3564,
+      "step": 10373
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019027239594886732,
+      "loss": 0.3739,
+      "step": 10374
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001902541108791396,
+      "loss": 0.183,
+      "step": 10375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001902358251648027,
+      "loss": 0.3509,
+      "step": 10376
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019021753880614942,
+      "loss": 0.0089,
+      "step": 10377
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019019925180347266,
+      "loss": 0.4999,
+      "step": 10378
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019018096415706522,
+      "loss": 0.0053,
+      "step": 10379
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019016267586721987,
+      "loss": 0.0019,
+      "step": 10380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001901443869342296,
+      "loss": 0.2746,
+      "step": 10381
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001901260973583872,
+      "loss": 0.1578,
+      "step": 10382
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019010780713998556,
+      "loss": 0.1621,
+      "step": 10383
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001900895162793176,
+      "loss": 1.0351,
+      "step": 10384
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001900712247766762,
+      "loss": 0.0065,
+      "step": 10385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019005293263235428,
+      "loss": 0.8029,
+      "step": 10386
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019003463984664472,
+      "loss": 0.1659,
+      "step": 10387
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00019001634641984056,
+      "loss": 0.2935,
+      "step": 10388
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018999805235223461,
+      "loss": 0.2904,
+      "step": 10389
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018997975764411992,
+      "loss": 0.4704,
+      "step": 10390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018996146229578937,
+      "loss": 0.0054,
+      "step": 10391
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018994316630753597,
+      "loss": 0.1842,
+      "step": 10392
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018992486967965276,
+      "loss": 0.649,
+      "step": 10393
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018990657241243263,
+      "loss": 0.8866,
+      "step": 10394
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018988827450616865,
+      "loss": 0.2386,
+      "step": 10395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018986997596115383,
+      "loss": 0.8035,
+      "step": 10396
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001898516767776812,
+      "loss": 0.4389,
+      "step": 10397
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018983337695604373,
+      "loss": 0.7765,
+      "step": 10398
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018981507649653453,
+      "loss": 1.5515,
+      "step": 10399
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018979677539944663,
+      "loss": 0.4442,
+      "step": 10400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018977847366507308,
+      "loss": 0.6948,
+      "step": 10401
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018976017129370703,
+      "loss": 0.2659,
+      "step": 10402
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018974186828564146,
+      "loss": 0.6062,
+      "step": 10403
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001897235646411695,
+      "loss": 0.3372,
+      "step": 10404
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018970526036058433,
+      "loss": 0.6496,
+      "step": 10405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018968695544417896,
+      "loss": 0.5535,
+      "step": 10406
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018966864989224659,
+      "loss": 0.0017,
+      "step": 10407
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018965034370508032,
+      "loss": 0.0014,
+      "step": 10408
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018963203688297328,
+      "loss": 1.3208,
+      "step": 10409
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018961372942621863,
+      "loss": 0.9907,
+      "step": 10410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001895954213351096,
+      "loss": 0.5037,
+      "step": 10411
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001895771126099393,
+      "loss": 0.3454,
+      "step": 10412
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018955880325100091,
+      "loss": 0.554,
+      "step": 10413
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001895404932585877,
+      "loss": 2.7715,
+      "step": 10414
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018952218263299284,
+      "loss": 0.3533,
+      "step": 10415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018950387137450954,
+      "loss": 0.3383,
+      "step": 10416
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018948555948343097,
+      "loss": 0.0094,
+      "step": 10417
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018946724696005047,
+      "loss": 0.0231,
+      "step": 10418
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018944893380466125,
+      "loss": 0.7094,
+      "step": 10419
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018943062001755654,
+      "loss": 0.8826,
+      "step": 10420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001894123055990296,
+      "loss": 0.3897,
+      "step": 10421
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018939399054937377,
+      "loss": 0.5977,
+      "step": 10422
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018937567486888233,
+      "loss": 0.8332,
+      "step": 10423
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001893573585578485,
+      "loss": 0.6374,
+      "step": 10424
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018933904161656567,
+      "loss": 0.0354,
+      "step": 10425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018932072404532715,
+      "loss": 0.2379,
+      "step": 10426
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001893024058444262,
+      "loss": 0.9867,
+      "step": 10427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018928408701415623,
+      "loss": 0.1978,
+      "step": 10428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018926576755481052,
+      "loss": 0.3141,
+      "step": 10429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018924744746668252,
+      "loss": 0.2707,
+      "step": 10430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018922912675006556,
+      "loss": 0.2694,
+      "step": 10431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018921080540525295,
+      "loss": 1.0119,
+      "step": 10432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018919248343253822,
+      "loss": 0.6146,
+      "step": 10433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018917416083221464,
+      "loss": 0.7328,
+      "step": 10434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018915583760457567,
+      "loss": 0.0046,
+      "step": 10435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018913751374991474,
+      "loss": 0.0098,
+      "step": 10436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001891191892685252,
+      "loss": 2.6222,
+      "step": 10437
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018910086416070061,
+      "loss": 0.836,
+      "step": 10438
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001890825384267344,
+      "loss": 0.0021,
+      "step": 10439
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018906421206691992,
+      "loss": 0.1749,
+      "step": 10440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018904588508155074,
+      "loss": 0.2164,
+      "step": 10441
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018902755747092032,
+      "loss": 1.3978,
+      "step": 10442
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018900922923532212,
+      "loss": 0.4988,
+      "step": 10443
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018899090037504963,
+      "loss": 0.2636,
+      "step": 10444
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018897257089039646,
+      "loss": 0.7424,
+      "step": 10445
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018895424078165603,
+      "loss": 0.2238,
+      "step": 10446
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018893591004912186,
+      "loss": 0.2463,
+      "step": 10447
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001889175786930876,
+      "loss": 0.227,
+      "step": 10448
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018889924671384664,
+      "loss": 0.0166,
+      "step": 10449
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018888091411169267,
+      "loss": 0.337,
+      "step": 10450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018886258088691922,
+      "loss": 0.1885,
+      "step": 10451
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018884424703981983,
+      "loss": 0.58,
+      "step": 10452
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018882591257068812,
+      "loss": 4.2964,
+      "step": 10453
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018880757747981775,
+      "loss": 0.9495,
+      "step": 10454
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018878924176750217,
+      "loss": 0.4221,
+      "step": 10455
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018877090543403514,
+      "loss": 0.0375,
+      "step": 10456
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018875256847971028,
+      "loss": 0.9246,
+      "step": 10457
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001887342309048212,
+      "loss": 0.2325,
+      "step": 10458
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001887158927096615,
+      "loss": 0.8364,
+      "step": 10459
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001886975538945249,
+      "loss": 0.371,
+      "step": 10460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018867921445970504,
+      "loss": 0.8108,
+      "step": 10461
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018866087440549563,
+      "loss": 0.0168,
+      "step": 10462
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018864253373219032,
+      "loss": 0.2841,
+      "step": 10463
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018862419244008286,
+      "loss": 0.9958,
+      "step": 10464
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018860585052946693,
+      "loss": 0.4325,
+      "step": 10465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018858750800063617,
+      "loss": 0.0088,
+      "step": 10466
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001885691648538844,
+      "loss": 0.0869,
+      "step": 10467
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018855082108950542,
+      "loss": 0.7236,
+      "step": 10468
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018853247670779283,
+      "loss": 0.0079,
+      "step": 10469
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018851413170904043,
+      "loss": 0.2729,
+      "step": 10470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001884957860935421,
+      "loss": 0.19,
+      "step": 10471
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018847743986159148,
+      "loss": 0.3514,
+      "step": 10472
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001884590930134824,
+      "loss": 0.0042,
+      "step": 10473
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018844074554950864,
+      "loss": 0.0025,
+      "step": 10474
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018842239746996408,
+      "loss": 0.8413,
+      "step": 10475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018840404877514243,
+      "loss": 0.6925,
+      "step": 10476
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018838569946533768,
+      "loss": 0.3152,
+      "step": 10477
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018836734954084345,
+      "loss": 0.539,
+      "step": 10478
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018834899900195375,
+      "loss": 0.9462,
+      "step": 10479
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001883306478489624,
+      "loss": 0.3476,
+      "step": 10480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001883122960821632,
+      "loss": 0.6839,
+      "step": 10481
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001882939437018501,
+      "loss": 0.4835,
+      "step": 10482
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000188275590708317,
+      "loss": 0.0018,
+      "step": 10483
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018825723710185773,
+      "loss": 0.543,
+      "step": 10484
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001882388828827662,
+      "loss": 0.5091,
+      "step": 10485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001882205280513364,
+      "loss": 1.4287,
+      "step": 10486
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018820217260786216,
+      "loss": 0.3769,
+      "step": 10487
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018818381655263746,
+      "loss": 0.6064,
+      "step": 10488
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018816545988595626,
+      "loss": 0.485,
+      "step": 10489
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001881471026081125,
+      "loss": 0.514,
+      "step": 10490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001881287447194002,
+      "loss": 0.8659,
+      "step": 10491
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018811038622011317,
+      "loss": 0.1555,
+      "step": 10492
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018809202711054555,
+      "loss": 0.7476,
+      "step": 10493
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018807366739099127,
+      "loss": 0.4446,
+      "step": 10494
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001880553070617444,
+      "loss": 0.1989,
+      "step": 10495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001880369461230988,
+      "loss": 0.2991,
+      "step": 10496
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001880185845753487,
+      "loss": 0.0173,
+      "step": 10497
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018800022241878794,
+      "loss": 1.0073,
+      "step": 10498
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001879818596537107,
+      "loss": 0.2634,
+      "step": 10499
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018796349628041095,
+      "loss": 0.0059,
+      "step": 10500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001879451322991828,
+      "loss": 0.0056,
+      "step": 10501
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001879267677103203,
+      "loss": 0.5516,
+      "step": 10502
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018790840251411755,
+      "loss": 0.5093,
+      "step": 10503
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018789003671086863,
+      "loss": 0.5969,
+      "step": 10504
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001878716703008676,
+      "loss": 0.4271,
+      "step": 10505
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018785330328440866,
+      "loss": 0.8851,
+      "step": 10506
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018783493566178582,
+      "loss": 0.3768,
+      "step": 10507
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001878165674332933,
+      "loss": 0.8389,
+      "step": 10508
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001877981985992252,
+      "loss": 0.371,
+      "step": 10509
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018777982915987567,
+      "loss": 0.3012,
+      "step": 10510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018776145911553887,
+      "loss": 0.2893,
+      "step": 10511
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187743088466509,
+      "loss": 0.0883,
+      "step": 10512
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018772471721308022,
+      "loss": 0.0033,
+      "step": 10513
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018770634535554667,
+      "loss": 0.5909,
+      "step": 10514
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001876879728942026,
+      "loss": 0.6488,
+      "step": 10515
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001876695998293422,
+      "loss": 0.4109,
+      "step": 10516
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018765122616125977,
+      "loss": 0.7351,
+      "step": 10517
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001876328518902494,
+      "loss": 0.189,
+      "step": 10518
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018761447701660534,
+      "loss": 0.1659,
+      "step": 10519
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018759610154062196,
+      "loss": 0.9186,
+      "step": 10520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001875777254625934,
+      "loss": 0.8754,
+      "step": 10521
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018755934878281398,
+      "loss": 0.0107,
+      "step": 10522
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001875409715015779,
+      "loss": 0.9599,
+      "step": 10523
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018752259361917958,
+      "loss": 0.0119,
+      "step": 10524
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001875042151359132,
+      "loss": 0.2527,
+      "step": 10525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018748583605207312,
+      "loss": 0.0052,
+      "step": 10526
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001874674563679536,
+      "loss": 0.7072,
+      "step": 10527
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187449076083849,
+      "loss": 0.4635,
+      "step": 10528
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018743069520005368,
+      "loss": 0.0425,
+      "step": 10529
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001874123137168619,
+      "loss": 0.4206,
+      "step": 10530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001873939316345681,
+      "loss": 0.2341,
+      "step": 10531
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018737554895346666,
+      "loss": 0.0009,
+      "step": 10532
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001873571656738518,
+      "loss": 0.3379,
+      "step": 10533
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018733878179601803,
+      "loss": 0.367,
+      "step": 10534
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018732039732025977,
+      "loss": 0.0006,
+      "step": 10535
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018730201224687127,
+      "loss": 0.0113,
+      "step": 10536
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018728362657614706,
+      "loss": 0.4975,
+      "step": 10537
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018726524030838153,
+      "loss": 0.4556,
+      "step": 10538
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018724685344386909,
+      "loss": 0.3812,
+      "step": 10539
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001872284659829042,
+      "loss": 0.4407,
+      "step": 10540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018721007792578135,
+      "loss": 0.7213,
+      "step": 10541
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018719168927279484,
+      "loss": 0.0028,
+      "step": 10542
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018717330002423932,
+      "loss": 0.1945,
+      "step": 10543
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018715491018040918,
+      "loss": 0.2592,
+      "step": 10544
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018713651974159884,
+      "loss": 0.0027,
+      "step": 10545
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000187118128708103,
+      "loss": 0.1516,
+      "step": 10546
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018709973708021596,
+      "loss": 2.8916,
+      "step": 10547
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001870813448582323,
+      "loss": 0.0031,
+      "step": 10548
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018706295204244658,
+      "loss": 0.2003,
+      "step": 10549
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001870445586331533,
+      "loss": 0.9783,
+      "step": 10550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018702616463064703,
+      "loss": 0.183,
+      "step": 10551
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001870077700352223,
+      "loss": 0.1779,
+      "step": 10552
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001869893748471736,
+      "loss": 0.4526,
+      "step": 10553
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018697097906679563,
+      "loss": 0.4096,
+      "step": 10554
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018695258269438297,
+      "loss": 0.459,
+      "step": 10555
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018693418573023011,
+      "loss": 0.4147,
+      "step": 10556
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018691578817463163,
+      "loss": 0.2902,
+      "step": 10557
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018689739002788227,
+      "loss": 0.6561,
+      "step": 10558
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001868789912902766,
+      "loss": 0.7207,
+      "step": 10559
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018686059196210922,
+      "loss": 0.0056,
+      "step": 10560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018684219204367477,
+      "loss": 0.9336,
+      "step": 10561
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018682379153526788,
+      "loss": 1.9099,
+      "step": 10562
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018680539043718327,
+      "loss": 0.007,
+      "step": 10563
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018678698874971555,
+      "loss": 0.2591,
+      "step": 10564
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018676858647315942,
+      "loss": 0.1164,
+      "step": 10565
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018675018360780958,
+      "loss": 3.8633,
+      "step": 10566
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001867317801539607,
+      "loss": 0.2245,
+      "step": 10567
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018671337611190743,
+      "loss": 0.4199,
+      "step": 10568
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001866949714819446,
+      "loss": 3.5283,
+      "step": 10569
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018667656626436687,
+      "loss": 0.278,
+      "step": 10570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018665816045946898,
+      "loss": 0.3626,
+      "step": 10571
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018663975406754566,
+      "loss": 0.2542,
+      "step": 10572
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018662134708889166,
+      "loss": 0.4173,
+      "step": 10573
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001866029395238017,
+      "loss": 2.4063,
+      "step": 10574
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001865845313725707,
+      "loss": 1.7482,
+      "step": 10575
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018656612263549322,
+      "loss": 0.4617,
+      "step": 10576
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018654771331286425,
+      "loss": 2.1735,
+      "step": 10577
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001865293034049785,
+      "loss": 0.5813,
+      "step": 10578
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018651089291213068,
+      "loss": 0.1385,
+      "step": 10579
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001864924818346158,
+      "loss": 0.5697,
+      "step": 10580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018647407017272858,
+      "loss": 0.2255,
+      "step": 10581
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018645565792676388,
+      "loss": 1.125,
+      "step": 10582
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018643724509701643,
+      "loss": 0.3084,
+      "step": 10583
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001864188316837813,
+      "loss": 0.3769,
+      "step": 10584
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018640041768735321,
+      "loss": 0.4686,
+      "step": 10585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000186382003108027,
+      "loss": 0.3707,
+      "step": 10586
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001863635879460977,
+      "loss": 0.2301,
+      "step": 10587
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018634517220186004,
+      "loss": 0.2198,
+      "step": 10588
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018632675587560907,
+      "loss": 0.0026,
+      "step": 10589
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001863083389676396,
+      "loss": 0.4626,
+      "step": 10590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018628992147824648,
+      "loss": 2.1267,
+      "step": 10591
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018627150340772483,
+      "loss": 0.0051,
+      "step": 10592
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00018625308475636945,
+      "loss": 0.3531,
+      "step": 10593
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018623466552447534,
+      "loss": 0.3249,
+      "step": 10594
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018621624571233745,
+      "loss": 0.051,
+      "step": 10595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018619782532025076,
+      "loss": 0.3615,
+      "step": 10596
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018617940434851015,
+      "loss": 0.6083,
+      "step": 10597
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018616098279741072,
+      "loss": 0.0018,
+      "step": 10598
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018614256066724742,
+      "loss": 3.1205,
+      "step": 10599
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018612413795831526,
+      "loss": 0.7228,
+      "step": 10600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018610571467090925,
+      "loss": 0.0151,
+      "step": 10601
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018608729080532438,
+      "loss": 0.3236,
+      "step": 10602
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018606886636185572,
+      "loss": 0.0447,
+      "step": 10603
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018605044134079832,
+      "loss": 0.5644,
+      "step": 10604
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018603201574244718,
+      "loss": 0.0056,
+      "step": 10605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018601358956709735,
+      "loss": 0.1614,
+      "step": 10606
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000185995162815044,
+      "loss": 0.16,
+      "step": 10607
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001859767354865821,
+      "loss": 1.4092,
+      "step": 10608
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018595830758200678,
+      "loss": 0.1759,
+      "step": 10609
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018593987910161313,
+      "loss": 0.2785,
+      "step": 10610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018592145004569624,
+      "loss": 0.254,
+      "step": 10611
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018590302041455127,
+      "loss": 0.5629,
+      "step": 10612
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018588459020847328,
+      "loss": 0.0045,
+      "step": 10613
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018586615942775745,
+      "loss": 3.0879,
+      "step": 10614
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018584772807269892,
+      "loss": 0.7071,
+      "step": 10615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018582929614359282,
+      "loss": 0.0092,
+      "step": 10616
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018581086364073427,
+      "loss": 0.4,
+      "step": 10617
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018579243056441854,
+      "loss": 0.0277,
+      "step": 10618
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018577399691494073,
+      "loss": 0.5485,
+      "step": 10619
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018575556269259606,
+      "loss": 1.4239,
+      "step": 10620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018573712789767967,
+      "loss": 0.6472,
+      "step": 10621
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018571869253048687,
+      "loss": 0.0261,
+      "step": 10622
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018570025659131277,
+      "loss": 0.6688,
+      "step": 10623
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018568182008045267,
+      "loss": 1.4305,
+      "step": 10624
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018566338299820176,
+      "loss": 0.3423,
+      "step": 10625
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018564494534485527,
+      "loss": 0.3138,
+      "step": 10626
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018562650712070853,
+      "loss": 0.0035,
+      "step": 10627
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001856080683260567,
+      "loss": 0.3091,
+      "step": 10628
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018558962896119508,
+      "loss": 0.8626,
+      "step": 10629
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018557118902641905,
+      "loss": 0.8495,
+      "step": 10630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018555274852202373,
+      "loss": 0.0089,
+      "step": 10631
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001855343074483045,
+      "loss": 0.8057,
+      "step": 10632
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018551586580555673,
+      "loss": 0.469,
+      "step": 10633
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018549742359407563,
+      "loss": 0.3355,
+      "step": 10634
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018547898081415658,
+      "loss": 0.6155,
+      "step": 10635
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018546053746609488,
+      "loss": 0.5495,
+      "step": 10636
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001854420935501859,
+      "loss": 0.0028,
+      "step": 10637
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018542364906672497,
+      "loss": 0.0925,
+      "step": 10638
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018540520401600754,
+      "loss": 1.7994,
+      "step": 10639
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001853867583983288,
+      "loss": 0.282,
+      "step": 10640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018536831221398428,
+      "loss": 1.3025,
+      "step": 10641
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018534986546326932,
+      "loss": 0.7554,
+      "step": 10642
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018533141814647928,
+      "loss": 0.8376,
+      "step": 10643
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018531297026390967,
+      "loss": 0.7186,
+      "step": 10644
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001852945218158558,
+      "loss": 0.2769,
+      "step": 10645
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018527607280261312,
+      "loss": 0.4045,
+      "step": 10646
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001852576232244771,
+      "loss": 0.2944,
+      "step": 10647
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018523917308174313,
+      "loss": 0.3982,
+      "step": 10648
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018522072237470673,
+      "loss": 0.7401,
+      "step": 10649
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018520227110366326,
+      "loss": 0.2874,
+      "step": 10650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001851838192689083,
+      "loss": 0.2049,
+      "step": 10651
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018516536687073726,
+      "loss": 0.622,
+      "step": 10652
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001851469139094457,
+      "loss": 0.3162,
+      "step": 10653
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018512846038532901,
+      "loss": 0.2163,
+      "step": 10654
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018511000629868273,
+      "loss": 0.1406,
+      "step": 10655
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018509155164980244,
+      "loss": 0.5123,
+      "step": 10656
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018507309643898357,
+      "loss": 0.6367,
+      "step": 10657
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018505464066652169,
+      "loss": 0.4199,
+      "step": 10658
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001850361843327124,
+      "loss": 0.5111,
+      "step": 10659
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001850177274378512,
+      "loss": 0.0042,
+      "step": 10660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001849992699822336,
+      "loss": 0.3312,
+      "step": 10661
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018498081196615524,
+      "loss": 0.4451,
+      "step": 10662
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018496235338991168,
+      "loss": 0.3256,
+      "step": 10663
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001849438942537985,
+      "loss": 0.8959,
+      "step": 10664
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018492543455811132,
+      "loss": 1.1673,
+      "step": 10665
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018490697430314568,
+      "loss": 0.4109,
+      "step": 10666
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018488851348919723,
+      "loss": 0.0156,
+      "step": 10667
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018487005211656165,
+      "loss": 0.327,
+      "step": 10668
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018485159018553449,
+      "loss": 0.7891,
+      "step": 10669
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001848331276964114,
+      "loss": 0.9547,
+      "step": 10670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018481466464948809,
+      "loss": 0.692,
+      "step": 10671
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018479620104506011,
+      "loss": 0.4258,
+      "step": 10672
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018477773688342324,
+      "loss": 0.2797,
+      "step": 10673
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018475927216487313,
+      "loss": 0.0059,
+      "step": 10674
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018474080688970538,
+      "loss": 0.1789,
+      "step": 10675
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001847223410582158,
+      "loss": 0.8278,
+      "step": 10676
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018470387467069997,
+      "loss": 1.5695,
+      "step": 10677
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018468540772745374,
+      "loss": 0.5646,
+      "step": 10678
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018466694022877274,
+      "loss": 0.3645,
+      "step": 10679
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001846484721749527,
+      "loss": 0.0168,
+      "step": 10680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001846300035662894,
+      "loss": 0.2965,
+      "step": 10681
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018461153440307856,
+      "loss": 0.1809,
+      "step": 10682
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018459306468561595,
+      "loss": 0.3388,
+      "step": 10683
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018457459441419732,
+      "loss": 0.3527,
+      "step": 10684
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018455612358911846,
+      "loss": 0.0025,
+      "step": 10685
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018453765221067511,
+      "loss": 0.6475,
+      "step": 10686
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001845191802791631,
+      "loss": 0.4168,
+      "step": 10687
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018450070779487828,
+      "loss": 0.052,
+      "step": 10688
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018448223475811635,
+      "loss": 0.0037,
+      "step": 10689
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018446376116917315,
+      "loss": 1.8087,
+      "step": 10690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001844452870283446,
+      "loss": 0.0297,
+      "step": 10691
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018442681233592646,
+      "loss": 0.4018,
+      "step": 10692
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018440833709221455,
+      "loss": 0.4724,
+      "step": 10693
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018438986129750483,
+      "loss": 1.0252,
+      "step": 10694
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018437138495209303,
+      "loss": 0.0044,
+      "step": 10695
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018435290805627512,
+      "loss": 0.2637,
+      "step": 10696
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018433443061034694,
+      "loss": 0.1883,
+      "step": 10697
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018431595261460436,
+      "loss": 0.2547,
+      "step": 10698
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018429747406934334,
+      "loss": 0.0057,
+      "step": 10699
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001842789949748597,
+      "loss": 0.442,
+      "step": 10700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018426051533144942,
+      "loss": 0.455,
+      "step": 10701
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018424203513940843,
+      "loss": 0.2865,
+      "step": 10702
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001842235543990326,
+      "loss": 0.4312,
+      "step": 10703
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001842050731106179,
+      "loss": 0.985,
+      "step": 10704
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018418659127446036,
+      "loss": 0.0024,
+      "step": 10705
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001841681088908558,
+      "loss": 0.4865,
+      "step": 10706
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018414962596010023,
+      "loss": 0.5605,
+      "step": 10707
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018413114248248973,
+      "loss": 1.2147,
+      "step": 10708
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018411265845832018,
+      "loss": 0.9312,
+      "step": 10709
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001840941738878876,
+      "loss": 0.5362,
+      "step": 10710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018407568877148797,
+      "loss": 0.653,
+      "step": 10711
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018405720310941736,
+      "loss": 0.5237,
+      "step": 10712
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001840387169019717,
+      "loss": 0.6041,
+      "step": 10713
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018402023014944718,
+      "loss": 0.0007,
+      "step": 10714
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001840017428521396,
+      "loss": 0.0479,
+      "step": 10715
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001839832550103452,
+      "loss": 0.9631,
+      "step": 10716
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018396476662436,
+      "loss": 0.3657,
+      "step": 10717
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018394627769448003,
+      "loss": 0.216,
+      "step": 10718
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018392778822100132,
+      "loss": 0.4847,
+      "step": 10719
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018390929820422003,
+      "loss": 0.4603,
+      "step": 10720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001838908076444322,
+      "loss": 1.005,
+      "step": 10721
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018387231654193399,
+      "loss": 0.4789,
+      "step": 10722
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018385382489702142,
+      "loss": 0.2194,
+      "step": 10723
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018383533270999067,
+      "loss": 0.1151,
+      "step": 10724
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018381683998113787,
+      "loss": 0.0048,
+      "step": 10725
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018379834671075908,
+      "loss": 0.0098,
+      "step": 10726
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018377985289915052,
+      "loss": 0.8799,
+      "step": 10727
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018376135854660837,
+      "loss": 0.3914,
+      "step": 10728
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018374286365342864,
+      "loss": 0.0288,
+      "step": 10729
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018372436821990763,
+      "loss": 1.131,
+      "step": 10730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001837058722463415,
+      "loss": 0.1187,
+      "step": 10731
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018368737573302638,
+      "loss": 0.3371,
+      "step": 10732
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018366887868025847,
+      "loss": 0.004,
+      "step": 10733
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018365038108833403,
+      "loss": 0.2573,
+      "step": 10734
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018363188295754922,
+      "loss": 0.6128,
+      "step": 10735
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001836133842882003,
+      "loss": 0.3706,
+      "step": 10736
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001835948850805835,
+      "loss": 0.186,
+      "step": 10737
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018357638533499497,
+      "loss": 0.0202,
+      "step": 10738
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018355788505173106,
+      "loss": 0.55,
+      "step": 10739
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000183539384231088,
+      "loss": 0.0115,
+      "step": 10740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018352088287336196,
+      "loss": 0.2517,
+      "step": 10741
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018350238097884937,
+      "loss": 0.8364,
+      "step": 10742
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018348387854784644,
+      "loss": 3.3419,
+      "step": 10743
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001834653755806494,
+      "loss": 0.002,
+      "step": 10744
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001834468720775546,
+      "loss": 0.1922,
+      "step": 10745
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018342836803885833,
+      "loss": 0.5209,
+      "step": 10746
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018340986346485695,
+      "loss": 0.258,
+      "step": 10747
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001833913583558467,
+      "loss": 0.2111,
+      "step": 10748
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000183372852712124,
+      "loss": 1.0787,
+      "step": 10749
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018335434653398515,
+      "loss": 0.5032,
+      "step": 10750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018333583982172648,
+      "loss": 0.0077,
+      "step": 10751
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018331733257564439,
+      "loss": 0.0215,
+      "step": 10752
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018329882479603516,
+      "loss": 0.4099,
+      "step": 10753
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001832803164831953,
+      "loss": 2.4765,
+      "step": 10754
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001832618076374211,
+      "loss": 0.4213,
+      "step": 10755
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001832432982590089,
+      "loss": 0.0122,
+      "step": 10756
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018322478834825525,
+      "loss": 0.0076,
+      "step": 10757
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018320627790545648,
+      "loss": 0.4293,
+      "step": 10758
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018318776693090896,
+      "loss": 0.6815,
+      "step": 10759
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018316925542490918,
+      "loss": 0.0064,
+      "step": 10760
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018315074338775354,
+      "loss": 0.8434,
+      "step": 10761
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018313223081973853,
+      "loss": 0.4717,
+      "step": 10762
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018311371772116055,
+      "loss": 0.3856,
+      "step": 10763
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018309520409231605,
+      "loss": 0.4912,
+      "step": 10764
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018307668993350153,
+      "loss": 0.4125,
+      "step": 10765
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018305817524501354,
+      "loss": 0.8076,
+      "step": 10766
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018303966002714844,
+      "loss": 0.0021,
+      "step": 10767
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001830211442802027,
+      "loss": 0.0018,
+      "step": 10768
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018300262800447298,
+      "loss": 0.5674,
+      "step": 10769
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018298411120025568,
+      "loss": 0.4401,
+      "step": 10770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018296559386784736,
+      "loss": 1.1087,
+      "step": 10771
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018294707600754445,
+      "loss": 0.9873,
+      "step": 10772
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018292855761964364,
+      "loss": 0.3999,
+      "step": 10773
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018291003870444137,
+      "loss": 0.524,
+      "step": 10774
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018289151926223425,
+      "loss": 0.0106,
+      "step": 10775
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018287299929331877,
+      "loss": 0.6149,
+      "step": 10776
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018285447879799158,
+      "loss": 0.9193,
+      "step": 10777
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828359577765492,
+      "loss": 0.3491,
+      "step": 10778
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001828174362292882,
+      "loss": 0.6025,
+      "step": 10779
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827989141565053,
+      "loss": 0.009,
+      "step": 10780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018278039155849695,
+      "loss": 0.0022,
+      "step": 10781
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018276186843555985,
+      "loss": 0.7667,
+      "step": 10782
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018274334478799058,
+      "loss": 0.55,
+      "step": 10783
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827248206160858,
+      "loss": 0.4447,
+      "step": 10784
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001827062959201421,
+      "loss": 2.4298,
+      "step": 10785
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001826877707004562,
+      "loss": 0.1942,
+      "step": 10786
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018266924495732466,
+      "loss": 0.1696,
+      "step": 10787
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001826507186910442,
+      "loss": 0.0151,
+      "step": 10788
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018263219190191153,
+      "loss": 0.0108,
+      "step": 10789
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018261366459022322,
+      "loss": 2.5275,
+      "step": 10790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018259513675627606,
+      "loss": 0.6812,
+      "step": 10791
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018257660840036673,
+      "loss": 0.8934,
+      "step": 10792
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018255807952279184,
+      "loss": 0.672,
+      "step": 10793
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018253955012384818,
+      "loss": 0.5117,
+      "step": 10794
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018252102020383247,
+      "loss": 0.5477,
+      "step": 10795
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018250248976304148,
+      "loss": 2.0269,
+      "step": 10796
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018248395880177183,
+      "loss": 0.5329,
+      "step": 10797
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018246542732032033,
+      "loss": 1.3726,
+      "step": 10798
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018244689531898375,
+      "loss": 1.0518,
+      "step": 10799
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018242836279805887,
+      "loss": 0.3746,
+      "step": 10800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018240982975784245,
+      "loss": 1.0948,
+      "step": 10801
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018239129619863114,
+      "loss": 0.0043,
+      "step": 10802
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823727621207219,
+      "loss": 0.6276,
+      "step": 10803
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823542275244115,
+      "loss": 0.8442,
+      "step": 10804
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823356924099966,
+      "loss": 0.0102,
+      "step": 10805
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001823171567777742,
+      "loss": 0.4997,
+      "step": 10806
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000182298620628041,
+      "loss": 1.0753,
+      "step": 10807
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018228008396109388,
+      "loss": 0.1662,
+      "step": 10808
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018226154677722964,
+      "loss": 0.411,
+      "step": 10809
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018224300907674514,
+      "loss": 0.5677,
+      "step": 10810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018222447085993726,
+      "loss": 3.1262,
+      "step": 10811
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018220593212710286,
+      "loss": 1.5185,
+      "step": 10812
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018218739287853875,
+      "loss": 0.1624,
+      "step": 10813
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018216885311454187,
+      "loss": 0.3098,
+      "step": 10814
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018215031283540912,
+      "loss": 0.0166,
+      "step": 10815
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018213177204143733,
+      "loss": 0.0747,
+      "step": 10816
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001821132307329234,
+      "loss": 1.6896,
+      "step": 10817
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018209468891016432,
+      "loss": 1.4595,
+      "step": 10818
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018207614657345697,
+      "loss": 0.1977,
+      "step": 10819
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018205760372309824,
+      "loss": 0.619,
+      "step": 10820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001820390603593851,
+      "loss": 0.8511,
+      "step": 10821
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018202051648261448,
+      "loss": 0.2785,
+      "step": 10822
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001820019720930834,
+      "loss": 0.0912,
+      "step": 10823
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018198342719108872,
+      "loss": 0.2873,
+      "step": 10824
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018196488177692747,
+      "loss": 1.0418,
+      "step": 10825
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001819463358508966,
+      "loss": 0.4548,
+      "step": 10826
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018192778941329314,
+      "loss": 0.0113,
+      "step": 10827
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000181909242464414,
+      "loss": 0.1559,
+      "step": 10828
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018189069500455627,
+      "loss": 0.5449,
+      "step": 10829
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001818721470340169,
+      "loss": 0.4356,
+      "step": 10830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018185359855309296,
+      "loss": 0.1311,
+      "step": 10831
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001818350495620814,
+      "loss": 0.2774,
+      "step": 10832
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001818165000612793,
+      "loss": 0.0089,
+      "step": 10833
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018179795005098375,
+      "loss": 0.3987,
+      "step": 10834
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001817793995314917,
+      "loss": 0.301,
+      "step": 10835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018176084850310027,
+      "loss": 0.3872,
+      "step": 10836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018174229696610653,
+      "loss": 0.4284,
+      "step": 10837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018172374492080758,
+      "loss": 0.0145,
+      "step": 10838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001817051923675004,
+      "loss": 1.3412,
+      "step": 10839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018168663930648215,
+      "loss": 0.0458,
+      "step": 10840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018166808573804997,
+      "loss": 0.2372,
+      "step": 10841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00018164953166250093,
+      "loss": 0.3534,
+      "step": 10842
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018163097708013206,
+      "loss": 0.2705,
+      "step": 10843
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018161242199124067,
+      "loss": 0.635,
+      "step": 10844
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001815938663961237,
+      "loss": 0.6122,
+      "step": 10845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018157531029507843,
+      "loss": 0.0147,
+      "step": 10846
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018155675368840193,
+      "loss": 0.4746,
+      "step": 10847
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018153819657639136,
+      "loss": 0.628,
+      "step": 10848
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018151963895934393,
+      "loss": 0.7033,
+      "step": 10849
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018150108083755685,
+      "loss": 0.3552,
+      "step": 10850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001814825222113271,
+      "loss": 0.344,
+      "step": 10851
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018146396308095204,
+      "loss": 0.9538,
+      "step": 10852
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018144540344672895,
+      "loss": 0.4069,
+      "step": 10853
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001814268433089548,
+      "loss": 0.0104,
+      "step": 10854
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018140828266792697,
+      "loss": 0.6231,
+      "step": 10855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001813897215239426,
+      "loss": 0.017,
+      "step": 10856
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018137115987729898,
+      "loss": 0.4698,
+      "step": 10857
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018135259772829326,
+      "loss": 0.2732,
+      "step": 10858
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018133403507722282,
+      "loss": 0.5847,
+      "step": 10859
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018131547192438478,
+      "loss": 1.0674,
+      "step": 10860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018129690827007645,
+      "loss": 0.3661,
+      "step": 10861
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018127834411459513,
+      "loss": 0.8159,
+      "step": 10862
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018125977945823805,
+      "loss": 0.3931,
+      "step": 10863
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001812412143013026,
+      "loss": 0.2329,
+      "step": 10864
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001812226486440859,
+      "loss": 0.3406,
+      "step": 10865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018120408248688532,
+      "loss": 1.0667,
+      "step": 10866
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018118551582999826,
+      "loss": 0.3828,
+      "step": 10867
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018116694867372193,
+      "loss": 0.2027,
+      "step": 10868
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001811483810183537,
+      "loss": 0.6642,
+      "step": 10869
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001811298128641909,
+      "loss": 0.312,
+      "step": 10870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018111124421153085,
+      "loss": 0.4434,
+      "step": 10871
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018109267506067093,
+      "loss": 0.089,
+      "step": 10872
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018107410541190846,
+      "loss": 0.5127,
+      "step": 10873
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018105553526554086,
+      "loss": 0.5112,
+      "step": 10874
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018103696462186546,
+      "loss": 0.3306,
+      "step": 10875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018101839348117965,
+      "loss": 0.4577,
+      "step": 10876
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001809998218437808,
+      "loss": 0.2261,
+      "step": 10877
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018098124970996634,
+      "loss": 0.0024,
+      "step": 10878
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001809626770800337,
+      "loss": 2.7641,
+      "step": 10879
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001809441039542802,
+      "loss": 0.2122,
+      "step": 10880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018092553033300336,
+      "loss": 0.3658,
+      "step": 10881
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018090695621650053,
+      "loss": 0.3658,
+      "step": 10882
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018088838160506922,
+      "loss": 0.5187,
+      "step": 10883
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001808698064990068,
+      "loss": 1.6924,
+      "step": 10884
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018085123089861074,
+      "loss": 0.1985,
+      "step": 10885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018083265480417854,
+      "loss": 0.705,
+      "step": 10886
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018081407821600768,
+      "loss": 0.0063,
+      "step": 10887
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018079550113439557,
+      "loss": 0.1478,
+      "step": 10888
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018077692355963974,
+      "loss": 0.8262,
+      "step": 10889
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001807583454920377,
+      "loss": 0.9108,
+      "step": 10890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018073976693188686,
+      "loss": 0.2542,
+      "step": 10891
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018072118787948476,
+      "loss": 0.1891,
+      "step": 10892
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018070260833512902,
+      "loss": 0.6029,
+      "step": 10893
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018068402829911706,
+      "loss": 0.1446,
+      "step": 10894
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018066544777174643,
+      "loss": 0.0862,
+      "step": 10895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018064686675331467,
+      "loss": 0.4348,
+      "step": 10896
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018062828524411935,
+      "loss": 0.0043,
+      "step": 10897
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000180609703244458,
+      "loss": 0.5377,
+      "step": 10898
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001805911207546282,
+      "loss": 1.905,
+      "step": 10899
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018057253777492747,
+      "loss": 0.2561,
+      "step": 10900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018055395430565348,
+      "loss": 2.3246,
+      "step": 10901
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018053537034710377,
+      "loss": 0.0475,
+      "step": 10902
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001805167858995759,
+      "loss": 0.986,
+      "step": 10903
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001804982009633675,
+      "loss": 0.0083,
+      "step": 10904
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018047961553877624,
+      "loss": 0.2544,
+      "step": 10905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001804610296260996,
+      "loss": 0.0071,
+      "step": 10906
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018044244322563534,
+      "loss": 0.0027,
+      "step": 10907
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018042385633768103,
+      "loss": 0.5346,
+      "step": 10908
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001804052689625343,
+      "loss": 1.2685,
+      "step": 10909
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018038668110049284,
+      "loss": 0.2247,
+      "step": 10910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001803680927518543,
+      "loss": 0.2753,
+      "step": 10911
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018034950391691628,
+      "loss": 0.6494,
+      "step": 10912
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018033091459597654,
+      "loss": 0.1754,
+      "step": 10913
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018031232478933276,
+      "loss": 0.3959,
+      "step": 10914
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018029373449728248,
+      "loss": 0.3919,
+      "step": 10915
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018027514372012362,
+      "loss": 0.0019,
+      "step": 10916
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018025655245815372,
+      "loss": 0.2555,
+      "step": 10917
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018023796071167054,
+      "loss": 0.3928,
+      "step": 10918
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018021936848097183,
+      "loss": 0.2813,
+      "step": 10919
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018020077576635525,
+      "loss": 0.4815,
+      "step": 10920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018018218256811856,
+      "loss": 2.5258,
+      "step": 10921
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018016358888655953,
+      "loss": 0.1242,
+      "step": 10922
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001801449947219759,
+      "loss": 0.2859,
+      "step": 10923
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018012640007466544,
+      "loss": 0.3337,
+      "step": 10924
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001801078049449259,
+      "loss": 0.0026,
+      "step": 10925
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000180089209333055,
+      "loss": 0.2137,
+      "step": 10926
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018007061323935058,
+      "loss": 0.6979,
+      "step": 10927
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018005201666411048,
+      "loss": 0.6927,
+      "step": 10928
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00018003341960763237,
+      "loss": 1.5199,
+      "step": 10929
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001800148220702141,
+      "loss": 0.8671,
+      "step": 10930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017999622405215357,
+      "loss": 1.7826,
+      "step": 10931
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001799776255537485,
+      "loss": 0.2686,
+      "step": 10932
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017995902657529675,
+      "loss": 0.401,
+      "step": 10933
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017994042711709616,
+      "loss": 0.3456,
+      "step": 10934
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017992182717944453,
+      "loss": 0.8192,
+      "step": 10935
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017990322676263978,
+      "loss": 0.585,
+      "step": 10936
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001798846258669797,
+      "loss": 0.615,
+      "step": 10937
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017986602449276222,
+      "loss": 0.4986,
+      "step": 10938
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017984742264028517,
+      "loss": 0.0318,
+      "step": 10939
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017982882030984643,
+      "loss": 0.4296,
+      "step": 10940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017981021750174388,
+      "loss": 0.092,
+      "step": 10941
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017979161421627547,
+      "loss": 0.599,
+      "step": 10942
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001797730104537391,
+      "loss": 0.4,
+      "step": 10943
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001797544062144326,
+      "loss": 0.5289,
+      "step": 10944
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017973580149865394,
+      "loss": 0.3373,
+      "step": 10945
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017971719630670104,
+      "loss": 0.003,
+      "step": 10946
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017969859063887188,
+      "loss": 0.0081,
+      "step": 10947
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017967998449546437,
+      "loss": 0.64,
+      "step": 10948
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017966137787677638,
+      "loss": 0.849,
+      "step": 10949
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017964277078310597,
+      "loss": 0.1841,
+      "step": 10950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017962416321475112,
+      "loss": 0.3383,
+      "step": 10951
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017960555517200968,
+      "loss": 0.0035,
+      "step": 10952
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017958694665517975,
+      "loss": 0.5361,
+      "step": 10953
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001795683376645593,
+      "loss": 0.91,
+      "step": 10954
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017954972820044627,
+      "loss": 1.1063,
+      "step": 10955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017953111826313865,
+      "loss": 0.0061,
+      "step": 10956
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017951250785293456,
+      "loss": 0.0015,
+      "step": 10957
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017949389697013193,
+      "loss": 0.319,
+      "step": 10958
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017947528561502876,
+      "loss": 0.8697,
+      "step": 10959
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017945667378792317,
+      "loss": 0.297,
+      "step": 10960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017943806148911312,
+      "loss": 0.3234,
+      "step": 10961
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017941944871889675,
+      "loss": 3.3981,
+      "step": 10962
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017940083547757202,
+      "loss": 0.2697,
+      "step": 10963
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017938222176543705,
+      "loss": 0.9823,
+      "step": 10964
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017936360758278992,
+      "loss": 0.2661,
+      "step": 10965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017934499292992866,
+      "loss": 0.1019,
+      "step": 10966
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017932637780715136,
+      "loss": 0.2315,
+      "step": 10967
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017930776221475615,
+      "loss": 0.0063,
+      "step": 10968
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001792891461530411,
+      "loss": 0.0106,
+      "step": 10969
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017927052962230433,
+      "loss": 0.3966,
+      "step": 10970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017925191262284395,
+      "loss": 0.0362,
+      "step": 10971
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017923329515495812,
+      "loss": 0.1611,
+      "step": 10972
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017921467721894493,
+      "loss": 0.3981,
+      "step": 10973
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017919605881510257,
+      "loss": 0.3752,
+      "step": 10974
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017917743994372905,
+      "loss": 0.3156,
+      "step": 10975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017915882060512268,
+      "loss": 1.1382,
+      "step": 10976
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017914020079958158,
+      "loss": 0.7768,
+      "step": 10977
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017912158052740387,
+      "loss": 0.1754,
+      "step": 10978
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001791029597888877,
+      "loss": 1.9818,
+      "step": 10979
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001790843385843314,
+      "loss": 0.5966,
+      "step": 10980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017906571691403303,
+      "loss": 1.1225,
+      "step": 10981
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017904709477829082,
+      "loss": 2.0323,
+      "step": 10982
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000179028472177403,
+      "loss": 0.4101,
+      "step": 10983
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017900984911166772,
+      "loss": 0.0043,
+      "step": 10984
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017899122558138328,
+      "loss": 0.0016,
+      "step": 10985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017897260158684784,
+      "loss": 1.0115,
+      "step": 10986
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017895397712835968,
+      "loss": 0.1013,
+      "step": 10987
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017893535220621707,
+      "loss": 0.9787,
+      "step": 10988
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001789167268207182,
+      "loss": 0.0009,
+      "step": 10989
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017889810097216128,
+      "loss": 0.4557,
+      "step": 10990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017887947466084476,
+      "loss": 0.7611,
+      "step": 10991
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017886084788706671,
+      "loss": 0.9258,
+      "step": 10992
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001788422206511255,
+      "loss": 0.475,
+      "step": 10993
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017882359295331947,
+      "loss": 2.7423,
+      "step": 10994
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001788049647939468,
+      "loss": 0.6919,
+      "step": 10995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001787863361733059,
+      "loss": 0.2783,
+      "step": 10996
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017876770709169496,
+      "loss": 0.2234,
+      "step": 10997
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001787490775494124,
+      "loss": 0.5803,
+      "step": 10998
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001787304475467565,
+      "loss": 0.2881,
+      "step": 10999
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017871181708402567,
+      "loss": 2.6667,
+      "step": 11000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017869318616151812,
+      "loss": 0.5855,
+      "step": 11001
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017867455477953219,
+      "loss": 0.719,
+      "step": 11002
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017865592293836641,
+      "loss": 0.3094,
+      "step": 11003
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017863729063831898,
+      "loss": 0.3327,
+      "step": 11004
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001786186578796883,
+      "loss": 0.7458,
+      "step": 11005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017860002466277283,
+      "loss": 0.9742,
+      "step": 11006
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017858139098787085,
+      "loss": 0.0015,
+      "step": 11007
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017856275685528076,
+      "loss": 0.0542,
+      "step": 11008
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000178544122265301,
+      "loss": 0.3753,
+      "step": 11009
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017852548721823,
+      "loss": 0.3924,
+      "step": 11010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017850685171436612,
+      "loss": 0.0045,
+      "step": 11011
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001784882157540078,
+      "loss": 0.2409,
+      "step": 11012
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017846957933745342,
+      "loss": 1.1514,
+      "step": 11013
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017845094246500147,
+      "loss": 0.1213,
+      "step": 11014
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017843230513695042,
+      "loss": 0.5458,
+      "step": 11015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017841366735359859,
+      "loss": 0.4614,
+      "step": 11016
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017839502911524456,
+      "loss": 0.3121,
+      "step": 11017
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017837639042218677,
+      "loss": 0.4273,
+      "step": 11018
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017835775127472367,
+      "loss": 0.5702,
+      "step": 11019
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001783391116731537,
+      "loss": 1.3389,
+      "step": 11020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017832047161777542,
+      "loss": 0.0369,
+      "step": 11021
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017830183110888732,
+      "loss": 0.2963,
+      "step": 11022
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017828319014678786,
+      "loss": 0.2738,
+      "step": 11023
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001782645487317755,
+      "loss": 0.0013,
+      "step": 11024
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017824590686414885,
+      "loss": 0.0063,
+      "step": 11025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017822726454420643,
+      "loss": 0.1965,
+      "step": 11026
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017820862177224668,
+      "loss": 0.7163,
+      "step": 11027
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017818997854856817,
+      "loss": 0.2557,
+      "step": 11028
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001781713348734695,
+      "loss": 0.0085,
+      "step": 11029
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001781526907472492,
+      "loss": 0.2682,
+      "step": 11030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017813404617020577,
+      "loss": 0.7913,
+      "step": 11031
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017811540114263784,
+      "loss": 0.3647,
+      "step": 11032
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017809675566484391,
+      "loss": 0.7586,
+      "step": 11033
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017807810973712265,
+      "loss": 1.1192,
+      "step": 11034
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017805946335977256,
+      "loss": 0.2349,
+      "step": 11035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017804081653309233,
+      "loss": 1.1887,
+      "step": 11036
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017802216925738047,
+      "loss": 0.0037,
+      "step": 11037
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001780035215329357,
+      "loss": 0.4288,
+      "step": 11038
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017798487336005643,
+      "loss": 0.037,
+      "step": 11039
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017796622473904147,
+      "loss": 0.33,
+      "step": 11040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017794757567018944,
+      "loss": 0.5806,
+      "step": 11041
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017792892615379888,
+      "loss": 0.0025,
+      "step": 11042
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001779102761901685,
+      "loss": 0.0543,
+      "step": 11043
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017789162577959693,
+      "loss": 0.2795,
+      "step": 11044
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017787297492238282,
+      "loss": 0.8237,
+      "step": 11045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017785432361882485,
+      "loss": 0.7962,
+      "step": 11046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001778356718692217,
+      "loss": 0.5622,
+      "step": 11047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017781701967387202,
+      "loss": 0.8073,
+      "step": 11048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017779836703307457,
+      "loss": 0.5252,
+      "step": 11049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017777971394712796,
+      "loss": 0.4283,
+      "step": 11050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001777610604163309,
+      "loss": 0.6623,
+      "step": 11051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017774240644098218,
+      "loss": 0.4463,
+      "step": 11052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001777237520213804,
+      "loss": 0.8834,
+      "step": 11053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017770509715782433,
+      "loss": 0.0058,
+      "step": 11054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017768644185061277,
+      "loss": 0.5278,
+      "step": 11055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017766778610004433,
+      "loss": 0.1841,
+      "step": 11056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001776491299064179,
+      "loss": 1.0174,
+      "step": 11057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017763047327003207,
+      "loss": 0.4596,
+      "step": 11058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017761181619118573,
+      "loss": 0.778,
+      "step": 11059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017759315867017756,
+      "loss": 0.6079,
+      "step": 11060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017757450070730638,
+      "loss": 1.3788,
+      "step": 11061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017755584230287093,
+      "loss": 0.0076,
+      "step": 11062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017753718345717005,
+      "loss": 0.3981,
+      "step": 11063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001775185241705025,
+      "loss": 2.2091,
+      "step": 11064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001774998644431671,
+      "loss": 1.3359,
+      "step": 11065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001774812042754626,
+      "loss": 0.5198,
+      "step": 11066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001774625436676879,
+      "loss": 0.4766,
+      "step": 11067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017744388262014174,
+      "loss": 0.3429,
+      "step": 11068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000177425221133123,
+      "loss": 0.0044,
+      "step": 11069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017740655920693053,
+      "loss": 0.1446,
+      "step": 11070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017738789684186313,
+      "loss": 0.3349,
+      "step": 11071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017736923403821963,
+      "loss": 0.5578,
+      "step": 11072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017735057079629895,
+      "loss": 0.0395,
+      "step": 11073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017733190711639995,
+      "loss": 0.3744,
+      "step": 11074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001773132429988215,
+      "loss": 0.0035,
+      "step": 11075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001772945784438624,
+      "loss": 0.2175,
+      "step": 11076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017727591345182157,
+      "loss": 1.7577,
+      "step": 11077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017725724802299797,
+      "loss": 0.4702,
+      "step": 11078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017723858215769043,
+      "loss": 2.0156,
+      "step": 11079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001772199158561979,
+      "loss": 0.8322,
+      "step": 11080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017720124911881923,
+      "loss": 0.4877,
+      "step": 11081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017718258194585342,
+      "loss": 0.643,
+      "step": 11082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017716391433759932,
+      "loss": 0.0034,
+      "step": 11083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001771452462943559,
+      "loss": 0.3107,
+      "step": 11084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017712657781642212,
+      "loss": 0.671,
+      "step": 11085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001771079089040969,
+      "loss": 0.4273,
+      "step": 11086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001770892395576792,
+      "loss": 0.7485,
+      "step": 11087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017707056977746797,
+      "loss": 0.0649,
+      "step": 11088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001770518995637622,
+      "loss": 0.0843,
+      "step": 11089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001770332289168609,
+      "loss": 0.4912,
+      "step": 11090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017701455783706297,
+      "loss": 0.0042,
+      "step": 11091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00017699588632466744,
+      "loss": 0.2811,
+      "step": 11092
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001769772143799733,
+      "loss": 0.6801,
+      "step": 11093
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001769585420032796,
+      "loss": 0.0033,
+      "step": 11094
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017693986919488524,
+      "loss": 0.4963,
+      "step": 11095
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017692119595508934,
+      "loss": 2.0785,
+      "step": 11096
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001769025222841909,
+      "loss": 0.6176,
+      "step": 11097
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017688384818248895,
+      "loss": 0.3136,
+      "step": 11098
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017686517365028256,
+      "loss": 0.4483,
+      "step": 11099
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001768464986878706,
+      "loss": 0.2043,
+      "step": 11100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017682782329555237,
+      "loss": 0.5281,
+      "step": 11101
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017680914747362676,
+      "loss": 0.0041,
+      "step": 11102
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017679047122239286,
+      "loss": 0.1001,
+      "step": 11103
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017677179454214984,
+      "loss": 0.5336,
+      "step": 11104
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001767531174331967,
+      "loss": 0.0026,
+      "step": 11105
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001767344398958325,
+      "loss": 0.3789,
+      "step": 11106
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017671576193035638,
+      "loss": 2.4629,
+      "step": 11107
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017669708353706745,
+      "loss": 0.6677,
+      "step": 11108
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017667840471626478,
+      "loss": 0.0018,
+      "step": 11109
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001766597254682475,
+      "loss": 0.1873,
+      "step": 11110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001766410457933147,
+      "loss": 0.0025,
+      "step": 11111
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017662236569176553,
+      "loss": 0.6184,
+      "step": 11112
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001766036851638992,
+      "loss": 0.4008,
+      "step": 11113
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017658500421001473,
+      "loss": 0.8572,
+      "step": 11114
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017656632283041126,
+      "loss": 0.4685,
+      "step": 11115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017654764102538811,
+      "loss": 0.4352,
+      "step": 11116
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017652895879524428,
+      "loss": 0.3993,
+      "step": 11117
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017651027614027897,
+      "loss": 0.1686,
+      "step": 11118
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017649159306079137,
+      "loss": 1.3772,
+      "step": 11119
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001764729095570807,
+      "loss": 0.0028,
+      "step": 11120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017645422562944608,
+      "loss": 0.3916,
+      "step": 11121
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017643554127818678,
+      "loss": 0.2238,
+      "step": 11122
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017641685650360194,
+      "loss": 0.095,
+      "step": 11123
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001763981713059908,
+      "loss": 1.0231,
+      "step": 11124
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017637948568565256,
+      "loss": 0.0046,
+      "step": 11125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001763607996428864,
+      "loss": 0.2736,
+      "step": 11126
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017634211317799168,
+      "loss": 0.8612,
+      "step": 11127
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017632342629126752,
+      "loss": 1.0108,
+      "step": 11128
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017630473898301317,
+      "loss": 0.6823,
+      "step": 11129
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017628605125352792,
+      "loss": 0.6202,
+      "step": 11130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017626736310311097,
+      "loss": 0.2833,
+      "step": 11131
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001762486745320617,
+      "loss": 1.8417,
+      "step": 11132
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017622998554067923,
+      "loss": 0.3999,
+      "step": 11133
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017621129612926294,
+      "loss": 0.3438,
+      "step": 11134
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017619260629811207,
+      "loss": 0.2538,
+      "step": 11135
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017617391604752596,
+      "loss": 0.7775,
+      "step": 11136
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001761552253778038,
+      "loss": 0.6106,
+      "step": 11137
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000176136534289245,
+      "loss": 0.6111,
+      "step": 11138
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017611784278214883,
+      "loss": 0.3612,
+      "step": 11139
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017609915085681457,
+      "loss": 0.198,
+      "step": 11140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017608045851354157,
+      "loss": 0.7212,
+      "step": 11141
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017606176575262923,
+      "loss": 0.1399,
+      "step": 11142
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017604307257437681,
+      "loss": 0.3208,
+      "step": 11143
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017602437897908365,
+      "loss": 0.429,
+      "step": 11144
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001760056849670491,
+      "loss": 0.9761,
+      "step": 11145
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017598699053857257,
+      "loss": 1.1709,
+      "step": 11146
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017596829569395335,
+      "loss": 0.9104,
+      "step": 11147
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001759496004334909,
+      "loss": 0.1859,
+      "step": 11148
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017593090475748449,
+      "loss": 0.0008,
+      "step": 11149
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017591220866623356,
+      "loss": 0.5564,
+      "step": 11150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017589351216003757,
+      "loss": 0.5463,
+      "step": 11151
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017587481523919575,
+      "loss": 0.6097,
+      "step": 11152
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017585611790400766,
+      "loss": 0.0014,
+      "step": 11153
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017583742015477263,
+      "loss": 0.5151,
+      "step": 11154
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001758187219917901,
+      "loss": 0.4997,
+      "step": 11155
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017580002341535946,
+      "loss": 2.709,
+      "step": 11156
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017578132442578015,
+      "loss": 0.2454,
+      "step": 11157
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017576262502335167,
+      "loss": 0.4615,
+      "step": 11158
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017574392520837342,
+      "loss": 0.4452,
+      "step": 11159
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001757252249811448,
+      "loss": 1.1819,
+      "step": 11160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001757065243419653,
+      "loss": 0.2673,
+      "step": 11161
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017568782329113446,
+      "loss": 0.0086,
+      "step": 11162
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017566912182895165,
+      "loss": 0.9744,
+      "step": 11163
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017565041995571632,
+      "loss": 0.3296,
+      "step": 11164
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001756317176717281,
+      "loss": 0.7302,
+      "step": 11165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017561301497728637,
+      "loss": 0.2883,
+      "step": 11166
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017559431187269057,
+      "loss": 0.3449,
+      "step": 11167
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017557560835824036,
+      "loss": 0.3929,
+      "step": 11168
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017555690443423516,
+      "loss": 0.5214,
+      "step": 11169
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017553820010097446,
+      "loss": 0.1134,
+      "step": 11170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017551949535875782,
+      "loss": 0.4971,
+      "step": 11171
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017550079020788473,
+      "loss": 0.2857,
+      "step": 11172
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001754820846486548,
+      "loss": 0.2505,
+      "step": 11173
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017546337868136756,
+      "loss": 0.6878,
+      "step": 11174
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017544467230632243,
+      "loss": 4.5127,
+      "step": 11175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017542596552381915,
+      "loss": 0.7272,
+      "step": 11176
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017540725833415716,
+      "loss": 0.3914,
+      "step": 11177
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017538855073763603,
+      "loss": 0.605,
+      "step": 11178
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017536984273455538,
+      "loss": 0.4873,
+      "step": 11179
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001753511343252148,
+      "loss": 0.5667,
+      "step": 11180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001753324255099138,
+      "loss": 0.9228,
+      "step": 11181
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017531371628895205,
+      "loss": 0.3135,
+      "step": 11182
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017529500666262914,
+      "loss": 0.8742,
+      "step": 11183
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017527629663124464,
+      "loss": 0.8858,
+      "step": 11184
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017525758619509825,
+      "loss": 0.1255,
+      "step": 11185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017523887535448942,
+      "loss": 0.98,
+      "step": 11186
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017522016410971791,
+      "loss": 0.2613,
+      "step": 11187
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017520145246108338,
+      "loss": 0.054,
+      "step": 11188
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017518274040888537,
+      "loss": 0.5707,
+      "step": 11189
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017516402795342352,
+      "loss": 0.9155,
+      "step": 11190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001751453150949976,
+      "loss": 0.3323,
+      "step": 11191
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017512660183390718,
+      "loss": 0.0073,
+      "step": 11192
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017510788817045193,
+      "loss": 0.6789,
+      "step": 11193
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017508917410493153,
+      "loss": 0.4919,
+      "step": 11194
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017507045963764564,
+      "loss": 0.0088,
+      "step": 11195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017505174476889398,
+      "loss": 0.3338,
+      "step": 11196
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017503302949897626,
+      "loss": 0.2481,
+      "step": 11197
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017501431382819209,
+      "loss": 0.8812,
+      "step": 11198
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017499559775684126,
+      "loss": 0.0262,
+      "step": 11199
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017497688128522348,
+      "loss": 0.3546,
+      "step": 11200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017495816441363832,
+      "loss": 0.1888,
+      "step": 11201
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001749394471423857,
+      "loss": 0.2384,
+      "step": 11202
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017492072947176526,
+      "loss": 0.2821,
+      "step": 11203
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017490201140207674,
+      "loss": 0.9567,
+      "step": 11204
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017488329293361982,
+      "loss": 0.2854,
+      "step": 11205
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017486457406669435,
+      "loss": 0.8414,
+      "step": 11206
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017484585480160005,
+      "loss": 0.7201,
+      "step": 11207
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017482713513863668,
+      "loss": 0.5046,
+      "step": 11208
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017480841507810398,
+      "loss": 0.7478,
+      "step": 11209
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017478969462030174,
+      "loss": 0.3715,
+      "step": 11210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001747709737655298,
+      "loss": 0.3416,
+      "step": 11211
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017475225251408784,
+      "loss": 0.3555,
+      "step": 11212
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017473353086627568,
+      "loss": 0.5695,
+      "step": 11213
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017471480882239322,
+      "loss": 0.7797,
+      "step": 11214
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017469608638274014,
+      "loss": 0.6786,
+      "step": 11215
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017467736354761625,
+      "loss": 0.5546,
+      "step": 11216
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001746586403173215,
+      "loss": 0.0909,
+      "step": 11217
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017463991669215562,
+      "loss": 0.5176,
+      "step": 11218
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017462119267241843,
+      "loss": 0.3171,
+      "step": 11219
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001746024682584098,
+      "loss": 0.7376,
+      "step": 11220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017458374345042955,
+      "loss": 0.2021,
+      "step": 11221
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017456501824877752,
+      "loss": 0.1359,
+      "step": 11222
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001745462926537537,
+      "loss": 0.6381,
+      "step": 11223
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017452756666565773,
+      "loss": 0.7888,
+      "step": 11224
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017450884028478962,
+      "loss": 0.4312,
+      "step": 11225
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017449011351144925,
+      "loss": 0.222,
+      "step": 11226
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017447138634593643,
+      "loss": 1.4936,
+      "step": 11227
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017445265878855113,
+      "loss": 0.8488,
+      "step": 11228
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001744339308395931,
+      "loss": 0.6904,
+      "step": 11229
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017441520249936244,
+      "loss": 0.3021,
+      "step": 11230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001743964737681589,
+      "loss": 1.242,
+      "step": 11231
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017437774464628244,
+      "loss": 0.4001,
+      "step": 11232
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000174359015134033,
+      "loss": 0.7659,
+      "step": 11233
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017434028523171053,
+      "loss": 0.2023,
+      "step": 11234
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017432155493961486,
+      "loss": 1.7289,
+      "step": 11235
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017430282425804603,
+      "loss": 0.3935,
+      "step": 11236
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017428409318730398,
+      "loss": 0.279,
+      "step": 11237
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017426536172768855,
+      "loss": 0.1267,
+      "step": 11238
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017424662987949977,
+      "loss": 0.0102,
+      "step": 11239
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017422789764303767,
+      "loss": 0.1772,
+      "step": 11240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001742091650186021,
+      "loss": 0.3952,
+      "step": 11241
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017419043200649306,
+      "loss": 0.9961,
+      "step": 11242
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001741716986070106,
+      "loss": 0.0856,
+      "step": 11243
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017415296482045463,
+      "loss": 0.5177,
+      "step": 11244
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017413423064712517,
+      "loss": 0.6685,
+      "step": 11245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001741154960873223,
+      "loss": 0.8316,
+      "step": 11246
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017409676114134588,
+      "loss": 0.2912,
+      "step": 11247
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017407802580949602,
+      "loss": 0.4793,
+      "step": 11248
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017405929009207277,
+      "loss": 0.6618,
+      "step": 11249
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017404055398937597,
+      "loss": 0.1485,
+      "step": 11250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017402181750170585,
+      "loss": 0.569,
+      "step": 11251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017400308062936242,
+      "loss": 0.5483,
+      "step": 11252
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017398434337264563,
+      "loss": 0.9323,
+      "step": 11253
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017396560573185556,
+      "loss": 0.5695,
+      "step": 11254
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017394686770729227,
+      "loss": 0.3265,
+      "step": 11255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017392812929925588,
+      "loss": 0.413,
+      "step": 11256
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017390939050804637,
+      "loss": 0.3497,
+      "step": 11257
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017389065133396387,
+      "loss": 0.9483,
+      "step": 11258
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017387191177730844,
+      "loss": 0.2141,
+      "step": 11259
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001738531718383802,
+      "loss": 0.8493,
+      "step": 11260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001738344315174792,
+      "loss": 0.5232,
+      "step": 11261
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001738156908149055,
+      "loss": 0.0023,
+      "step": 11262
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017379694973095934,
+      "loss": 0.5246,
+      "step": 11263
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001737782082659407,
+      "loss": 0.6238,
+      "step": 11264
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017375946642014968,
+      "loss": 0.8134,
+      "step": 11265
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017374072419388655,
+      "loss": 0.4835,
+      "step": 11266
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017372198158745132,
+      "loss": 0.0046,
+      "step": 11267
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017370323860114417,
+      "loss": 0.354,
+      "step": 11268
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017368449523526524,
+      "loss": 0.9506,
+      "step": 11269
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017366575149011461,
+      "loss": 0.7111,
+      "step": 11270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001736470073659925,
+      "loss": 0.572,
+      "step": 11271
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017362826286319914,
+      "loss": 0.701,
+      "step": 11272
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017360951798203451,
+      "loss": 1.0215,
+      "step": 11273
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001735907727227989,
+      "loss": 0.2617,
+      "step": 11274
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001735720270857925,
+      "loss": 0.9565,
+      "step": 11275
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017355328107131546,
+      "loss": 0.5516,
+      "step": 11276
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017353453467966793,
+      "loss": 0.1968,
+      "step": 11277
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001735157879111502,
+      "loss": 0.7744,
+      "step": 11278
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017349704076606242,
+      "loss": 0.0052,
+      "step": 11279
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017347829324470477,
+      "loss": 0.833,
+      "step": 11280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001734595453473775,
+      "loss": 0.9849,
+      "step": 11281
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001734407970743808,
+      "loss": 0.5683,
+      "step": 11282
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017342204842601494,
+      "loss": 0.3259,
+      "step": 11283
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001734032994025801,
+      "loss": 0.0043,
+      "step": 11284
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017338455000437655,
+      "loss": 0.5168,
+      "step": 11285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017336580023170458,
+      "loss": 0.3653,
+      "step": 11286
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017334705008486434,
+      "loss": 0.9485,
+      "step": 11287
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017332829956415608,
+      "loss": 0.1167,
+      "step": 11288
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017330954866988022,
+      "loss": 0.8442,
+      "step": 11289
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017329079740233686,
+      "loss": 0.4286,
+      "step": 11290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017327204576182633,
+      "loss": 0.5797,
+      "step": 11291
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017325329374864892,
+      "loss": 1.1943,
+      "step": 11292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017323454136310493,
+      "loss": 0.2747,
+      "step": 11293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001732157886054946,
+      "loss": 0.3275,
+      "step": 11294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017319703547611832,
+      "loss": 0.4431,
+      "step": 11295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017317828197527623,
+      "loss": 0.2281,
+      "step": 11296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001731595281032688,
+      "loss": 0.7793,
+      "step": 11297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001731407738603963,
+      "loss": 0.5556,
+      "step": 11298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017312201924695898,
+      "loss": 0.3464,
+      "step": 11299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017310326426325727,
+      "loss": 0.7021,
+      "step": 11300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017308450890959147,
+      "loss": 0.0051,
+      "step": 11301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017306575318626186,
+      "loss": 0.2778,
+      "step": 11302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017304699709356884,
+      "loss": 0.6113,
+      "step": 11303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017302824063181277,
+      "loss": 0.0071,
+      "step": 11304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017300948380129396,
+      "loss": 0.9708,
+      "step": 11305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017299072660231283,
+      "loss": 0.3971,
+      "step": 11306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001729719690351697,
+      "loss": 0.5388,
+      "step": 11307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017295321110016496,
+      "loss": 0.507,
+      "step": 11308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017293445279759897,
+      "loss": 0.485,
+      "step": 11309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001729156941277722,
+      "loss": 0.2792,
+      "step": 11310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017289693509098492,
+      "loss": 0.2317,
+      "step": 11311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017287817568753763,
+      "loss": 0.2538,
+      "step": 11312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017285941591773068,
+      "loss": 2.2159,
+      "step": 11313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017284065578186445,
+      "loss": 1.832,
+      "step": 11314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017282189528023946,
+      "loss": 0.7313,
+      "step": 11315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017280313441315603,
+      "loss": 0.124,
+      "step": 11316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017278437318091463,
+      "loss": 0.2443,
+      "step": 11317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001727656115838157,
+      "loss": 0.0239,
+      "step": 11318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017274684962215964,
+      "loss": 0.2199,
+      "step": 11319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017272808729624698,
+      "loss": 0.6886,
+      "step": 11320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017270932460637808,
+      "loss": 0.6004,
+      "step": 11321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017269056155285337,
+      "loss": 0.3414,
+      "step": 11322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017267179813597344,
+      "loss": 1.1344,
+      "step": 11323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001726530343560387,
+      "loss": 0.2027,
+      "step": 11324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017263427021334957,
+      "loss": 0.3438,
+      "step": 11325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017261550570820658,
+      "loss": 0.0011,
+      "step": 11326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017259674084091022,
+      "loss": 0.4644,
+      "step": 11327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017257797561176096,
+      "loss": 0.0029,
+      "step": 11328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001725592100210593,
+      "loss": 0.8366,
+      "step": 11329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017254044406910576,
+      "loss": 0.3631,
+      "step": 11330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001725216777562008,
+      "loss": 0.5366,
+      "step": 11331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000172502911082645,
+      "loss": 0.3473,
+      "step": 11332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017248414404873888,
+      "loss": 0.6082,
+      "step": 11333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001724653766547829,
+      "loss": 0.334,
+      "step": 11334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017244660890107765,
+      "loss": 0.2224,
+      "step": 11335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017242784078792364,
+      "loss": 1.2532,
+      "step": 11336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001724090723156214,
+      "loss": 0.5513,
+      "step": 11337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017239030348447154,
+      "loss": 0.3571,
+      "step": 11338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017237153429477451,
+      "loss": 0.5467,
+      "step": 11339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017235276474683098,
+      "loss": 1.1258,
+      "step": 11340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00017233399484094147,
+      "loss": 1.6953,
+      "step": 11341
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017231522457740654,
+      "loss": 0.2987,
+      "step": 11342
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017229645395652678,
+      "loss": 0.2093,
+      "step": 11343
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017227768297860274,
+      "loss": 0.2361,
+      "step": 11344
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017225891164393507,
+      "loss": 0.5115,
+      "step": 11345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017224013995282433,
+      "loss": 0.6408,
+      "step": 11346
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017222136790557118,
+      "loss": 0.2033,
+      "step": 11347
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017220259550247607,
+      "loss": 0.3438,
+      "step": 11348
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001721838227438398,
+      "loss": 0.4076,
+      "step": 11349
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017216504962996288,
+      "loss": 0.9188,
+      "step": 11350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017214627616114593,
+      "loss": 0.1673,
+      "step": 11351
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001721275023376896,
+      "loss": 0.2708,
+      "step": 11352
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017210872815989458,
+      "loss": 0.4844,
+      "step": 11353
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017208995362806142,
+      "loss": 2.6055,
+      "step": 11354
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017207117874249085,
+      "loss": 0.2589,
+      "step": 11355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017205240350348343,
+      "loss": 0.8835,
+      "step": 11356
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017203362791133988,
+      "loss": 0.0032,
+      "step": 11357
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001720148519663608,
+      "loss": 0.6651,
+      "step": 11358
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000171996075668847,
+      "loss": 0.1565,
+      "step": 11359
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017197729901909895,
+      "loss": 0.8889,
+      "step": 11360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017195852201741752,
+      "loss": 0.5572,
+      "step": 11361
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001719397446641033,
+      "loss": 0.0035,
+      "step": 11362
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017192096695945697,
+      "loss": 0.6925,
+      "step": 11363
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001719021889037793,
+      "loss": 0.4874,
+      "step": 11364
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001718834104973709,
+      "loss": 1.4131,
+      "step": 11365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017186463174053256,
+      "loss": 0.2784,
+      "step": 11366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017184585263356494,
+      "loss": 0.2561,
+      "step": 11367
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017182707317676873,
+      "loss": 0.1002,
+      "step": 11368
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017180829337044476,
+      "loss": 0.8682,
+      "step": 11369
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017178951321489372,
+      "loss": 0.706,
+      "step": 11370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017177073271041626,
+      "loss": 0.003,
+      "step": 11371
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017175195185731323,
+      "loss": 0.2841,
+      "step": 11372
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017173317065588534,
+      "loss": 0.3584,
+      "step": 11373
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017171438910643333,
+      "loss": 0.6256,
+      "step": 11374
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017169560720925795,
+      "loss": 0.1862,
+      "step": 11375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017167682496465998,
+      "loss": 0.007,
+      "step": 11376
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017165804237294021,
+      "loss": 0.3928,
+      "step": 11377
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017163925943439935,
+      "loss": 0.7229,
+      "step": 11378
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001716204761493383,
+      "loss": 0.1526,
+      "step": 11379
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017160169251805775,
+      "loss": 0.4282,
+      "step": 11380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001715829085408585,
+      "loss": 0.4185,
+      "step": 11381
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017156412421804134,
+      "loss": 0.1592,
+      "step": 11382
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017154533954990712,
+      "loss": 0.2652,
+      "step": 11383
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017152655453675663,
+      "loss": 0.4988,
+      "step": 11384
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017150776917889072,
+      "loss": 0.4114,
+      "step": 11385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017148898347661004,
+      "loss": 0.2028,
+      "step": 11386
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017147019743021563,
+      "loss": 1.1008,
+      "step": 11387
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017145141104000822,
+      "loss": 0.4762,
+      "step": 11388
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017143262430628867,
+      "loss": 0.5033,
+      "step": 11389
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017141383722935776,
+      "loss": 0.1117,
+      "step": 11390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001713950498095164,
+      "loss": 0.2967,
+      "step": 11391
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017137626204706544,
+      "loss": 0.1304,
+      "step": 11392
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017135747394230575,
+      "loss": 0.1179,
+      "step": 11393
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017133868549553812,
+      "loss": 0.4985,
+      "step": 11394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017131989670706347,
+      "loss": 0.2655,
+      "step": 11395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017130110757718274,
+      "loss": 0.5979,
+      "step": 11396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017128231810619663,
+      "loss": 0.3286,
+      "step": 11397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001712635282944062,
+      "loss": 0.3908,
+      "step": 11398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001712447381421123,
+      "loss": 0.2598,
+      "step": 11399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017122594764961576,
+      "loss": 0.2793,
+      "step": 11400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001712071568172175,
+      "loss": 2.168,
+      "step": 11401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017118836564521855,
+      "loss": 0.9336,
+      "step": 11402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017116957413391962,
+      "loss": 0.3906,
+      "step": 11403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017115078228362176,
+      "loss": 0.5769,
+      "step": 11404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017113199009462587,
+      "loss": 0.0076,
+      "step": 11405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017111319756723292,
+      "loss": 0.2778,
+      "step": 11406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017109440470174374,
+      "loss": 0.1868,
+      "step": 11407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017107561149845937,
+      "loss": 0.6178,
+      "step": 11408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017105681795768067,
+      "loss": 0.1376,
+      "step": 11409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017103802407970862,
+      "loss": 0.3285,
+      "step": 11410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017101922986484426,
+      "loss": 0.6133,
+      "step": 11411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001710004353133884,
+      "loss": 0.3397,
+      "step": 11412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017098164042564213,
+      "loss": 0.7638,
+      "step": 11413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017096284520190638,
+      "loss": 0.341,
+      "step": 11414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017094404964248212,
+      "loss": 0.0575,
+      "step": 11415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017092525374767033,
+      "loss": 0.3515,
+      "step": 11416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017090645751777198,
+      "loss": 0.1897,
+      "step": 11417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017088766095308812,
+      "loss": 0.2541,
+      "step": 11418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001708688640539197,
+      "loss": 1.3683,
+      "step": 11419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017085006682056778,
+      "loss": 0.1509,
+      "step": 11420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017083126925333328,
+      "loss": 0.4679,
+      "step": 11421
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017081247135251735,
+      "loss": 0.0054,
+      "step": 11422
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017079367311842084,
+      "loss": 0.0101,
+      "step": 11423
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017077487455134486,
+      "loss": 0.5751,
+      "step": 11424
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017075607565159053,
+      "loss": 0.3439,
+      "step": 11425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017073727641945874,
+      "loss": 0.8758,
+      "step": 11426
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017071847685525059,
+      "loss": 0.9097,
+      "step": 11427
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017069967695926714,
+      "loss": 0.7581,
+      "step": 11428
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001706808767318094,
+      "loss": 0.9685,
+      "step": 11429
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001706620761731785,
+      "loss": 1.2163,
+      "step": 11430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001706432752836754,
+      "loss": 0.8364,
+      "step": 11431
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017062447406360128,
+      "loss": 0.5834,
+      "step": 11432
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017060567251325713,
+      "loss": 0.2081,
+      "step": 11433
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017058687063294413,
+      "loss": 0.0251,
+      "step": 11434
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001705680684229632,
+      "loss": 0.3897,
+      "step": 11435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017054926588361554,
+      "loss": 0.3517,
+      "step": 11436
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001705304630152023,
+      "loss": 0.217,
+      "step": 11437
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017051165981802444,
+      "loss": 0.6293,
+      "step": 11438
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017049285629238314,
+      "loss": 0.5738,
+      "step": 11439
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001704740524385795,
+      "loss": 0.3992,
+      "step": 11440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017045524825691467,
+      "loss": 0.2469,
+      "step": 11441
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001704364437476897,
+      "loss": 0.0021,
+      "step": 11442
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017041763891120575,
+      "loss": 0.0024,
+      "step": 11443
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000170398833747764,
+      "loss": 3.004,
+      "step": 11444
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017038002825766547,
+      "loss": 0.3182,
+      "step": 11445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017036122244121142,
+      "loss": 0.5232,
+      "step": 11446
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017034241629870298,
+      "loss": 0.1432,
+      "step": 11447
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017032360983044125,
+      "loss": 0.6262,
+      "step": 11448
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017030480303672742,
+      "loss": 0.2832,
+      "step": 11449
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017028599591786256,
+      "loss": 1.0324,
+      "step": 11450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017026718847414803,
+      "loss": 0.7588,
+      "step": 11451
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017024838070588485,
+      "loss": 0.0315,
+      "step": 11452
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017022957261337424,
+      "loss": 0.5231,
+      "step": 11453
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017021076419691737,
+      "loss": 1.0724,
+      "step": 11454
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017019195545681546,
+      "loss": 0.1754,
+      "step": 11455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017017314639336967,
+      "loss": 0.2297,
+      "step": 11456
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001701543370068813,
+      "loss": 0.4557,
+      "step": 11457
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017013552729765134,
+      "loss": 0.251,
+      "step": 11458
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001701167172659812,
+      "loss": 0.8681,
+      "step": 11459
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017009790691217204,
+      "loss": 0.341,
+      "step": 11460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017007909623652502,
+      "loss": 0.1599,
+      "step": 11461
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017006028523934143,
+      "loss": 0.6662,
+      "step": 11462
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017004147392092248,
+      "loss": 0.7385,
+      "step": 11463
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001700226622815694,
+      "loss": 0.5619,
+      "step": 11464
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00017000385032158344,
+      "loss": 0.2779,
+      "step": 11465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016998503804126582,
+      "loss": 0.3066,
+      "step": 11466
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016996622544091782,
+      "loss": 0.0524,
+      "step": 11467
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001699474125208407,
+      "loss": 0.8836,
+      "step": 11468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016992859928133567,
+      "loss": 0.0039,
+      "step": 11469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016990978572270404,
+      "loss": 0.0317,
+      "step": 11470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016989097184524711,
+      "loss": 0.2886,
+      "step": 11471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016987215764926612,
+      "loss": 0.5201,
+      "step": 11472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016985334313506228,
+      "loss": 0.4142,
+      "step": 11473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000169834528302937,
+      "loss": 0.0242,
+      "step": 11474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001698157131531915,
+      "loss": 1.6837,
+      "step": 11475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016979689768612706,
+      "loss": 0.4292,
+      "step": 11476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016977808190204508,
+      "loss": 0.8759,
+      "step": 11477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016975926580124677,
+      "loss": 1.3409,
+      "step": 11478
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016974044938403347,
+      "loss": 0.2725,
+      "step": 11479
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016972163265070652,
+      "loss": 0.1803,
+      "step": 11480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016970281560156717,
+      "loss": 0.1699,
+      "step": 11481
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016968399823691683,
+      "loss": 0.6152,
+      "step": 11482
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016966518055705685,
+      "loss": 0.7694,
+      "step": 11483
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016964636256228845,
+      "loss": 0.5982,
+      "step": 11484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016962754425291305,
+      "loss": 0.736,
+      "step": 11485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016960872562923201,
+      "loss": 0.3002,
+      "step": 11486
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016958990669154662,
+      "loss": 0.625,
+      "step": 11487
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016957108744015828,
+      "loss": 0.2759,
+      "step": 11488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001695522678753684,
+      "loss": 0.7451,
+      "step": 11489
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016953344799747823,
+      "loss": 0.7166,
+      "step": 11490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016951462780678925,
+      "loss": 1.2163,
+      "step": 11491
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016949580730360277,
+      "loss": 3.9286,
+      "step": 11492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694769864882202,
+      "loss": 1.2188,
+      "step": 11493
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694581653609429,
+      "loss": 0.2267,
+      "step": 11494
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016943934392207231,
+      "loss": 0.734,
+      "step": 11495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694205221719098,
+      "loss": 0.7526,
+      "step": 11496
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001694017001107568,
+      "loss": 0.2254,
+      "step": 11497
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016938287773891468,
+      "loss": 0.3444,
+      "step": 11498
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001693640550566848,
+      "loss": 0.0339,
+      "step": 11499
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001693452320643687,
+      "loss": 0.022,
+      "step": 11500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016932640876226774,
+      "loss": 0.506,
+      "step": 11501
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016930758515068336,
+      "loss": 0.1392,
+      "step": 11502
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001692887612299169,
+      "loss": 0.3018,
+      "step": 11503
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016926993700026996,
+      "loss": 0.2467,
+      "step": 11504
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016925111246204384,
+      "loss": 0.9027,
+      "step": 11505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016923228761554012,
+      "loss": 0.0062,
+      "step": 11506
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001692134624610601,
+      "loss": 0.3019,
+      "step": 11507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016919463699890534,
+      "loss": 0.5807,
+      "step": 11508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001691758112293773,
+      "loss": 0.3312,
+      "step": 11509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001691569851527774,
+      "loss": 0.004,
+      "step": 11510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016913815876940715,
+      "loss": 0.1951,
+      "step": 11511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016911933207956802,
+      "loss": 1.5522,
+      "step": 11512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016910050508356145,
+      "loss": 0.0035,
+      "step": 11513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016908167778168892,
+      "loss": 0.7723,
+      "step": 11514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016906285017425206,
+      "loss": 0.8097,
+      "step": 11515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016904402226155217,
+      "loss": 1.4436,
+      "step": 11516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001690251940438909,
+      "loss": 0.1792,
+      "step": 11517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016900636552156966,
+      "loss": 0.8725,
+      "step": 11518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016898753669489008,
+      "loss": 0.4447,
+      "step": 11519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001689687075641535,
+      "loss": 0.619,
+      "step": 11520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016894987812966162,
+      "loss": 0.0036,
+      "step": 11521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001689310483917158,
+      "loss": 2.9376,
+      "step": 11522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016891221835061774,
+      "loss": 0.3678,
+      "step": 11523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016889338800666882,
+      "loss": 1.1475,
+      "step": 11524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001688745573601706,
+      "loss": 0.5922,
+      "step": 11525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016885572641142478,
+      "loss": 0.0078,
+      "step": 11526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016883689516073276,
+      "loss": 0.4765,
+      "step": 11527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001688180636083961,
+      "loss": 0.0424,
+      "step": 11528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016879923175471646,
+      "loss": 0.0082,
+      "step": 11529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016878039959999529,
+      "loss": 0.0163,
+      "step": 11530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016876156714453418,
+      "loss": 0.1069,
+      "step": 11531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001687427343886348,
+      "loss": 0.2811,
+      "step": 11532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001687239013325986,
+      "loss": 1.2421,
+      "step": 11533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016870506797672723,
+      "loss": 1.5838,
+      "step": 11534
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001686862343213223,
+      "loss": 0.111,
+      "step": 11535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001686674003666854,
+      "loss": 1.096,
+      "step": 11536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016864856611311797,
+      "loss": 1.1409,
+      "step": 11537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016862973156092187,
+      "loss": 1.013,
+      "step": 11538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016861089671039853,
+      "loss": 1.3809,
+      "step": 11539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001685920615618496,
+      "loss": 0.377,
+      "step": 11540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016857322611557672,
+      "loss": 0.0028,
+      "step": 11541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016855439037188148,
+      "loss": 0.5722,
+      "step": 11542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016853555433106557,
+      "loss": 0.0341,
+      "step": 11543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016851671799343054,
+      "loss": 0.1579,
+      "step": 11544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016849788135927805,
+      "loss": 1.5413,
+      "step": 11545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016847904442890977,
+      "loss": 0.8795,
+      "step": 11546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016846020720262734,
+      "loss": 1.3014,
+      "step": 11547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016844136968073236,
+      "loss": 0.6329,
+      "step": 11548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016842253186352655,
+      "loss": 1.0597,
+      "step": 11549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016840369375131155,
+      "loss": 0.8238,
+      "step": 11550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016838485534438898,
+      "loss": 3.4025,
+      "step": 11551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016836601664306055,
+      "loss": 0.0302,
+      "step": 11552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016834717764762795,
+      "loss": 0.2743,
+      "step": 11553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016832833835839282,
+      "loss": 0.9773,
+      "step": 11554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016830949877565688,
+      "loss": 0.2505,
+      "step": 11555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016829065889972176,
+      "loss": 0.87,
+      "step": 11556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016827181873088919,
+      "loss": 0.5137,
+      "step": 11557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016825297826946094,
+      "loss": 0.2198,
+      "step": 11558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016823413751573853,
+      "loss": 0.4607,
+      "step": 11559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001682152964700238,
+      "loss": 1.3282,
+      "step": 11560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016819645513261848,
+      "loss": 0.2218,
+      "step": 11561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016817761350382423,
+      "loss": 0.4942,
+      "step": 11562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016815877158394272,
+      "loss": 0.0519,
+      "step": 11563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001681399293732758,
+      "loss": 0.1784,
+      "step": 11564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016812108687212513,
+      "loss": 0.0127,
+      "step": 11565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016810224408079244,
+      "loss": 0.3227,
+      "step": 11566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016808340099957946,
+      "loss": 0.3394,
+      "step": 11567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016806455762878798,
+      "loss": 0.6101,
+      "step": 11568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001680457139687197,
+      "loss": 0.0521,
+      "step": 11569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016802687001967644,
+      "loss": 0.5339,
+      "step": 11570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016800802578195983,
+      "loss": 0.2408,
+      "step": 11571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016798918125587173,
+      "loss": 0.678,
+      "step": 11572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016797033644171396,
+      "loss": 0.4232,
+      "step": 11573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016795149133978814,
+      "loss": 0.7215,
+      "step": 11574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001679326459503962,
+      "loss": 0.4797,
+      "step": 11575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016791380027383982,
+      "loss": 0.6304,
+      "step": 11576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016789495431042078,
+      "loss": 0.0979,
+      "step": 11577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016787610806044096,
+      "loss": 1.512,
+      "step": 11578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016785726152420205,
+      "loss": 0.2894,
+      "step": 11579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001678384147020059,
+      "loss": 0.8378,
+      "step": 11580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016781956759415434,
+      "loss": 0.0241,
+      "step": 11581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016780072020094913,
+      "loss": 0.3154,
+      "step": 11582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677818725226921,
+      "loss": 0.6793,
+      "step": 11583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677630245596851,
+      "loss": 2.1945,
+      "step": 11584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677441763122299,
+      "loss": 0.4243,
+      "step": 11585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001677253277806283,
+      "loss": 0.2407,
+      "step": 11586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016770647896518226,
+      "loss": 0.7149,
+      "step": 11587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016768762986619349,
+      "loss": 0.1202,
+      "step": 11588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016766878048396386,
+      "loss": 0.547,
+      "step": 11589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00016764993081879524,
+      "loss": 0.3977,
+      "step": 11590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016763108087098947,
+      "loss": 0.3878,
+      "step": 11591
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001676122306408484,
+      "loss": 0.7307,
+      "step": 11592
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016759338012867388,
+      "loss": 0.3256,
+      "step": 11593
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016757452933476778,
+      "loss": 1.0933,
+      "step": 11594
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016755567825943197,
+      "loss": 0.7523,
+      "step": 11595
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016753682690296836,
+      "loss": 0.2171,
+      "step": 11596
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001675179752656787,
+      "loss": 0.4834,
+      "step": 11597
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016749912334786503,
+      "loss": 0.2111,
+      "step": 11598
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016748027114982917,
+      "loss": 1.0662,
+      "step": 11599
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016746141867187298,
+      "loss": 0.437,
+      "step": 11600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016744256591429837,
+      "loss": 0.4426,
+      "step": 11601
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001674237128774072,
+      "loss": 0.3729,
+      "step": 11602
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001674048595615015,
+      "loss": 0.232,
+      "step": 11603
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016738600596688302,
+      "loss": 0.1609,
+      "step": 11604
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016736715209385379,
+      "loss": 0.2248,
+      "step": 11605
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001673482979427157,
+      "loss": 0.7621,
+      "step": 11606
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001673294435137706,
+      "loss": 2.1316,
+      "step": 11607
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016731058880732052,
+      "loss": 0.5441,
+      "step": 11608
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016729173382366733,
+      "loss": 0.5291,
+      "step": 11609
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000167272878563113,
+      "loss": 0.6452,
+      "step": 11610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016725402302595938,
+      "loss": 0.2095,
+      "step": 11611
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001672351672125085,
+      "loss": 2.6006,
+      "step": 11612
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001672163111230623,
+      "loss": 0.0664,
+      "step": 11613
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016719745475792268,
+      "loss": 0.2883,
+      "step": 11614
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016717859811739167,
+      "loss": 0.5812,
+      "step": 11615
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001671597412017712,
+      "loss": 0.0071,
+      "step": 11616
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016714088401136316,
+      "loss": 0.8524,
+      "step": 11617
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016712202654646966,
+      "loss": 0.6948,
+      "step": 11618
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001671031688073926,
+      "loss": 0.0053,
+      "step": 11619
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016708431079443388,
+      "loss": 0.0026,
+      "step": 11620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016706545250789562,
+      "loss": 0.0147,
+      "step": 11621
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001670465939480798,
+      "loss": 0.0096,
+      "step": 11622
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016702773511528826,
+      "loss": 0.2868,
+      "step": 11623
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016700887600982315,
+      "loss": 0.5168,
+      "step": 11624
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016699001663198642,
+      "loss": 0.4683,
+      "step": 11625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016697115698208008,
+      "loss": 0.2379,
+      "step": 11626
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001669522970604061,
+      "loss": 0.0032,
+      "step": 11627
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016693343686726656,
+      "loss": 0.0096,
+      "step": 11628
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001669145764029634,
+      "loss": 0.0059,
+      "step": 11629
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016689571566779873,
+      "loss": 0.2949,
+      "step": 11630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016687685466207448,
+      "loss": 0.9223,
+      "step": 11631
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016685799338609275,
+      "loss": 0.416,
+      "step": 11632
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001668391318401556,
+      "loss": 0.3045,
+      "step": 11633
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000166820270024565,
+      "loss": 0.3308,
+      "step": 11634
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016680140793962298,
+      "loss": 0.352,
+      "step": 11635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001667825455856317,
+      "loss": 0.0025,
+      "step": 11636
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016676368296289306,
+      "loss": 1.6922,
+      "step": 11637
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016674482007170924,
+      "loss": 0.6762,
+      "step": 11638
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016672595691238223,
+      "loss": 0.0021,
+      "step": 11639
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016670709348521416,
+      "loss": 0.2898,
+      "step": 11640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016668822979050706,
+      "loss": 0.0045,
+      "step": 11641
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016666936582856299,
+      "loss": 0.266,
+      "step": 11642
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000166650501599684,
+      "loss": 2.017,
+      "step": 11643
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016663163710417227,
+      "loss": 0.8524,
+      "step": 11644
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016661277234232986,
+      "loss": 0.5632,
+      "step": 11645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016659390731445874,
+      "loss": 0.8126,
+      "step": 11646
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016657504202086115,
+      "loss": 0.0251,
+      "step": 11647
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016655617646183917,
+      "loss": 0.4918,
+      "step": 11648
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016653731063769484,
+      "loss": 0.3505,
+      "step": 11649
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016651844454873029,
+      "loss": 0.4836,
+      "step": 11650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016649957819524767,
+      "loss": 0.3176,
+      "step": 11651
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016648071157754905,
+      "loss": 0.5492,
+      "step": 11652
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001664618446959366,
+      "loss": 0.7828,
+      "step": 11653
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016644297755071232,
+      "loss": 0.5658,
+      "step": 11654
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001664241101421785,
+      "loss": 0.379,
+      "step": 11655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001664052424706372,
+      "loss": 0.3006,
+      "step": 11656
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016638637453639057,
+      "loss": 0.4209,
+      "step": 11657
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016636750633974073,
+      "loss": 0.6482,
+      "step": 11658
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016634863788098986,
+      "loss": 0.5408,
+      "step": 11659
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001663297691604401,
+      "loss": 0.001,
+      "step": 11660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016631090017839352,
+      "loss": 0.0343,
+      "step": 11661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001662920309351524,
+      "loss": 0.6495,
+      "step": 11662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016627316143101887,
+      "loss": 0.0114,
+      "step": 11663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016625429166629508,
+      "loss": 0.7151,
+      "step": 11664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016623542164128318,
+      "loss": 0.0214,
+      "step": 11665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001662165513562854,
+      "loss": 0.5094,
+      "step": 11666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016619768081160385,
+      "loss": 0.0028,
+      "step": 11667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016617881000754078,
+      "loss": 0.4114,
+      "step": 11668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016615993894439834,
+      "loss": 0.011,
+      "step": 11669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001661410676224787,
+      "loss": 0.2859,
+      "step": 11670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016612219604208414,
+      "loss": 1.1286,
+      "step": 11671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016610332420351675,
+      "loss": 0.4656,
+      "step": 11672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016608445210707884,
+      "loss": 0.6844,
+      "step": 11673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016606557975307255,
+      "loss": 2.8946,
+      "step": 11674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016604670714180013,
+      "loss": 0.4517,
+      "step": 11675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016602783427356377,
+      "loss": 0.3531,
+      "step": 11676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016600896114866566,
+      "loss": 2.4551,
+      "step": 11677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001659900877674081,
+      "loss": 0.4767,
+      "step": 11678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001659712141300933,
+      "loss": 0.4138,
+      "step": 11679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016595234023702342,
+      "loss": 0.0032,
+      "step": 11680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016593346608850077,
+      "loss": 1.3577,
+      "step": 11681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001659145916848276,
+      "loss": 0.5815,
+      "step": 11682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016589571702630614,
+      "loss": 0.5397,
+      "step": 11683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001658768421132386,
+      "loss": 0.2452,
+      "step": 11684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016585796694592727,
+      "loss": 1.0204,
+      "step": 11685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016583909152467438,
+      "loss": 0.0015,
+      "step": 11686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016582021584978223,
+      "loss": 0.2382,
+      "step": 11687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016580133992155304,
+      "loss": 0.1446,
+      "step": 11688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016578246374028912,
+      "loss": 0.8097,
+      "step": 11689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016576358730629274,
+      "loss": 1.3707,
+      "step": 11690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016574471061986615,
+      "loss": 0.0215,
+      "step": 11691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016572583368131168,
+      "loss": 0.3585,
+      "step": 11692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016570695649093155,
+      "loss": 0.0701,
+      "step": 11693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016568807904902814,
+      "loss": 0.2624,
+      "step": 11694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001656692013559036,
+      "loss": 0.4105,
+      "step": 11695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016565032341186035,
+      "loss": 0.0999,
+      "step": 11696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001656314452172007,
+      "loss": 0.4446,
+      "step": 11697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001656125667722269,
+      "loss": 0.3417,
+      "step": 11698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016559368807724125,
+      "loss": 0.0868,
+      "step": 11699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001655748091325461,
+      "loss": 0.0832,
+      "step": 11700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016555592993844375,
+      "loss": 0.1333,
+      "step": 11701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016553705049523654,
+      "loss": 0.752,
+      "step": 11702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016551817080322678,
+      "loss": 1.7192,
+      "step": 11703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016549929086271677,
+      "loss": 0.6162,
+      "step": 11704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016548041067400895,
+      "loss": 0.7031,
+      "step": 11705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016546153023740554,
+      "loss": 0.0022,
+      "step": 11706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016544264955320893,
+      "loss": 0.0014,
+      "step": 11707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016542376862172144,
+      "loss": 0.3928,
+      "step": 11708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001654048874432455,
+      "loss": 0.6544,
+      "step": 11709
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016538600601808336,
+      "loss": 0.7333,
+      "step": 11710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016536712434653747,
+      "loss": 5.2402,
+      "step": 11711
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016534824242891012,
+      "loss": 0.5168,
+      "step": 11712
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016532936026550368,
+      "loss": 3.5358,
+      "step": 11713
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016531047785662057,
+      "loss": 0.8879,
+      "step": 11714
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016529159520256315,
+      "loss": 0.8249,
+      "step": 11715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016527271230363372,
+      "loss": 0.1124,
+      "step": 11716
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016525382916013478,
+      "loss": 0.0024,
+      "step": 11717
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016523494577236862,
+      "loss": 0.3687,
+      "step": 11718
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016521606214063773,
+      "loss": 2.0686,
+      "step": 11719
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001651971782652444,
+      "loss": 0.5135,
+      "step": 11720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016517829414649103,
+      "loss": 0.0253,
+      "step": 11721
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001651594097846801,
+      "loss": 0.3029,
+      "step": 11722
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000165140525180114,
+      "loss": 0.2893,
+      "step": 11723
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016512164033309507,
+      "loss": 3.3125,
+      "step": 11724
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016510275524392575,
+      "loss": 0.4862,
+      "step": 11725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001650838699129085,
+      "loss": 1.7607,
+      "step": 11726
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016506498434034576,
+      "loss": 0.3458,
+      "step": 11727
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016504609852653984,
+      "loss": 0.0499,
+      "step": 11728
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016502721247179324,
+      "loss": 0.9508,
+      "step": 11729
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001650083261764084,
+      "loss": 0.2598,
+      "step": 11730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016498943964068772,
+      "loss": 0.1855,
+      "step": 11731
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001649705528649337,
+      "loss": 0.0923,
+      "step": 11732
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001649516658494487,
+      "loss": 0.396,
+      "step": 11733
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016493277859453518,
+      "loss": 0.454,
+      "step": 11734
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016491389110049573,
+      "loss": 1.0137,
+      "step": 11735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016489500336763262,
+      "loss": 0.607,
+      "step": 11736
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001648761153962484,
+      "loss": 0.4405,
+      "step": 11737
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016485722718664552,
+      "loss": 0.2618,
+      "step": 11738
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016483833873912645,
+      "loss": 0.3748,
+      "step": 11739
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016481945005399366,
+      "loss": 0.7552,
+      "step": 11740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016480056113154959,
+      "loss": 0.3668,
+      "step": 11741
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016478167197209677,
+      "loss": 0.3076,
+      "step": 11742
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016476278257593767,
+      "loss": 0.0087,
+      "step": 11743
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001647438929433747,
+      "loss": 0.7638,
+      "step": 11744
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016472500307471044,
+      "loss": 0.1329,
+      "step": 11745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016470611297024742,
+      "loss": 0.4102,
+      "step": 11746
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016468722263028802,
+      "loss": 0.0098,
+      "step": 11747
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016466833205513475,
+      "loss": 0.0331,
+      "step": 11748
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016464944124509022,
+      "loss": 0.5919,
+      "step": 11749
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016463055020045686,
+      "loss": 0.2733,
+      "step": 11750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016461165892153718,
+      "loss": 0.8916,
+      "step": 11751
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016459276740863374,
+      "loss": 0.3721,
+      "step": 11752
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000164573875662049,
+      "loss": 0.5417,
+      "step": 11753
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016455498368208552,
+      "loss": 0.8086,
+      "step": 11754
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016453609146904584,
+      "loss": 0.2775,
+      "step": 11755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016451719902323242,
+      "loss": 0.3818,
+      "step": 11756
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016449830634494787,
+      "loss": 0.1205,
+      "step": 11757
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016447941343449476,
+      "loss": 1.2826,
+      "step": 11758
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001644605202921755,
+      "loss": 1.4683,
+      "step": 11759
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016444162691829276,
+      "loss": 0.0252,
+      "step": 11760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000164422733313149,
+      "loss": 0.8931,
+      "step": 11761
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016440383947704683,
+      "loss": 0.4776,
+      "step": 11762
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016438494541028877,
+      "loss": 0.003,
+      "step": 11763
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016436605111317743,
+      "loss": 0.2732,
+      "step": 11764
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016434715658601533,
+      "loss": 0.326,
+      "step": 11765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016432826182910504,
+      "loss": 0.6211,
+      "step": 11766
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016430936684274915,
+      "loss": 0.4641,
+      "step": 11767
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001642904716272502,
+      "loss": 0.0411,
+      "step": 11768
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016427157618291086,
+      "loss": 0.0024,
+      "step": 11769
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016425268051003356,
+      "loss": 0.0027,
+      "step": 11770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000164233784608921,
+      "loss": 0.5297,
+      "step": 11771
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016421488847987578,
+      "loss": 0.871,
+      "step": 11772
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016419599212320044,
+      "loss": 2.9122,
+      "step": 11773
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016417709553919756,
+      "loss": 0.8549,
+      "step": 11774
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001641581987281698,
+      "loss": 0.0048,
+      "step": 11775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016413930169041974,
+      "loss": 0.939,
+      "step": 11776
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016412040442624994,
+      "loss": 0.2356,
+      "step": 11777
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016410150693596312,
+      "loss": 0.0079,
+      "step": 11778
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016408260921986178,
+      "loss": 0.3763,
+      "step": 11779
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016406371127824858,
+      "loss": 0.6222,
+      "step": 11780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016404481311142622,
+      "loss": 0.2423,
+      "step": 11781
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001640259147196972,
+      "loss": 1.558,
+      "step": 11782
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001640070161033642,
+      "loss": 0.2564,
+      "step": 11783
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016398811726272988,
+      "loss": 0.9545,
+      "step": 11784
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016396921819809685,
+      "loss": 0.2549,
+      "step": 11785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016395031890976776,
+      "loss": 0.394,
+      "step": 11786
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016393141939804524,
+      "loss": 0.4043,
+      "step": 11787
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000163912519663232,
+      "loss": 0.4513,
+      "step": 11788
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016389361970563056,
+      "loss": 0.0524,
+      "step": 11789
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016387471952554366,
+      "loss": 0.8354,
+      "step": 11790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016385581912327398,
+      "loss": 0.1623,
+      "step": 11791
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016383691849912414,
+      "loss": 0.2411,
+      "step": 11792
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016381801765339682,
+      "loss": 0.0088,
+      "step": 11793
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016379911658639465,
+      "loss": 0.3124,
+      "step": 11794
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016378021529842043,
+      "loss": 0.6375,
+      "step": 11795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016376131378977667,
+      "loss": 0.3285,
+      "step": 11796
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016374241206076613,
+      "loss": 2.3477,
+      "step": 11797
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016372351011169153,
+      "loss": 1.0136,
+      "step": 11798
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016370460794285546,
+      "loss": 0.9223,
+      "step": 11799
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016368570555456066,
+      "loss": 0.6815,
+      "step": 11800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016366680294710984,
+      "loss": 0.1269,
+      "step": 11801
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001636479001208057,
+      "loss": 0.197,
+      "step": 11802
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016362899707595088,
+      "loss": 0.6939,
+      "step": 11803
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016361009381284813,
+      "loss": 0.4523,
+      "step": 11804
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001635911903318002,
+      "loss": 0.1384,
+      "step": 11805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016357228663310973,
+      "loss": 0.122,
+      "step": 11806
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016355338271707947,
+      "loss": 1.8025,
+      "step": 11807
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016353447858401208,
+      "loss": 0.7042,
+      "step": 11808
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016351557423421038,
+      "loss": 0.373,
+      "step": 11809
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016349666966797703,
+      "loss": 0.7326,
+      "step": 11810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016347776488561476,
+      "loss": 0.2656,
+      "step": 11811
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016345885988742632,
+      "loss": 1.2681,
+      "step": 11812
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016343995467371444,
+      "loss": 0.0481,
+      "step": 11813
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016342104924478183,
+      "loss": 0.4578,
+      "step": 11814
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001634021436009313,
+      "loss": 0.5992,
+      "step": 11815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016338323774246554,
+      "loss": 0.0699,
+      "step": 11816
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016336433166968728,
+      "loss": 0.2847,
+      "step": 11817
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016334542538289936,
+      "loss": 0.3224,
+      "step": 11818
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016332651888240442,
+      "loss": 0.6437,
+      "step": 11819
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016330761216850533,
+      "loss": 0.373,
+      "step": 11820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001632887052415048,
+      "loss": 0.8208,
+      "step": 11821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001632697981017056,
+      "loss": 1.7829,
+      "step": 11822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016325089074941047,
+      "loss": 0.7597,
+      "step": 11823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016323198318492224,
+      "loss": 0.2256,
+      "step": 11824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016321307540854364,
+      "loss": 0.0667,
+      "step": 11825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016319416742057745,
+      "loss": 0.2636,
+      "step": 11826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001631752592213265,
+      "loss": 0.754,
+      "step": 11827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016315635081109355,
+      "loss": 0.2358,
+      "step": 11828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016313744219018136,
+      "loss": 0.3704,
+      "step": 11829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016311853335889282,
+      "loss": 0.6703,
+      "step": 11830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016309962431753055,
+      "loss": 0.6712,
+      "step": 11831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630807150663975,
+      "loss": 0.4585,
+      "step": 11832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630618056057965,
+      "loss": 0.379,
+      "step": 11833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630428959360302,
+      "loss": 1.2048,
+      "step": 11834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001630239860574015,
+      "loss": 0.0023,
+      "step": 11835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016300507597021325,
+      "loss": 0.002,
+      "step": 11836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001629861656747682,
+      "loss": 0.3972,
+      "step": 11837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001629672551713692,
+      "loss": 0.7285,
+      "step": 11838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00016294834446031907,
+      "loss": 0.8782,
+      "step": 11839
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016292943354192065,
+      "loss": 0.0439,
+      "step": 11840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016291052241647675,
+      "loss": 1.7705,
+      "step": 11841
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016289161108429018,
+      "loss": 0.3706,
+      "step": 11842
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016287269954566385,
+      "loss": 0.108,
+      "step": 11843
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001628537878009005,
+      "loss": 0.5016,
+      "step": 11844
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001628348758503031,
+      "loss": 0.4287,
+      "step": 11845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016281596369417437,
+      "loss": 0.207,
+      "step": 11846
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016279705133281726,
+      "loss": 0.9302,
+      "step": 11847
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016277813876653456,
+      "loss": 0.4989,
+      "step": 11848
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016275922599562912,
+      "loss": 2.2168,
+      "step": 11849
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016274031302040384,
+      "loss": 0.4281,
+      "step": 11850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016272139984116157,
+      "loss": 0.1402,
+      "step": 11851
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016270248645820518,
+      "loss": 0.6196,
+      "step": 11852
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016268357287183753,
+      "loss": 0.0311,
+      "step": 11853
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001626646590823615,
+      "loss": 0.3473,
+      "step": 11854
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016264574509007997,
+      "loss": 1.4074,
+      "step": 11855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001626268308952958,
+      "loss": 0.0039,
+      "step": 11856
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016260791649831186,
+      "loss": 0.3397,
+      "step": 11857
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016258900189943106,
+      "loss": 0.0086,
+      "step": 11858
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016257008709895634,
+      "loss": 0.4941,
+      "step": 11859
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001625511720971905,
+      "loss": 0.2491,
+      "step": 11860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016253225689443648,
+      "loss": 0.3478,
+      "step": 11861
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016251334149099722,
+      "loss": 0.4635,
+      "step": 11862
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001624944258871755,
+      "loss": 0.3862,
+      "step": 11863
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016247551008327432,
+      "loss": 3.4719,
+      "step": 11864
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001624565940795966,
+      "loss": 0.4238,
+      "step": 11865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001624376778764452,
+      "loss": 0.5661,
+      "step": 11866
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016241876147412307,
+      "loss": 0.5565,
+      "step": 11867
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623998448729331,
+      "loss": 0.009,
+      "step": 11868
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623809280731782,
+      "loss": 0.343,
+      "step": 11869
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623620110751614,
+      "loss": 1.5506,
+      "step": 11870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016234309387918547,
+      "loss": 0.211,
+      "step": 11871
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623241764855534,
+      "loss": 5.0328,
+      "step": 11872
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001623052588945682,
+      "loss": 0.1975,
+      "step": 11873
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016228634110653273,
+      "loss": 0.3088,
+      "step": 11874
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016226742312174994,
+      "loss": 0.3409,
+      "step": 11875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016224850494052278,
+      "loss": 0.4818,
+      "step": 11876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016222958656315418,
+      "loss": 0.389,
+      "step": 11877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016221066798994714,
+      "loss": 0.316,
+      "step": 11878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016219174922120456,
+      "loss": 0.8027,
+      "step": 11879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001621728302572294,
+      "loss": 0.5238,
+      "step": 11880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001621539110983246,
+      "loss": 0.1945,
+      "step": 11881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016213499174479325,
+      "loss": 0.2094,
+      "step": 11882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016211607219693814,
+      "loss": 0.0133,
+      "step": 11883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016209715245506234,
+      "loss": 0.7997,
+      "step": 11884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001620782325194688,
+      "loss": 0.1731,
+      "step": 11885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016205931239046052,
+      "loss": 1.1439,
+      "step": 11886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016204039206834037,
+      "loss": 0.57,
+      "step": 11887
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001620214715534115,
+      "loss": 0.0256,
+      "step": 11888
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016200255084597678,
+      "loss": 0.5726,
+      "step": 11889
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016198362994633922,
+      "loss": 0.0042,
+      "step": 11890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016196470885480177,
+      "loss": 0.4635,
+      "step": 11891
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001619457875716675,
+      "loss": 0.2988,
+      "step": 11892
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016192686609723937,
+      "loss": 0.6422,
+      "step": 11893
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016190794443182043,
+      "loss": 2.5918,
+      "step": 11894
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016188902257571354,
+      "loss": 0.5158,
+      "step": 11895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016187010052922185,
+      "loss": 0.3944,
+      "step": 11896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016185117829264832,
+      "loss": 0.6811,
+      "step": 11897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016183225586629592,
+      "loss": 0.3025,
+      "step": 11898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001618133332504677,
+      "loss": 0.3594,
+      "step": 11899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016179441044546672,
+      "loss": 0.5061,
+      "step": 11900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016177548745159592,
+      "loss": 0.2574,
+      "step": 11901
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016175656426915835,
+      "loss": 0.5715,
+      "step": 11902
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016173764089845707,
+      "loss": 0.1851,
+      "step": 11903
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001617187173397951,
+      "loss": 0.552,
+      "step": 11904
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016169979359347548,
+      "loss": 0.4302,
+      "step": 11905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016168086965980113,
+      "loss": 1.0799,
+      "step": 11906
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016166194553907527,
+      "loss": 0.344,
+      "step": 11907
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016164302123160087,
+      "loss": 0.4959,
+      "step": 11908
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001616240967376809,
+      "loss": 0.6692,
+      "step": 11909
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016160517205761842,
+      "loss": 0.4256,
+      "step": 11910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016158624719171663,
+      "loss": 0.7689,
+      "step": 11911
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016156732214027845,
+      "loss": 0.3891,
+      "step": 11912
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016154839690360695,
+      "loss": 0.113,
+      "step": 11913
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001615294714820052,
+      "loss": 0.2029,
+      "step": 11914
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016151054587577627,
+      "loss": 0.2482,
+      "step": 11915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016149162008522323,
+      "loss": 0.6882,
+      "step": 11916
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016147269411064913,
+      "loss": 0.3126,
+      "step": 11917
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016145376795235707,
+      "loss": 1.0376,
+      "step": 11918
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016143484161065008,
+      "loss": 0.007,
+      "step": 11919
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016141591508583134,
+      "loss": 0.7056,
+      "step": 11920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016139698837820376,
+      "loss": 0.0036,
+      "step": 11921
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016137806148807055,
+      "loss": 0.2608,
+      "step": 11922
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016135913441573478,
+      "loss": 0.5656,
+      "step": 11923
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016134020716149949,
+      "loss": 2.548,
+      "step": 11924
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001613212797256678,
+      "loss": 0.4015,
+      "step": 11925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016130235210854284,
+      "loss": 0.2879,
+      "step": 11926
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016128342431042763,
+      "loss": 0.2723,
+      "step": 11927
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016126449633162534,
+      "loss": 0.6015,
+      "step": 11928
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016124556817243905,
+      "loss": 0.7677,
+      "step": 11929
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016122663983317184,
+      "loss": 0.5602,
+      "step": 11930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001612077113141269,
+      "loss": 0.2891,
+      "step": 11931
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016118878261560724,
+      "loss": 0.3893,
+      "step": 11932
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000161169853737916,
+      "loss": 5.9749,
+      "step": 11933
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016115092468135635,
+      "loss": 0.0183,
+      "step": 11934
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016113199544623134,
+      "loss": 0.9021,
+      "step": 11935
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016111306603284415,
+      "loss": 0.2781,
+      "step": 11936
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016109413644149788,
+      "loss": 0.0017,
+      "step": 11937
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016107520667249568,
+      "loss": 0.742,
+      "step": 11938
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016105627672614067,
+      "loss": 0.1052,
+      "step": 11939
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016103734660273596,
+      "loss": 0.0394,
+      "step": 11940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016101841630258474,
+      "loss": 0.0029,
+      "step": 11941
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016099948582599008,
+      "loss": 0.5033,
+      "step": 11942
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001609805551732552,
+      "loss": 0.0034,
+      "step": 11943
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016096162434468318,
+      "loss": 0.3046,
+      "step": 11944
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001609426933405772,
+      "loss": 0.3956,
+      "step": 11945
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016092376216124041,
+      "loss": 0.3277,
+      "step": 11946
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000160904830806976,
+      "loss": 0.3938,
+      "step": 11947
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016088589927808704,
+      "loss": 1.1056,
+      "step": 11948
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016086696757487676,
+      "loss": 1.3504,
+      "step": 11949
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016084803569764832,
+      "loss": 0.0156,
+      "step": 11950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016082910364670483,
+      "loss": 0.4044,
+      "step": 11951
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016081017142234951,
+      "loss": 0.3253,
+      "step": 11952
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016079123902488552,
+      "loss": 0.5141,
+      "step": 11953
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016077230645461605,
+      "loss": 0.3014,
+      "step": 11954
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016075337371184426,
+      "loss": 0.0018,
+      "step": 11955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001607344407968733,
+      "loss": 0.3317,
+      "step": 11956
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016071550771000644,
+      "loss": 0.3519,
+      "step": 11957
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016069657445154674,
+      "loss": 1.4226,
+      "step": 11958
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016067764102179746,
+      "loss": 1.9639,
+      "step": 11959
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016065870742106183,
+      "loss": 0.7632,
+      "step": 11960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016063977364964296,
+      "loss": 0.1967,
+      "step": 11961
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001606208397078441,
+      "loss": 0.204,
+      "step": 11962
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001606019055959684,
+      "loss": 0.2393,
+      "step": 11963
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016058297131431914,
+      "loss": 0.8762,
+      "step": 11964
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016056403686319945,
+      "loss": 0.4484,
+      "step": 11965
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016054510224291257,
+      "loss": 0.7913,
+      "step": 11966
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001605261674537617,
+      "loss": 0.2756,
+      "step": 11967
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016050723249605005,
+      "loss": 0.212,
+      "step": 11968
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016048829737008087,
+      "loss": 1.0749,
+      "step": 11969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016046936207615728,
+      "loss": 0.1141,
+      "step": 11970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016045042661458263,
+      "loss": 1.6873,
+      "step": 11971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001604314909856601,
+      "loss": 0.0027,
+      "step": 11972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016041255518969288,
+      "loss": 0.3684,
+      "step": 11973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001603936192269842,
+      "loss": 0.1593,
+      "step": 11974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016037468309783727,
+      "loss": 0.0135,
+      "step": 11975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016035574680255542,
+      "loss": 0.3776,
+      "step": 11976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016033681034144178,
+      "loss": 0.2172,
+      "step": 11977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016031787371479966,
+      "loss": 0.8456,
+      "step": 11978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016029893692293227,
+      "loss": 0.3996,
+      "step": 11979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001602799999661429,
+      "loss": 0.1315,
+      "step": 11980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001602610628447347,
+      "loss": 1.9333,
+      "step": 11981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000160242125559011,
+      "loss": 0.0387,
+      "step": 11982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016022318810927506,
+      "loss": 0.5164,
+      "step": 11983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016020425049583008,
+      "loss": 2.6294,
+      "step": 11984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016018531271897932,
+      "loss": 0.3101,
+      "step": 11985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016016637477902612,
+      "loss": 0.636,
+      "step": 11986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016014743667627364,
+      "loss": 0.9038,
+      "step": 11987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001601284984110252,
+      "loss": 0.6706,
+      "step": 11988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016010955998358407,
+      "loss": 0.0039,
+      "step": 11989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016009062139425348,
+      "loss": 0.379,
+      "step": 11990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016007168264333673,
+      "loss": 0.5886,
+      "step": 11991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016005274373113716,
+      "loss": 0.3648,
+      "step": 11992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016003380465795793,
+      "loss": 0.1097,
+      "step": 11993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00016001486542410238,
+      "loss": 0.0037,
+      "step": 11994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015999592602987384,
+      "loss": 1.005,
+      "step": 11995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015997698647557553,
+      "loss": 0.3594,
+      "step": 11996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015995804676151068,
+      "loss": 0.5651,
+      "step": 11997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015993910688798276,
+      "loss": 1.877,
+      "step": 11998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001599201668552949,
+      "loss": 0.0054,
+      "step": 11999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015990122666375046,
+      "loss": 0.4084,
+      "step": 12000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015988228631365277,
+      "loss": 0.9974,
+      "step": 12001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015986334580530504,
+      "loss": 0.3915,
+      "step": 12002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015984440513901066,
+      "loss": 0.6688,
+      "step": 12003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001598254643150729,
+      "loss": 0.3752,
+      "step": 12004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015980652333379508,
+      "loss": 0.5279,
+      "step": 12005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015978758219548054,
+      "loss": 0.544,
+      "step": 12006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015976864090043252,
+      "loss": 0.3465,
+      "step": 12007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015974969944895434,
+      "loss": 0.3585,
+      "step": 12008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015973075784134943,
+      "loss": 0.815,
+      "step": 12009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159711816077921,
+      "loss": 0.7576,
+      "step": 12010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001596928741589724,
+      "loss": 0.2045,
+      "step": 12011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159673932084807,
+      "loss": 0.1649,
+      "step": 12012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015965498985572808,
+      "loss": 0.0089,
+      "step": 12013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159636047472039,
+      "loss": 0.6258,
+      "step": 12014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001596171049340431,
+      "loss": 0.494,
+      "step": 12015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001595981622420436,
+      "loss": 0.0419,
+      "step": 12016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015957921939634403,
+      "loss": 0.1344,
+      "step": 12017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015956027639724762,
+      "loss": 3.0753,
+      "step": 12018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015954133324505768,
+      "loss": 0.0966,
+      "step": 12019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015952238994007765,
+      "loss": 0.3902,
+      "step": 12020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001595034464826109,
+      "loss": 0.4913,
+      "step": 12021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001594845028729606,
+      "loss": 0.6472,
+      "step": 12022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001594655591114303,
+      "loss": 0.4168,
+      "step": 12023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015944661519832324,
+      "loss": 0.3021,
+      "step": 12024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015942767113394282,
+      "loss": 0.4781,
+      "step": 12025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001594087269185924,
+      "loss": 0.1161,
+      "step": 12026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015938978255257535,
+      "loss": 0.009,
+      "step": 12027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159370838036195,
+      "loss": 1.6083,
+      "step": 12028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015935189336975476,
+      "loss": 0.0586,
+      "step": 12029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000159332948553558,
+      "loss": 0.7136,
+      "step": 12030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015931400358790802,
+      "loss": 0.0016,
+      "step": 12031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015929505847310833,
+      "loss": 4.686,
+      "step": 12032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015927611320946217,
+      "loss": 0.9536,
+      "step": 12033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015925716779727295,
+      "loss": 0.3403,
+      "step": 12034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015923822223684418,
+      "loss": 1.2447,
+      "step": 12035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015921927652847906,
+      "loss": 0.2926,
+      "step": 12036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001592003306724811,
+      "loss": 0.0019,
+      "step": 12037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015918138466915365,
+      "loss": 0.3406,
+      "step": 12038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001591624385188001,
+      "loss": 0.1532,
+      "step": 12039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015914349222172385,
+      "loss": 1.1974,
+      "step": 12040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015912454577822832,
+      "loss": 0.9887,
+      "step": 12041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015910559918861682,
+      "loss": 0.0061,
+      "step": 12042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015908665245319283,
+      "loss": 0.2414,
+      "step": 12043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015906770557225978,
+      "loss": 0.5789,
+      "step": 12044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015904875854612103,
+      "loss": 0.1894,
+      "step": 12045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015902981137507993,
+      "loss": 1.3916,
+      "step": 12046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015901086405944003,
+      "loss": 0.7283,
+      "step": 12047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015899191659950464,
+      "loss": 0.0025,
+      "step": 12048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001589729689955772,
+      "loss": 1.0291,
+      "step": 12049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001589540212479611,
+      "loss": 1.5586,
+      "step": 12050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001589350733569598,
+      "loss": 0.2298,
+      "step": 12051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015891612532287672,
+      "loss": 0.4011,
+      "step": 12052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015889717714601527,
+      "loss": 0.6721,
+      "step": 12053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015887822882667887,
+      "loss": 0.0302,
+      "step": 12054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015885928036517098,
+      "loss": 0.5612,
+      "step": 12055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000158840331761795,
+      "loss": 0.0017,
+      "step": 12056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015882138301685436,
+      "loss": 0.498,
+      "step": 12057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015880243413065256,
+      "loss": 0.3317,
+      "step": 12058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015878348510349295,
+      "loss": 0.3942,
+      "step": 12059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015876453593567901,
+      "loss": 0.6164,
+      "step": 12060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015874558662751418,
+      "loss": 0.4644,
+      "step": 12061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015872663717930193,
+      "loss": 0.3039,
+      "step": 12062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015870768759134566,
+      "loss": 0.0044,
+      "step": 12063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015868873786394882,
+      "loss": 0.6052,
+      "step": 12064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001586697879974149,
+      "loss": 0.2864,
+      "step": 12065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015865083799204734,
+      "loss": 0.4052,
+      "step": 12066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015863188784814963,
+      "loss": 0.1069,
+      "step": 12067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015861293756602512,
+      "loss": 0.4886,
+      "step": 12068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015859398714597738,
+      "loss": 0.3292,
+      "step": 12069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001585750365883099,
+      "loss": 0.7439,
+      "step": 12070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015855608589332597,
+      "loss": 0.0017,
+      "step": 12071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001585371350613292,
+      "loss": 0.0469,
+      "step": 12072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015851818409262305,
+      "loss": 0.4383,
+      "step": 12073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015849923298751092,
+      "loss": 2.4257,
+      "step": 12074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015848028174629636,
+      "loss": 0.0046,
+      "step": 12075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015846133036928283,
+      "loss": 0.1661,
+      "step": 12076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015844237885677376,
+      "loss": 0.7,
+      "step": 12077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015842342720907265,
+      "loss": 0.0013,
+      "step": 12078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015840447542648308,
+      "loss": 0.001,
+      "step": 12079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015838552350930834,
+      "loss": 0.0016,
+      "step": 12080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015836657145785208,
+      "loss": 0.2127,
+      "step": 12081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015834761927241775,
+      "loss": 0.0576,
+      "step": 12082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015832866695330878,
+      "loss": 0.2256,
+      "step": 12083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015830971450082876,
+      "loss": 0.0383,
+      "step": 12084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015829076191528111,
+      "loss": 0.5345,
+      "step": 12085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015827180919696934,
+      "loss": 0.0037,
+      "step": 12086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015825285634619697,
+      "loss": 0.0115,
+      "step": 12087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001582339033632675,
+      "loss": 0.2759,
+      "step": 12088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00015821495024848442,
+      "loss": 0.2553,
+      "step": 12089
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015819599700215125,
+      "loss": 0.5362,
+      "step": 12090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015817704362457144,
+      "loss": 0.376,
+      "step": 12091
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015815809011604856,
+      "loss": 0.1453,
+      "step": 12092
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015813913647688617,
+      "loss": 0.9176,
+      "step": 12093
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015812018270738772,
+      "loss": 0.4875,
+      "step": 12094
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015810122880785662,
+      "loss": 0.4925,
+      "step": 12095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001580822747785966,
+      "loss": 0.0012,
+      "step": 12096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015806332061991108,
+      "loss": 3.1709,
+      "step": 12097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015804436633210352,
+      "loss": 1.0824,
+      "step": 12098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015802541191547756,
+      "loss": 0.3773,
+      "step": 12099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015800645737033668,
+      "loss": 0.4712,
+      "step": 12100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015798750269698436,
+      "loss": 0.2925,
+      "step": 12101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015796854789572418,
+      "loss": 1.2088,
+      "step": 12102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015794959296685967,
+      "loss": 4.0822,
+      "step": 12103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015793063791069437,
+      "loss": 0.5371,
+      "step": 12104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015791168272753182,
+      "loss": 0.0078,
+      "step": 12105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001578927274176755,
+      "loss": 0.419,
+      "step": 12106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015787377198142903,
+      "loss": 0.2713,
+      "step": 12107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015785481641909595,
+      "loss": 0.2116,
+      "step": 12108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001578358607309797,
+      "loss": 0.9605,
+      "step": 12109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001578169049173839,
+      "loss": 0.3226,
+      "step": 12110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015779794897861216,
+      "loss": 0.1343,
+      "step": 12111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015777899291496796,
+      "loss": 0.3931,
+      "step": 12112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015776003672675484,
+      "loss": 0.1659,
+      "step": 12113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015774108041427637,
+      "loss": 0.5716,
+      "step": 12114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015772212397783611,
+      "loss": 3.0157,
+      "step": 12115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001577031674177377,
+      "loss": 0.0024,
+      "step": 12116
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015768421073428452,
+      "loss": 0.3566,
+      "step": 12117
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015766525392778028,
+      "loss": 0.3073,
+      "step": 12118
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015764629699852853,
+      "loss": 0.3366,
+      "step": 12119
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015762733994683278,
+      "loss": 0.3884,
+      "step": 12120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001576083827729966,
+      "loss": 0.0349,
+      "step": 12121
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015758942547732365,
+      "loss": 0.7955,
+      "step": 12122
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001575704680601174,
+      "loss": 0.135,
+      "step": 12123
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015755151052168148,
+      "loss": 0.1702,
+      "step": 12124
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015753255286231945,
+      "loss": 0.1942,
+      "step": 12125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015751359508233488,
+      "loss": 0.8864,
+      "step": 12126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001574946371820314,
+      "loss": 2.0548,
+      "step": 12127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015747567916171254,
+      "loss": 0.3936,
+      "step": 12128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015745672102168186,
+      "loss": 0.52,
+      "step": 12129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015743776276224298,
+      "loss": 0.3937,
+      "step": 12130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015741880438369956,
+      "loss": 0.2421,
+      "step": 12131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015739984588635506,
+      "loss": 0.2462,
+      "step": 12132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015738088727051318,
+      "loss": 0.0036,
+      "step": 12133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015736192853647743,
+      "loss": 0.0009,
+      "step": 12134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015734296968455144,
+      "loss": 0.7788,
+      "step": 12135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015732401071503883,
+      "loss": 0.2273,
+      "step": 12136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015730505162824318,
+      "loss": 1.0501,
+      "step": 12137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015728609242446811,
+      "loss": 0.0026,
+      "step": 12138
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015726713310401717,
+      "loss": 0.3063,
+      "step": 12139
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000157248173667194,
+      "loss": 0.4608,
+      "step": 12140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001572292141143022,
+      "loss": 0.5086,
+      "step": 12141
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015721025444564546,
+      "loss": 0.9906,
+      "step": 12142
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015719129466152724,
+      "loss": 0.003,
+      "step": 12143
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015717233476225122,
+      "loss": 0.8993,
+      "step": 12144
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001571533747481211,
+      "loss": 0.5545,
+      "step": 12145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015713441461944033,
+      "loss": 0.6797,
+      "step": 12146
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015711545437651264,
+      "loss": 0.2271,
+      "step": 12147
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015709649401964167,
+      "loss": 0.2901,
+      "step": 12148
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000157077533549131,
+      "loss": 0.002,
+      "step": 12149
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015705857296528423,
+      "loss": 0.1328,
+      "step": 12150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000157039612268405,
+      "loss": 0.4929,
+      "step": 12151
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001570206514587969,
+      "loss": 0.817,
+      "step": 12152
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015700169053676366,
+      "loss": 0.1211,
+      "step": 12153
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015698272950260888,
+      "loss": 1.9209,
+      "step": 12154
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015696376835663608,
+      "loss": 0.3086,
+      "step": 12155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015694480709914905,
+      "loss": 0.5654,
+      "step": 12156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015692584573045136,
+      "loss": 0.2815,
+      "step": 12157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015690688425084665,
+      "loss": 2.627,
+      "step": 12158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001568879226606385,
+      "loss": 1.5153,
+      "step": 12159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015686896096013065,
+      "loss": 0.3094,
+      "step": 12160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001568499991496267,
+      "loss": 0.9373,
+      "step": 12161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015683103722943025,
+      "loss": 0.0051,
+      "step": 12162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015681207519984504,
+      "loss": 0.9751,
+      "step": 12163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015679311306117461,
+      "loss": 0.0008,
+      "step": 12164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015677415081372272,
+      "loss": 0.287,
+      "step": 12165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015675518845779295,
+      "loss": 0.5968,
+      "step": 12166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015673622599368897,
+      "loss": 0.8253,
+      "step": 12167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001567172634217145,
+      "loss": 0.0008,
+      "step": 12168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566983007421731,
+      "loss": 0.2635,
+      "step": 12169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566793379553684,
+      "loss": 0.4626,
+      "step": 12170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015666037506160423,
+      "loss": 0.3111,
+      "step": 12171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566414120611841,
+      "loss": 0.2311,
+      "step": 12172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001566224489544117,
+      "loss": 0.9055,
+      "step": 12173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015660348574159075,
+      "loss": 0.033,
+      "step": 12174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001565845224230249,
+      "loss": 0.4951,
+      "step": 12175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001565655589990178,
+      "loss": 0.0039,
+      "step": 12176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015654659546987315,
+      "loss": 0.4781,
+      "step": 12177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001565276318358945,
+      "loss": 0.3697,
+      "step": 12178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001565086680973857,
+      "loss": 0.6775,
+      "step": 12179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015648970425465036,
+      "loss": 0.4834,
+      "step": 12180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015647074030799207,
+      "loss": 0.5004,
+      "step": 12181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015645177625771465,
+      "loss": 0.7319,
+      "step": 12182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015643281210412174,
+      "loss": 0.8148,
+      "step": 12183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015641384784751696,
+      "loss": 0.223,
+      "step": 12184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000156394883488204,
+      "loss": 0.1904,
+      "step": 12185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015637591902648663,
+      "loss": 0.0012,
+      "step": 12186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015635695446266845,
+      "loss": 0.0015,
+      "step": 12187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015633798979705323,
+      "loss": 0.7969,
+      "step": 12188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015631902502994457,
+      "loss": 0.7389,
+      "step": 12189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015630006016164623,
+      "loss": 0.8319,
+      "step": 12190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015628109519246185,
+      "loss": 0.043,
+      "step": 12191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015626213012269523,
+      "loss": 0.3203,
+      "step": 12192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001562431649526499,
+      "loss": 0.2972,
+      "step": 12193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015622419968262973,
+      "loss": 0.3703,
+      "step": 12194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015620523431293833,
+      "loss": 0.428,
+      "step": 12195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015618626884387934,
+      "loss": 0.0033,
+      "step": 12196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015616730327575664,
+      "loss": 0.0049,
+      "step": 12197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001561483376088738,
+      "loss": 0.7111,
+      "step": 12198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015612937184353453,
+      "loss": 1.3782,
+      "step": 12199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001561104059800426,
+      "loss": 0.322,
+      "step": 12200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015609144001870167,
+      "loss": 0.678,
+      "step": 12201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015607247395981545,
+      "loss": 0.7572,
+      "step": 12202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015605350780368774,
+      "loss": 0.5992,
+      "step": 12203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015603454155062213,
+      "loss": 0.2621,
+      "step": 12204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015601557520092242,
+      "loss": 0.0138,
+      "step": 12205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015599660875489232,
+      "loss": 0.0019,
+      "step": 12206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001559776422128355,
+      "loss": 0.3061,
+      "step": 12207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015595867557505568,
+      "loss": 0.0092,
+      "step": 12208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015593970884185667,
+      "loss": 0.0253,
+      "step": 12209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015592074201354212,
+      "loss": 0.7082,
+      "step": 12210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015590177509041577,
+      "loss": 0.9898,
+      "step": 12211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015588280807278133,
+      "loss": 0.2796,
+      "step": 12212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015586384096094257,
+      "loss": 0.1846,
+      "step": 12213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015584487375520318,
+      "loss": 0.0229,
+      "step": 12214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015582590645586694,
+      "loss": 0.0021,
+      "step": 12215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015580693906323752,
+      "loss": 0.6665,
+      "step": 12216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001557879715776187,
+      "loss": 0.2676,
+      "step": 12217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015576900399931425,
+      "loss": 0.5428,
+      "step": 12218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015575003632862776,
+      "loss": 0.5248,
+      "step": 12219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015573106856586316,
+      "loss": 0.5934,
+      "step": 12220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015571210071132406,
+      "loss": 0.4432,
+      "step": 12221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015569313276531422,
+      "loss": 1.336,
+      "step": 12222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015567416472813742,
+      "loss": 0.1635,
+      "step": 12223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001556551966000974,
+      "loss": 1.0847,
+      "step": 12224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015563622838149786,
+      "loss": 0.3172,
+      "step": 12225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015561726007264263,
+      "loss": 1.6112,
+      "step": 12226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015559829167383535,
+      "loss": 0.2655,
+      "step": 12227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015557932318537986,
+      "loss": 0.0685,
+      "step": 12228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001555603546075799,
+      "loss": 0.1217,
+      "step": 12229
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015554138594073917,
+      "loss": 0.6734,
+      "step": 12230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015552241718516148,
+      "loss": 1.1231,
+      "step": 12231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015550344834115061,
+      "loss": 0.3933,
+      "step": 12232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015548447940901024,
+      "loss": 0.3665,
+      "step": 12233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001554655103890441,
+      "loss": 0.0503,
+      "step": 12234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001554465412815561,
+      "loss": 0.339,
+      "step": 12235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001554275720868499,
+      "loss": 0.4949,
+      "step": 12236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015540860280522924,
+      "loss": 0.3738,
+      "step": 12237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015538963343699795,
+      "loss": 0.2311,
+      "step": 12238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015537066398245978,
+      "loss": 0.4369,
+      "step": 12239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015535169444191844,
+      "loss": 0.398,
+      "step": 12240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015533272481567782,
+      "loss": 0.4784,
+      "step": 12241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015531375510404155,
+      "loss": 0.0033,
+      "step": 12242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001552947853073135,
+      "loss": 0.0017,
+      "step": 12243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015527581542579743,
+      "loss": 0.8292,
+      "step": 12244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015525684545979703,
+      "loss": 0.2995,
+      "step": 12245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015523787540961618,
+      "loss": 0.5882,
+      "step": 12246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001552189052755586,
+      "loss": 0.4552,
+      "step": 12247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001551999350579281,
+      "loss": 0.2104,
+      "step": 12248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015518096475702846,
+      "loss": 0.2853,
+      "step": 12249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015516199437316341,
+      "loss": 0.386,
+      "step": 12250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015514302390663676,
+      "loss": 0.1059,
+      "step": 12251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015512405335775235,
+      "loss": 0.3286,
+      "step": 12252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015510508272681384,
+      "loss": 0.3404,
+      "step": 12253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015508611201412513,
+      "loss": 0.535,
+      "step": 12254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015506714121999005,
+      "loss": 0.0045,
+      "step": 12255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001550481703447122,
+      "loss": 0.3289,
+      "step": 12256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001550291993885955,
+      "loss": 0.675,
+      "step": 12257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015501022835194377,
+      "loss": 0.2305,
+      "step": 12258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001549912572350607,
+      "loss": 0.1727,
+      "step": 12259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015497228603825013,
+      "loss": 0.0022,
+      "step": 12260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015495331476181591,
+      "loss": 0.3638,
+      "step": 12261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015493434340606174,
+      "loss": 0.1727,
+      "step": 12262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001549153719712915,
+      "loss": 0.4232,
+      "step": 12263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015489640045780894,
+      "loss": 0.3483,
+      "step": 12264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015487742886591786,
+      "loss": 0.4222,
+      "step": 12265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015485845719592209,
+      "loss": 0.9918,
+      "step": 12266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015483948544812548,
+      "loss": 0.2495,
+      "step": 12267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015482051362283164,
+      "loss": 0.5123,
+      "step": 12268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015480154172034465,
+      "loss": 0.2754,
+      "step": 12269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015478256974096812,
+      "loss": 0.2297,
+      "step": 12270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001547635976850059,
+      "loss": 0.9973,
+      "step": 12271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001547446255527618,
+      "loss": 0.0012,
+      "step": 12272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015472565334453966,
+      "loss": 0.9432,
+      "step": 12273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015470668106064324,
+      "loss": 0.008,
+      "step": 12274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015468770870137642,
+      "loss": 0.4151,
+      "step": 12275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000154668736267043,
+      "loss": 0.2605,
+      "step": 12276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015464976375794676,
+      "loss": 0.7354,
+      "step": 12277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015463079117439155,
+      "loss": 0.3958,
+      "step": 12278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001546118185166811,
+      "loss": 0.3867,
+      "step": 12279
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015459284578511938,
+      "loss": 0.3538,
+      "step": 12280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015457387298001008,
+      "loss": 0.4886,
+      "step": 12281
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015455490010165708,
+      "loss": 0.6423,
+      "step": 12282
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015453592715036416,
+      "loss": 0.2941,
+      "step": 12283
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015451695412643524,
+      "loss": 0.6617,
+      "step": 12284
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015449798103017406,
+      "loss": 0.215,
+      "step": 12285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015447900786188444,
+      "loss": 0.1454,
+      "step": 12286
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015446003462187024,
+      "loss": 0.2995,
+      "step": 12287
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015444106131043528,
+      "loss": 0.5558,
+      "step": 12288
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015442208792788338,
+      "loss": 0.2671,
+      "step": 12289
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001544031144745184,
+      "loss": 0.3123,
+      "step": 12290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015438414095064412,
+      "loss": 0.3648,
+      "step": 12291
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001543651673565644,
+      "loss": 0.5017,
+      "step": 12292
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001543461936925831,
+      "loss": 0.7675,
+      "step": 12293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015432721995900403,
+      "loss": 0.8233,
+      "step": 12294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015430824615613104,
+      "loss": 0.3254,
+      "step": 12295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015428927228426794,
+      "loss": 0.0146,
+      "step": 12296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015427029834371857,
+      "loss": 0.5188,
+      "step": 12297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015425132433478678,
+      "loss": 0.0004,
+      "step": 12298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015423235025777642,
+      "loss": 2.5757,
+      "step": 12299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015421337611299133,
+      "loss": 0.0011,
+      "step": 12300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015419440190073532,
+      "loss": 0.0032,
+      "step": 12301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001541754276213123,
+      "loss": 0.3236,
+      "step": 12302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000154156453275026,
+      "loss": 2.752,
+      "step": 12303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015413747886218042,
+      "loss": 0.5041,
+      "step": 12304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001541185043830793,
+      "loss": 0.0089,
+      "step": 12305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015409952983802647,
+      "loss": 0.5349,
+      "step": 12306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015408055522732588,
+      "loss": 0.3959,
+      "step": 12307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001540615805512813,
+      "loss": 1.1056,
+      "step": 12308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001540426058101966,
+      "loss": 0.525,
+      "step": 12309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015402363100437564,
+      "loss": 0.8131,
+      "step": 12310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015400465613412228,
+      "loss": 0.0019,
+      "step": 12311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015398568119974032,
+      "loss": 0.845,
+      "step": 12312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001539667062015337,
+      "loss": 0.0012,
+      "step": 12313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001539477311398062,
+      "loss": 0.3489,
+      "step": 12314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015392875601486172,
+      "loss": 0.6915,
+      "step": 12315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015390978082700416,
+      "loss": 0.0005,
+      "step": 12316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015389080557653723,
+      "loss": 0.0052,
+      "step": 12317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015387183026376497,
+      "loss": 2.3496,
+      "step": 12318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015385285488899114,
+      "loss": 1.0522,
+      "step": 12319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015383387945251963,
+      "loss": 0.1859,
+      "step": 12320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015381490395465427,
+      "loss": 0.7335,
+      "step": 12321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015379592839569894,
+      "loss": 0.3014,
+      "step": 12322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015377695277595755,
+      "loss": 0.0015,
+      "step": 12323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001537579770957339,
+      "loss": 0.6035,
+      "step": 12324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015373900135533188,
+      "loss": 0.4619,
+      "step": 12325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015372002555505537,
+      "loss": 0.5757,
+      "step": 12326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001537010496952083,
+      "loss": 0.4484,
+      "step": 12327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001536820737760944,
+      "loss": 0.398,
+      "step": 12328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015366309779801763,
+      "loss": 1.4785,
+      "step": 12329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015364412176128187,
+      "loss": 0.6411,
+      "step": 12330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015362514566619092,
+      "loss": 0.4217,
+      "step": 12331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015360616951304873,
+      "loss": 0.4638,
+      "step": 12332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015358719330215914,
+      "loss": 0.0514,
+      "step": 12333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015356821703382606,
+      "loss": 0.472,
+      "step": 12334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015354924070835332,
+      "loss": 0.1194,
+      "step": 12335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001535302643260448,
+      "loss": 0.005,
+      "step": 12336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015351128788720442,
+      "loss": 1.3994,
+      "step": 12337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00015349231139213603,
+      "loss": 0.703,
+      "step": 12338
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015347333484114353,
+      "loss": 0.9245,
+      "step": 12339
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015345435823453074,
+      "loss": 0.4321,
+      "step": 12340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001534353815726016,
+      "loss": 0.3901,
+      "step": 12341
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015341640485566003,
+      "loss": 0.5277,
+      "step": 12342
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001533974280840098,
+      "loss": 0.5653,
+      "step": 12343
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015337845125795488,
+      "loss": 4.4279,
+      "step": 12344
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015335947437779916,
+      "loss": 1.2679,
+      "step": 12345
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015334049744384647,
+      "loss": 0.079,
+      "step": 12346
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015332152045640073,
+      "loss": 1.5586,
+      "step": 12347
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015330254341576585,
+      "loss": 0.8915,
+      "step": 12348
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015328356632224568,
+      "loss": 0.2206,
+      "step": 12349
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015326458917614409,
+      "loss": 0.5888,
+      "step": 12350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015324561197776503,
+      "loss": 1.0546,
+      "step": 12351
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015322663472741238,
+      "loss": 0.1328,
+      "step": 12352
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015320765742539002,
+      "loss": 0.7294,
+      "step": 12353
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015318868007200184,
+      "loss": 0.8796,
+      "step": 12354
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015316970266755168,
+      "loss": 0.0945,
+      "step": 12355
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015315072521234356,
+      "loss": 0.6777,
+      "step": 12356
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015313174770668125,
+      "loss": 0.0061,
+      "step": 12357
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001531127701508687,
+      "loss": 2.647,
+      "step": 12358
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015309379254520986,
+      "loss": 0.2698,
+      "step": 12359
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001530748148900085,
+      "loss": 0.2654,
+      "step": 12360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015305583718556865,
+      "loss": 0.3378,
+      "step": 12361
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015303685943219414,
+      "loss": 0.9066,
+      "step": 12362
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015301788163018889,
+      "loss": 0.4638,
+      "step": 12363
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001529989037798568,
+      "loss": 0.2249,
+      "step": 12364
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015297992588150178,
+      "loss": 1.367,
+      "step": 12365
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015296094793542763,
+      "loss": 0.0043,
+      "step": 12366
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015294196994193842,
+      "loss": 0.7261,
+      "step": 12367
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000152922991901338,
+      "loss": 0.1515,
+      "step": 12368
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015290401381393024,
+      "loss": 1.0129,
+      "step": 12369
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000152885035680019,
+      "loss": 0.8227,
+      "step": 12370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001528660574999083,
+      "loss": 0.4373,
+      "step": 12371
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015284707927390196,
+      "loss": 0.0328,
+      "step": 12372
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001528281010023039,
+      "loss": 1.0459,
+      "step": 12373
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001528091226854181,
+      "loss": 0.5416,
+      "step": 12374
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001527901443235484,
+      "loss": 0.553,
+      "step": 12375
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001527711659169987,
+      "loss": 0.0031,
+      "step": 12376
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015275218746607297,
+      "loss": 0.2472,
+      "step": 12377
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015273320897107504,
+      "loss": 0.4101,
+      "step": 12378
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015271423043230895,
+      "loss": 0.5003,
+      "step": 12379
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015269525185007847,
+      "loss": 1.1506,
+      "step": 12380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015267627322468758,
+      "loss": 0.0744,
+      "step": 12381
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015265729455644022,
+      "loss": 0.0104,
+      "step": 12382
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015263831584564025,
+      "loss": 0.2975,
+      "step": 12383
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001526193370925916,
+      "loss": 0.6177,
+      "step": 12384
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001526003582975982,
+      "loss": 0.6908,
+      "step": 12385
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000152581379460964,
+      "loss": 2.4468,
+      "step": 12386
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015256240058299285,
+      "loss": 0.049,
+      "step": 12387
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015254342166398874,
+      "loss": 0.3331,
+      "step": 12388
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001525244427042555,
+      "loss": 0.0199,
+      "step": 12389
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015250546370409714,
+      "loss": 0.5025,
+      "step": 12390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015248648466381753,
+      "loss": 0.0961,
+      "step": 12391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015246750558372055,
+      "loss": 0.3811,
+      "step": 12392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015244852646411022,
+      "loss": 0.5134,
+      "step": 12393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015242954730529043,
+      "loss": 0.0653,
+      "step": 12394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015241056810756505,
+      "loss": 0.7892,
+      "step": 12395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015239158887123808,
+      "loss": 0.3368,
+      "step": 12396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015237260959661335,
+      "loss": 0.6924,
+      "step": 12397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015235363028399485,
+      "loss": 0.1181,
+      "step": 12398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015233465093368652,
+      "loss": 0.5873,
+      "step": 12399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015231567154599224,
+      "loss": 0.4299,
+      "step": 12400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015229669212121595,
+      "loss": 0.5903,
+      "step": 12401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001522777126596616,
+      "loss": 0.3648,
+      "step": 12402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015225873316163313,
+      "loss": 0.1806,
+      "step": 12403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015223975362743436,
+      "loss": 0.0216,
+      "step": 12404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015222077405736934,
+      "loss": 0.435,
+      "step": 12405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015220179445174197,
+      "loss": 0.266,
+      "step": 12406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015218281481085616,
+      "loss": 0.0934,
+      "step": 12407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001521638351350158,
+      "loss": 0.9917,
+      "step": 12408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015214485542452491,
+      "loss": 0.6074,
+      "step": 12409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015212587567968737,
+      "loss": 0.1762,
+      "step": 12410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015210689590080712,
+      "loss": 0.4858,
+      "step": 12411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001520879160881881,
+      "loss": 0.35,
+      "step": 12412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015206893624213424,
+      "loss": 0.7838,
+      "step": 12413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015204995636294952,
+      "loss": 1.8852,
+      "step": 12414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015203097645093773,
+      "loss": 0.949,
+      "step": 12415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015201199650640294,
+      "loss": 0.718,
+      "step": 12416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015199301652964908,
+      "loss": 0.6904,
+      "step": 12417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015197403652098,
+      "loss": 0.2744,
+      "step": 12418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015195505648069968,
+      "loss": 0.0165,
+      "step": 12419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015193607640911211,
+      "loss": 0.2421,
+      "step": 12420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001519170963065212,
+      "loss": 0.8672,
+      "step": 12421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015189811617323084,
+      "loss": 0.5167,
+      "step": 12422
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015187913600954496,
+      "loss": 1.0246,
+      "step": 12423
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015186015581576757,
+      "loss": 1.0213,
+      "step": 12424
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015184117559220256,
+      "loss": 0.147,
+      "step": 12425
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015182219533915392,
+      "loss": 0.17,
+      "step": 12426
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015180321505692555,
+      "loss": 0.3711,
+      "step": 12427
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015178423474582136,
+      "loss": 0.7225,
+      "step": 12428
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001517652544061454,
+      "loss": 0.0033,
+      "step": 12429
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015174627403820147,
+      "loss": 0.9695,
+      "step": 12430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015172729364229363,
+      "loss": 0.3987,
+      "step": 12431
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015170831321872574,
+      "loss": 0.448,
+      "step": 12432
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015168933276780178,
+      "loss": 0.7744,
+      "step": 12433
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015167035228982572,
+      "loss": 0.4195,
+      "step": 12434
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015165137178510145,
+      "loss": 0.3727,
+      "step": 12435
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015163239125393294,
+      "loss": 0.6323,
+      "step": 12436
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015161341069662413,
+      "loss": 0.9478,
+      "step": 12437
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015159443011347897,
+      "loss": 0.1268,
+      "step": 12438
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001515754495048014,
+      "loss": 3.6514,
+      "step": 12439
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015155646887089543,
+      "loss": 0.3571,
+      "step": 12440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015153748821206483,
+      "loss": 0.0864,
+      "step": 12441
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015151850752861374,
+      "loss": 0.3715,
+      "step": 12442
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015149952682084607,
+      "loss": 0.3666,
+      "step": 12443
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015148054608906566,
+      "loss": 0.5057,
+      "step": 12444
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001514615653335765,
+      "loss": 0.2628,
+      "step": 12445
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015144258455468266,
+      "loss": 0.1964,
+      "step": 12446
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015142360375268796,
+      "loss": 0.4178,
+      "step": 12447
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015140462292789636,
+      "loss": 0.4344,
+      "step": 12448
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015138564208061185,
+      "loss": 0.5893,
+      "step": 12449
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015136666121113836,
+      "loss": 0.0078,
+      "step": 12450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015134768031977979,
+      "loss": 0.2803,
+      "step": 12451
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015132869940684025,
+      "loss": 0.3599,
+      "step": 12452
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015130971847262346,
+      "loss": 0.4222,
+      "step": 12453
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015129073751743357,
+      "loss": 0.6611,
+      "step": 12454
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015127175654157447,
+      "loss": 0.0024,
+      "step": 12455
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015125277554535008,
+      "loss": 0.9579,
+      "step": 12456
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015123379452906433,
+      "loss": 0.0122,
+      "step": 12457
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015121481349302123,
+      "loss": 0.0091,
+      "step": 12458
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015119583243752475,
+      "loss": 0.3805,
+      "step": 12459
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015117685136287876,
+      "loss": 1.3807,
+      "step": 12460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001511578702693873,
+      "loss": 0.617,
+      "step": 12461
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015113888915735426,
+      "loss": 0.2936,
+      "step": 12462
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015111990802708366,
+      "loss": 0.355,
+      "step": 12463
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015110092687887935,
+      "loss": 0.8052,
+      "step": 12464
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015108194571304535,
+      "loss": 0.4925,
+      "step": 12465
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001510629645298857,
+      "loss": 2.0049,
+      "step": 12466
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015104398332970419,
+      "loss": 0.3251,
+      "step": 12467
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015102500211280483,
+      "loss": 0.3263,
+      "step": 12468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015100602087949168,
+      "loss": 0.2388,
+      "step": 12469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001509870396300686,
+      "loss": 0.5339,
+      "step": 12470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015096805836483954,
+      "loss": 0.4359,
+      "step": 12471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015094907708410846,
+      "loss": 0.5755,
+      "step": 12472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015093009578817938,
+      "loss": 0.4742,
+      "step": 12473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001509111144773562,
+      "loss": 0.8254,
+      "step": 12474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015089213315194283,
+      "loss": 0.8646,
+      "step": 12475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015087315181224335,
+      "loss": 0.0007,
+      "step": 12476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015085417045856162,
+      "loss": 0.4305,
+      "step": 12477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001508351890912017,
+      "loss": 0.228,
+      "step": 12478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015081620771046738,
+      "loss": 0.7682,
+      "step": 12479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001507972263166628,
+      "loss": 0.2389,
+      "step": 12480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015077824491009178,
+      "loss": 0.4766,
+      "step": 12481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001507592634910584,
+      "loss": 0.0022,
+      "step": 12482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001507402820598665,
+      "loss": 0.0109,
+      "step": 12483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015072130061682013,
+      "loss": 0.642,
+      "step": 12484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015070231916222316,
+      "loss": 0.026,
+      "step": 12485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015068333769637963,
+      "loss": 0.0017,
+      "step": 12486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001506643562195935,
+      "loss": 0.1498,
+      "step": 12487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015064537473216864,
+      "loss": 0.0039,
+      "step": 12488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015062639323440915,
+      "loss": 0.1776,
+      "step": 12489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015060741172661887,
+      "loss": 0.6301,
+      "step": 12490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001505884302091018,
+      "loss": 0.0001,
+      "step": 12491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015056944868216196,
+      "loss": 0.6694,
+      "step": 12492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015055046714610316,
+      "loss": 0.1817,
+      "step": 12493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015053148560122947,
+      "loss": 0.0239,
+      "step": 12494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001505125040478449,
+      "loss": 0.2272,
+      "step": 12495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015049352248625332,
+      "loss": 0.8511,
+      "step": 12496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015047454091675867,
+      "loss": 0.8328,
+      "step": 12497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015045555933966502,
+      "loss": 0.365,
+      "step": 12498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015043657775527623,
+      "loss": 0.0098,
+      "step": 12499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015041759616389635,
+      "loss": 0.0025,
+      "step": 12500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001503986145658293,
+      "loss": 0.7966,
+      "step": 12501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015037963296137894,
+      "loss": 0.5919,
+      "step": 12502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001503606513508494,
+      "loss": 0.153,
+      "step": 12503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015034166973454459,
+      "loss": 0.5787,
+      "step": 12504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015032268811276844,
+      "loss": 0.5968,
+      "step": 12505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015030370648582488,
+      "loss": 2.4297,
+      "step": 12506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015028472485401797,
+      "loss": 0.4977,
+      "step": 12507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015026574321765158,
+      "loss": 0.4604,
+      "step": 12508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001502467615770297,
+      "loss": 0.0034,
+      "step": 12509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015022777993245633,
+      "loss": 0.2232,
+      "step": 12510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001502087982842354,
+      "loss": 0.2272,
+      "step": 12511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001501898166326709,
+      "loss": 0.3894,
+      "step": 12512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015017083497806673,
+      "loss": 0.9065,
+      "step": 12513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001501518533207269,
+      "loss": 0.416,
+      "step": 12514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015013287166095544,
+      "loss": 0.5654,
+      "step": 12515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015011388999905617,
+      "loss": 0.2144,
+      "step": 12516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001500949083353331,
+      "loss": 0.4744,
+      "step": 12517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015007592667009027,
+      "loss": 1.8608,
+      "step": 12518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015005694500363156,
+      "loss": 0.4755,
+      "step": 12519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000150037963336261,
+      "loss": 0.3048,
+      "step": 12520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015001898166828247,
+      "loss": 0.537,
+      "step": 12521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00015,
+      "loss": 0.2788,
+      "step": 12522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014998101833171753,
+      "loss": 0.0007,
+      "step": 12523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014996203666373903,
+      "loss": 0.5632,
+      "step": 12524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014994305499636842,
+      "loss": 0.3025,
+      "step": 12525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001499240733299097,
+      "loss": 0.587,
+      "step": 12526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001499050916646669,
+      "loss": 0.2178,
+      "step": 12527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014988611000094383,
+      "loss": 0.4434,
+      "step": 12528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014986712833904459,
+      "loss": 0.0074,
+      "step": 12529
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001498481466792731,
+      "loss": 0.0003,
+      "step": 12530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014982916502193324,
+      "loss": 0.211,
+      "step": 12531
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001498101833673291,
+      "loss": 0.7416,
+      "step": 12532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001497912017157646,
+      "loss": 0.1295,
+      "step": 12533
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014977222006754362,
+      "loss": 0.2681,
+      "step": 12534
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014975323842297028,
+      "loss": 0.2241,
+      "step": 12535
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014973425678234842,
+      "loss": 0.0716,
+      "step": 12536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001497152751459821,
+      "loss": 3.4453,
+      "step": 12537
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014969629351417512,
+      "loss": 0.4033,
+      "step": 12538
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001496773118872316,
+      "loss": 0.4279,
+      "step": 12539
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014965833026545544,
+      "loss": 1.0033,
+      "step": 12540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001496393486491506,
+      "loss": 0.4222,
+      "step": 12541
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000149620367038621,
+      "loss": 0.727,
+      "step": 12542
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014960138543417074,
+      "loss": 0.0002,
+      "step": 12543
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014958240383610366,
+      "loss": 0.3326,
+      "step": 12544
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014956342224472371,
+      "loss": 0.6115,
+      "step": 12545
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014954444066033498,
+      "loss": 0.0002,
+      "step": 12546
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001495254590832413,
+      "loss": 0.9567,
+      "step": 12547
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001495064775137467,
+      "loss": 0.8256,
+      "step": 12548
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001494874959521551,
+      "loss": 0.4705,
+      "step": 12549
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001494685143987705,
+      "loss": 0.3224,
+      "step": 12550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014944953285389687,
+      "loss": 0.7579,
+      "step": 12551
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014943055131783807,
+      "loss": 0.3467,
+      "step": 12552
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014941156979089818,
+      "loss": 0.0023,
+      "step": 12553
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014939258827338113,
+      "loss": 0.4445,
+      "step": 12554
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014937360676559082,
+      "loss": 1.0928,
+      "step": 12555
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001493546252678313,
+      "loss": 0.4628,
+      "step": 12556
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001493356437804065,
+      "loss": 0.4204,
+      "step": 12557
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014931666230362034,
+      "loss": 0.0018,
+      "step": 12558
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014929768083777687,
+      "loss": 4.1491,
+      "step": 12559
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001492786993831799,
+      "loss": 0.4622,
+      "step": 12560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014925971794013348,
+      "loss": 0.7881,
+      "step": 12561
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014924073650894164,
+      "loss": 0.344,
+      "step": 12562
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001492217550899082,
+      "loss": 0.2602,
+      "step": 12563
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014920277368333718,
+      "loss": 0.6782,
+      "step": 12564
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014918379228953262,
+      "loss": 0.4862,
+      "step": 12565
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001491648109087983,
+      "loss": 0.3399,
+      "step": 12566
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014914582954143836,
+      "loss": 0.0859,
+      "step": 12567
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014912684818775662,
+      "loss": 0.002,
+      "step": 12568
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014910786684805711,
+      "loss": 0.0002,
+      "step": 12569
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001490888855226438,
+      "loss": 0.3313,
+      "step": 12570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014906990421182062,
+      "loss": 0.788,
+      "step": 12571
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001490509229158915,
+      "loss": 0.5701,
+      "step": 12572
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001490319416351605,
+      "loss": 0.2583,
+      "step": 12573
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014901296036993138,
+      "loss": 0.6292,
+      "step": 12574
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001489939791205083,
+      "loss": 0.6977,
+      "step": 12575
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014897499788719514,
+      "loss": 0.382,
+      "step": 12576
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001489560166702958,
+      "loss": 0.3361,
+      "step": 12577
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014893703547011434,
+      "loss": 0.0885,
+      "step": 12578
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014891805428695462,
+      "loss": 0.0003,
+      "step": 12579
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014889907312112062,
+      "loss": 0.341,
+      "step": 12580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014888009197291637,
+      "loss": 0.0004,
+      "step": 12581
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014886111084264574,
+      "loss": 0.014,
+      "step": 12582
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014884212973061267,
+      "loss": 0.3791,
+      "step": 12583
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014882314863712124,
+      "loss": 0.1789,
+      "step": 12584
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014880416756247525,
+      "loss": 0.0611,
+      "step": 12585
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014878518650697877,
+      "loss": 0.2967,
+      "step": 12586
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014876620547093567,
+      "loss": 0.0014,
+      "step": 12587
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014874722445464993,
+      "loss": 0.2952,
+      "step": 12588
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014872824345842556,
+      "loss": 3.6445,
+      "step": 12589
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014870926248256643,
+      "loss": 0.3237,
+      "step": 12590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014869028152737649,
+      "loss": 0.4507,
+      "step": 12591
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014867130059315978,
+      "loss": 0.7888,
+      "step": 12592
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001486523196802202,
+      "loss": 0.6221,
+      "step": 12593
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014863333878886161,
+      "loss": 0.5786,
+      "step": 12594
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014861435791938815,
+      "loss": 0.7529,
+      "step": 12595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014859537707210364,
+      "loss": 0.2747,
+      "step": 12596
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014857639624731207,
+      "loss": 0.8432,
+      "step": 12597
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014855741544531732,
+      "loss": 0.0562,
+      "step": 12598
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014853843466642344,
+      "loss": 2.4649,
+      "step": 12599
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014851945391093434,
+      "loss": 1.1664,
+      "step": 12600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014850047317915393,
+      "loss": 0.3036,
+      "step": 12601
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001484814924713862,
+      "loss": 0.4399,
+      "step": 12602
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014846251178793514,
+      "loss": 0.897,
+      "step": 12603
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014844353112910457,
+      "loss": 0.011,
+      "step": 12604
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014842455049519855,
+      "loss": 0.4342,
+      "step": 12605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014840556988652103,
+      "loss": 0.6077,
+      "step": 12606
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014838658930337584,
+      "loss": 0.0041,
+      "step": 12607
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001483676087460671,
+      "loss": 0.0007,
+      "step": 12608
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014834862821489856,
+      "loss": 0.3753,
+      "step": 12609
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014832964771017428,
+      "loss": 2.4974,
+      "step": 12610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014831066723219825,
+      "loss": 0.5806,
+      "step": 12611
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014829168678127426,
+      "loss": 0.4568,
+      "step": 12612
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014827270635770637,
+      "loss": 3.4805,
+      "step": 12613
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014825372596179853,
+      "loss": 0.4281,
+      "step": 12614
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001482347455938546,
+      "loss": 0.0071,
+      "step": 12615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014821576525417862,
+      "loss": 0.5878,
+      "step": 12616
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014819678494307448,
+      "loss": 0.6958,
+      "step": 12617
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014817780466084603,
+      "loss": 1.7344,
+      "step": 12618
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001481588244077974,
+      "loss": 0.1706,
+      "step": 12619
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014813984418423243,
+      "loss": 0.0038,
+      "step": 12620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014812086399045498,
+      "loss": 0.3019,
+      "step": 12621
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001481018838267692,
+      "loss": 0.1679,
+      "step": 12622
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001480829036934788,
+      "loss": 0.357,
+      "step": 12623
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014806392359088783,
+      "loss": 1.0407,
+      "step": 12624
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001480449435193003,
+      "loss": 0.7926,
+      "step": 12625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014802596347901996,
+      "loss": 0.7117,
+      "step": 12626
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014800698347035095,
+      "loss": 0.2861,
+      "step": 12627
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014798800349359706,
+      "loss": 0.5712,
+      "step": 12628
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014796902354906224,
+      "loss": 0.7036,
+      "step": 12629
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001479500436370505,
+      "loss": 0.6025,
+      "step": 12630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014793106375786573,
+      "loss": 0.4081,
+      "step": 12631
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014791208391181184,
+      "loss": 0.2929,
+      "step": 12632
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014789310409919285,
+      "loss": 1.0556,
+      "step": 12633
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001478741243203126,
+      "loss": 0.6474,
+      "step": 12634
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014785514457547509,
+      "loss": 0.6178,
+      "step": 12635
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014783616486498417,
+      "loss": 0.3038,
+      "step": 12636
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014781718518914384,
+      "loss": 2.416,
+      "step": 12637
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014779820554825806,
+      "loss": 0.0145,
+      "step": 12638
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014777922594263064,
+      "loss": 0.5718,
+      "step": 12639
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001477602463725656,
+      "loss": 1.0049,
+      "step": 12640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014774126683836692,
+      "loss": 0.5018,
+      "step": 12641
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001477222873403384,
+      "loss": 0.4829,
+      "step": 12642
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014770330787878402,
+      "loss": 0.0042,
+      "step": 12643
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014768432845400776,
+      "loss": 0.3613,
+      "step": 12644
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014766534906631348,
+      "loss": 0.3093,
+      "step": 12645
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014764636971600515,
+      "loss": 0.2226,
+      "step": 12646
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014762739040338665,
+      "loss": 0.4292,
+      "step": 12647
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014760841112876195,
+      "loss": 0.3273,
+      "step": 12648
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014758943189243496,
+      "loss": 0.1903,
+      "step": 12649
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014757045269470954,
+      "loss": 0.3272,
+      "step": 12650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014755147353588975,
+      "loss": 0.4009,
+      "step": 12651
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014753249441627945,
+      "loss": 0.266,
+      "step": 12652
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014751351533618247,
+      "loss": 0.0116,
+      "step": 12653
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014749453629590284,
+      "loss": 0.7063,
+      "step": 12654
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014747555729574448,
+      "loss": 0.4974,
+      "step": 12655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014745657833601123,
+      "loss": 0.4583,
+      "step": 12656
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014743759941700715,
+      "loss": 2.0664,
+      "step": 12657
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014741862053903597,
+      "loss": 0.2392,
+      "step": 12658
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014739964170240176,
+      "loss": 0.3218,
+      "step": 12659
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001473806629074084,
+      "loss": 0.4332,
+      "step": 12660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014736168415435975,
+      "loss": 0.4377,
+      "step": 12661
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014734270544355978,
+      "loss": 0.3504,
+      "step": 12662
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014732372677531244,
+      "loss": 0.0707,
+      "step": 12663
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014730474814992153,
+      "loss": 0.2339,
+      "step": 12664
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014728576956769108,
+      "loss": 0.0106,
+      "step": 12665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014726679102892493,
+      "loss": 0.8067,
+      "step": 12666
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000147247812533927,
+      "loss": 0.339,
+      "step": 12667
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001472288340830013,
+      "loss": 0.5881,
+      "step": 12668
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001472098556764516,
+      "loss": 0.3387,
+      "step": 12669
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014719087731458185,
+      "loss": 0.3986,
+      "step": 12670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014717189899769606,
+      "loss": 0.5367,
+      "step": 12671
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014715292072609801,
+      "loss": 1.5782,
+      "step": 12672
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014713394250009167,
+      "loss": 0.0116,
+      "step": 12673
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014711496431998096,
+      "loss": 0.4601,
+      "step": 12674
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014709598618606976,
+      "loss": 0.3413,
+      "step": 12675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000147077008098662,
+      "loss": 1.5664,
+      "step": 12676
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014705803005806155,
+      "loss": 0.0136,
+      "step": 12677
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014703905206457228,
+      "loss": 0.835,
+      "step": 12678
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014702007411849824,
+      "loss": 2.7234,
+      "step": 12679
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001470010962201432,
+      "loss": 0.0079,
+      "step": 12680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014698211836981106,
+      "loss": 0.1665,
+      "step": 12681
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014696314056780583,
+      "loss": 0.6209,
+      "step": 12682
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014694416281443132,
+      "loss": 0.013,
+      "step": 12683
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001469251851099915,
+      "loss": 0.2454,
+      "step": 12684
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014690620745479014,
+      "loss": 0.4848,
+      "step": 12685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014688722984913127,
+      "loss": 0.1745,
+      "step": 12686
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014686825229331878,
+      "loss": 0.0122,
+      "step": 12687
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014684927478765644,
+      "loss": 0.3674,
+      "step": 12688
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001468302973324483,
+      "loss": 0.4569,
+      "step": 12689
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001468113199279982,
+      "loss": 0.0437,
+      "step": 12690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014679234257460996,
+      "loss": 0.0084,
+      "step": 12691
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001467733652725876,
+      "loss": 0.9253,
+      "step": 12692
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014675438802223497,
+      "loss": 0.5178,
+      "step": 12693
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001467354108238559,
+      "loss": 0.2982,
+      "step": 12694
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014671643367775438,
+      "loss": 0.6377,
+      "step": 12695
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014669745658423415,
+      "loss": 0.283,
+      "step": 12696
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014667847954359924,
+      "loss": 0.0699,
+      "step": 12697
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014665950255615353,
+      "loss": 0.6709,
+      "step": 12698
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001466405256222008,
+      "loss": 0.1637,
+      "step": 12699
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001466215487420451,
+      "loss": 0.5273,
+      "step": 12700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001466025719159902,
+      "loss": 0.08,
+      "step": 12701
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014658359514433997,
+      "loss": 0.7959,
+      "step": 12702
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014656461842739836,
+      "loss": 0.2456,
+      "step": 12703
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014654564176546926,
+      "loss": 0.5276,
+      "step": 12704
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014652666515885645,
+      "loss": 0.2509,
+      "step": 12705
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014650768860786397,
+      "loss": 0.0059,
+      "step": 12706
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014648871211279558,
+      "loss": 0.2311,
+      "step": 12707
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014646973567395514,
+      "loss": 0.9338,
+      "step": 12708
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001464507592916467,
+      "loss": 0.5942,
+      "step": 12709
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014643178296617395,
+      "loss": 0.5775,
+      "step": 12710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014641280669784083,
+      "loss": 0.6467,
+      "step": 12711
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014639383048695127,
+      "loss": 0.7641,
+      "step": 12712
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014637485433380905,
+      "loss": 0.0324,
+      "step": 12713
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014635587823871816,
+      "loss": 0.0863,
+      "step": 12714
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001463369022019824,
+      "loss": 0.483,
+      "step": 12715
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014631792622390557,
+      "loss": 0.0428,
+      "step": 12716
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014629895030479173,
+      "loss": 0.4826,
+      "step": 12717
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001462799744449446,
+      "loss": 0.4108,
+      "step": 12718
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014626099864466807,
+      "loss": 0.0054,
+      "step": 12719
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001462420229042661,
+      "loss": 1.0386,
+      "step": 12720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014622304722404245,
+      "loss": 0.0766,
+      "step": 12721
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000146204071604301,
+      "loss": 0.5288,
+      "step": 12722
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001461850960453457,
+      "loss": 0.25,
+      "step": 12723
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014616612054748037,
+      "loss": 0.5529,
+      "step": 12724
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014614714511100886,
+      "loss": 0.635,
+      "step": 12725
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014612816973623503,
+      "loss": 0.8647,
+      "step": 12726
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014610919442346271,
+      "loss": 0.0031,
+      "step": 12727
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014609021917299587,
+      "loss": 0.3229,
+      "step": 12728
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014607124398513825,
+      "loss": 0.7654,
+      "step": 12729
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014605226886019377,
+      "loss": 0.2834,
+      "step": 12730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001460332937984663,
+      "loss": 0.6694,
+      "step": 12731
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014601431880025966,
+      "loss": 0.184,
+      "step": 12732
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014599534386587778,
+      "loss": 1.0615,
+      "step": 12733
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014597636899562437,
+      "loss": 0.3691,
+      "step": 12734
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014595739418980338,
+      "loss": 0.4766,
+      "step": 12735
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001459384194487187,
+      "loss": 0.3724,
+      "step": 12736
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001459194447726741,
+      "loss": 0.6299,
+      "step": 12737
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001459004701619735,
+      "loss": 0.558,
+      "step": 12738
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001458814956169207,
+      "loss": 0.2423,
+      "step": 12739
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014586252113781955,
+      "loss": 0.3649,
+      "step": 12740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014584354672497393,
+      "loss": 0.0681,
+      "step": 12741
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014582457237868773,
+      "loss": 0.939,
+      "step": 12742
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014580559809926466,
+      "loss": 0.3692,
+      "step": 12743
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014578662388700872,
+      "loss": 0.2981,
+      "step": 12744
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014576764974222358,
+      "loss": 0.005,
+      "step": 12745
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001457486756652132,
+      "loss": 0.2761,
+      "step": 12746
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014572970165628146,
+      "loss": 0.8012,
+      "step": 12747
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014571072771573206,
+      "loss": 0.3143,
+      "step": 12748
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014569175384386896,
+      "loss": 1.0613,
+      "step": 12749
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014567278004099597,
+      "loss": 0.3345,
+      "step": 12750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014565380630741687,
+      "loss": 0.5274,
+      "step": 12751
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014563483264343555,
+      "loss": 0.6579,
+      "step": 12752
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014561585904935588,
+      "loss": 0.4243,
+      "step": 12753
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001455968855254816,
+      "loss": 0.0012,
+      "step": 12754
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014557791207211662,
+      "loss": 0.0026,
+      "step": 12755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014555893868956472,
+      "loss": 2.6446,
+      "step": 12756
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014553996537812973,
+      "loss": 0.3854,
+      "step": 12757
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014552099213811556,
+      "loss": 0.0031,
+      "step": 12758
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014550201896982594,
+      "loss": 0.3116,
+      "step": 12759
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001454830458735647,
+      "loss": 2.3255,
+      "step": 12760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014546407284963582,
+      "loss": 0.1136,
+      "step": 12761
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001454450998983429,
+      "loss": 0.0085,
+      "step": 12762
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014542612701998992,
+      "loss": 0.1112,
+      "step": 12763
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014540715421488065,
+      "loss": 0.408,
+      "step": 12764
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014538818148331884,
+      "loss": 0.0216,
+      "step": 12765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014536920882560848,
+      "loss": 0.4391,
+      "step": 12766
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014535023624205325,
+      "loss": 0.5625,
+      "step": 12767
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014533126373295695,
+      "loss": 0.0043,
+      "step": 12768
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014531229129862355,
+      "loss": 2.5692,
+      "step": 12769
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014529331893935673,
+      "loss": 0.0018,
+      "step": 12770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001452743466554603,
+      "loss": 0.7485,
+      "step": 12771
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001452553744472382,
+      "loss": 0.5991,
+      "step": 12772
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001452364023149941,
+      "loss": 0.342,
+      "step": 12773
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014521743025903194,
+      "loss": 1.2217,
+      "step": 12774
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014519845827965536,
+      "loss": 0.6269,
+      "step": 12775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014517948637716828,
+      "loss": 0.8965,
+      "step": 12776
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014516051455187455,
+      "loss": 0.292,
+      "step": 12777
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001451415428040779,
+      "loss": 0.8427,
+      "step": 12778
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014512257113408208,
+      "loss": 0.6428,
+      "step": 12779
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014510359954219106,
+      "loss": 0.5752,
+      "step": 12780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001450846280287085,
+      "loss": 0.0072,
+      "step": 12781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014506565659393826,
+      "loss": 0.3874,
+      "step": 12782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001450466852381841,
+      "loss": 0.2751,
+      "step": 12783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014502771396174985,
+      "loss": 0.6275,
+      "step": 12784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014500874276493934,
+      "loss": 0.2205,
+      "step": 12785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014498977164805623,
+      "loss": 1.9658,
+      "step": 12786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014497080061140447,
+      "loss": 0.5984,
+      "step": 12787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001449518296552878,
+      "loss": 0.7572,
+      "step": 12788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014493285878000996,
+      "loss": 0.5276,
+      "step": 12789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014491388798587479,
+      "loss": 0.419,
+      "step": 12790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014489491727318613,
+      "loss": 0.4861,
+      "step": 12791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014487594664224765,
+      "loss": 0.0016,
+      "step": 12792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014485697609336326,
+      "loss": 0.3299,
+      "step": 12793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001448380056268366,
+      "loss": 0.3588,
+      "step": 12794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014481903524297157,
+      "loss": 0.6495,
+      "step": 12795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014480006494207193,
+      "loss": 0.0214,
+      "step": 12796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014478109472444137,
+      "loss": 1.5152,
+      "step": 12797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001447621245903838,
+      "loss": 0.0186,
+      "step": 12798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014474315454020297,
+      "loss": 0.0022,
+      "step": 12799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001447241845742026,
+      "loss": 0.8128,
+      "step": 12800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014470521469268647,
+      "loss": 0.3502,
+      "step": 12801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014468624489595846,
+      "loss": 0.3078,
+      "step": 12802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014466727518432218,
+      "loss": 0.0318,
+      "step": 12803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014464830555808154,
+      "loss": 0.5694,
+      "step": 12804
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014462933601754022,
+      "loss": 0.0257,
+      "step": 12805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000144610366563002,
+      "loss": 0.8711,
+      "step": 12806
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014459139719477074,
+      "loss": 0.6383,
+      "step": 12807
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001445724279131501,
+      "loss": 0.3575,
+      "step": 12808
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014455345871844386,
+      "loss": 0.5426,
+      "step": 12809
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001445344896109559,
+      "loss": 0.2641,
+      "step": 12810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014451552059098979,
+      "loss": 0.213,
+      "step": 12811
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014449655165884941,
+      "loss": 0.3589,
+      "step": 12812
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014447758281483852,
+      "loss": 0.0108,
+      "step": 12813
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001444586140592608,
+      "loss": 0.0068,
+      "step": 12814
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014443964539242012,
+      "loss": 0.0137,
+      "step": 12815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014442067681462014,
+      "loss": 0.0071,
+      "step": 12816
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001444017083261646,
+      "loss": 0.8181,
+      "step": 12817
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001443827399273574,
+      "loss": 0.5492,
+      "step": 12818
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001443637716185021,
+      "loss": 1.8616,
+      "step": 12819
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014434480339990257,
+      "loss": 0.4528,
+      "step": 12820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014432583527186258,
+      "loss": 0.3175,
+      "step": 12821
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014430686723468578,
+      "loss": 0.4909,
+      "step": 12822
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014428789928867597,
+      "loss": 0.4738,
+      "step": 12823
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014426893143413685,
+      "loss": 0.4219,
+      "step": 12824
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001442499636713722,
+      "loss": 2.8594,
+      "step": 12825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014423099600068578,
+      "loss": 0.0063,
+      "step": 12826
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001442120284223813,
+      "loss": 1.2135,
+      "step": 12827
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014419306093676242,
+      "loss": 0.6387,
+      "step": 12828
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014417409354413306,
+      "loss": 0.2544,
+      "step": 12829
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001441551262447968,
+      "loss": 0.0047,
+      "step": 12830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014413615903905745,
+      "loss": 0.3457,
+      "step": 12831
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014411719192721864,
+      "loss": 0.1923,
+      "step": 12832
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001440982249095842,
+      "loss": 0.0046,
+      "step": 12833
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001440792579864579,
+      "loss": 0.0107,
+      "step": 12834
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014406029115814333,
+      "loss": 0.2842,
+      "step": 12835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00014404132442494426,
+      "loss": 0.0109,
+      "step": 12836
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014402235778716453,
+      "loss": 0.862,
+      "step": 12837
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014400339124510768,
+      "loss": 0.2216,
+      "step": 12838
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014398442479907756,
+      "loss": 1.8112,
+      "step": 12839
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014396545844937787,
+      "loss": 0.7378,
+      "step": 12840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014394649219631224,
+      "loss": 1.0675,
+      "step": 12841
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014392752604018452,
+      "loss": 0.2818,
+      "step": 12842
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014390855998129833,
+      "loss": 0.846,
+      "step": 12843
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014388959401995737,
+      "loss": 0.2899,
+      "step": 12844
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001438706281564655,
+      "loss": 0.6172,
+      "step": 12845
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001438516623911262,
+      "loss": 0.0116,
+      "step": 12846
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014383269672424336,
+      "loss": 0.719,
+      "step": 12847
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014381373115612066,
+      "loss": 2.1095,
+      "step": 12848
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014379476568706167,
+      "loss": 0.2742,
+      "step": 12849
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014377580031737028,
+      "loss": 0.4028,
+      "step": 12850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001437568350473501,
+      "loss": 0.8403,
+      "step": 12851
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014373786987730477,
+      "loss": 0.7941,
+      "step": 12852
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014371890480753815,
+      "loss": 0.8083,
+      "step": 12853
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001436999398383538,
+      "loss": 0.2736,
+      "step": 12854
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001436809749700554,
+      "loss": 0.3733,
+      "step": 12855
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001436620102029468,
+      "loss": 0.4642,
+      "step": 12856
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014364304553733152,
+      "loss": 0.7566,
+      "step": 12857
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014362408097351334,
+      "loss": 0.278,
+      "step": 12858
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014360511651179597,
+      "loss": 0.4484,
+      "step": 12859
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014358615215248304,
+      "loss": 0.643,
+      "step": 12860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001435671878958783,
+      "loss": 1.4632,
+      "step": 12861
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014354822374228535,
+      "loss": 0.5863,
+      "step": 12862
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014352925969200788,
+      "loss": 0.107,
+      "step": 12863
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014351029574534966,
+      "loss": 4.1417,
+      "step": 12864
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001434913319026143,
+      "loss": 0.374,
+      "step": 12865
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014347236816410547,
+      "loss": 0.0065,
+      "step": 12866
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014345340453012688,
+      "loss": 0.03,
+      "step": 12867
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014343444100098221,
+      "loss": 1.0283,
+      "step": 12868
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014341547757697507,
+      "loss": 0.5024,
+      "step": 12869
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014339651425840922,
+      "loss": 0.4104,
+      "step": 12870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014337755104558826,
+      "loss": 0.2679,
+      "step": 12871
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001433585879388159,
+      "loss": 0.2456,
+      "step": 12872
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014333962493839577,
+      "loss": 0.2328,
+      "step": 12873
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014332066204463155,
+      "loss": 0.2077,
+      "step": 12874
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014330169925782694,
+      "loss": 0.8265,
+      "step": 12875
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014328273657828548,
+      "loss": 0.3109,
+      "step": 12876
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014326377400631095,
+      "loss": 0.7937,
+      "step": 12877
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014324481154220702,
+      "loss": 0.0061,
+      "step": 12878
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014322584918627726,
+      "loss": 0.3087,
+      "step": 12879
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001432068869388254,
+      "loss": 0.1421,
+      "step": 12880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014318792480015496,
+      "loss": 0.0222,
+      "step": 12881
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014316896277056972,
+      "loss": 0.8784,
+      "step": 12882
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014315000085037336,
+      "loss": 0.5335,
+      "step": 12883
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014313103903986933,
+      "loss": 0.2256,
+      "step": 12884
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014311207733936149,
+      "loss": 0.4594,
+      "step": 12885
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014309311574915338,
+      "loss": 0.1563,
+      "step": 12886
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001430741542695486,
+      "loss": 0.1113,
+      "step": 12887
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014305519290085092,
+      "loss": 0.0171,
+      "step": 12888
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001430362316433639,
+      "loss": 0.7774,
+      "step": 12889
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014301727049739112,
+      "loss": 0.4858,
+      "step": 12890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014299830946323632,
+      "loss": 0.0341,
+      "step": 12891
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014297934854120306,
+      "loss": 2.4837,
+      "step": 12892
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014296038773159499,
+      "loss": 0.1712,
+      "step": 12893
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001429414270347158,
+      "loss": 0.3051,
+      "step": 12894
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000142922466450869,
+      "loss": 0.0029,
+      "step": 12895
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001429035059803583,
+      "loss": 0.6972,
+      "step": 12896
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014288454562348733,
+      "loss": 0.3972,
+      "step": 12897
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014286558538055964,
+      "loss": 0.9385,
+      "step": 12898
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001428466252518789,
+      "loss": 0.4117,
+      "step": 12899
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014282766523774875,
+      "loss": 0.4558,
+      "step": 12900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014280870533847273,
+      "loss": 0.3564,
+      "step": 12901
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014278974555435457,
+      "loss": 0.4786,
+      "step": 12902
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014277078588569776,
+      "loss": 0.0034,
+      "step": 12903
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014275182633280596,
+      "loss": 0.8699,
+      "step": 12904
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014273286689598283,
+      "loss": 0.0542,
+      "step": 12905
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001427139075755319,
+      "loss": 0.4618,
+      "step": 12906
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001426949483717568,
+      "loss": 0.0031,
+      "step": 12907
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014267598928496115,
+      "loss": 0.014,
+      "step": 12908
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014265703031544853,
+      "loss": 0.9497,
+      "step": 12909
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001426380714635226,
+      "loss": 0.3125,
+      "step": 12910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014261911272948682,
+      "loss": 1.5729,
+      "step": 12911
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001426001541136449,
+      "loss": 0.9581,
+      "step": 12912
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014258119561630044,
+      "loss": 0.2595,
+      "step": 12913
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000142562237237757,
+      "loss": 0.3085,
+      "step": 12914
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014254327897831812,
+      "loss": 0.2557,
+      "step": 12915
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014252432083828746,
+      "loss": 0.3507,
+      "step": 12916
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014250536281796859,
+      "loss": 0.3786,
+      "step": 12917
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014248640491766506,
+      "loss": 0.0071,
+      "step": 12918
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014246744713768052,
+      "loss": 0.7982,
+      "step": 12919
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001424484894783185,
+      "loss": 0.0051,
+      "step": 12920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014242953193988262,
+      "loss": 0.0093,
+      "step": 12921
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014241057452267633,
+      "loss": 0.2335,
+      "step": 12922
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014239161722700335,
+      "loss": 0.0775,
+      "step": 12923
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014237266005316722,
+      "loss": 1.6387,
+      "step": 12924
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014235370300147145,
+      "loss": 0.2457,
+      "step": 12925
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001423347460722197,
+      "loss": 0.8385,
+      "step": 12926
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014231578926571546,
+      "loss": 0.3348,
+      "step": 12927
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001422968325822623,
+      "loss": 0.3565,
+      "step": 12928
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001422778760221639,
+      "loss": 0.4047,
+      "step": 12929
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014225891958572363,
+      "loss": 0.3791,
+      "step": 12930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014223996327324516,
+      "loss": 0.7578,
+      "step": 12931
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014222100708503207,
+      "loss": 1.0191,
+      "step": 12932
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014220205102138782,
+      "loss": 2.0796,
+      "step": 12933
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014218309508261606,
+      "loss": 0.3785,
+      "step": 12934
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014216413926902032,
+      "loss": 0.2002,
+      "step": 12935
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014214518358090407,
+      "loss": 0.1597,
+      "step": 12936
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014212622801857094,
+      "loss": 0.8701,
+      "step": 12937
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014210727258232448,
+      "loss": 0.0338,
+      "step": 12938
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014208831727246816,
+      "loss": 2.1268,
+      "step": 12939
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014206936208930563,
+      "loss": 0.1794,
+      "step": 12940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001420504070331403,
+      "loss": 0.4227,
+      "step": 12941
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014203145210427577,
+      "loss": 0.0112,
+      "step": 12942
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014201249730301564,
+      "loss": 0.129,
+      "step": 12943
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014199354262966332,
+      "loss": 0.4209,
+      "step": 12944
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014197458808452238,
+      "loss": 0.0019,
+      "step": 12945
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014195563366789645,
+      "loss": 0.4801,
+      "step": 12946
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014193667938008892,
+      "loss": 0.7349,
+      "step": 12947
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014191772522140336,
+      "loss": 0.1258,
+      "step": 12948
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014189877119214335,
+      "loss": 0.5752,
+      "step": 12949
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001418798172926123,
+      "loss": 0.3256,
+      "step": 12950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014186086352311385,
+      "loss": 0.0037,
+      "step": 12951
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001418419098839514,
+      "loss": 0.356,
+      "step": 12952
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014182295637542853,
+      "loss": 0.4027,
+      "step": 12953
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014180400299784878,
+      "loss": 0.3379,
+      "step": 12954
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014178504975151558,
+      "loss": 0.003,
+      "step": 12955
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014176609663673248,
+      "loss": 0.179,
+      "step": 12956
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000141747143653803,
+      "loss": 0.123,
+      "step": 12957
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014172819080303064,
+      "loss": 0.539,
+      "step": 12958
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001417092380847189,
+      "loss": 0.4902,
+      "step": 12959
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001416902854991712,
+      "loss": 2.002,
+      "step": 12960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014167133304669117,
+      "loss": 0.0078,
+      "step": 12961
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014165238072758228,
+      "loss": 0.0016,
+      "step": 12962
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014163342854214793,
+      "loss": 0.4065,
+      "step": 12963
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001416144764906916,
+      "loss": 3.1671,
+      "step": 12964
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014159552457351695,
+      "loss": 0.3076,
+      "step": 12965
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014157657279092732,
+      "loss": 0.9856,
+      "step": 12966
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001415576211432262,
+      "loss": 0.0079,
+      "step": 12967
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014153866963071717,
+      "loss": 0.1314,
+      "step": 12968
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014151971825370364,
+      "loss": 0.0785,
+      "step": 12969
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014150076701248908,
+      "loss": 0.5581,
+      "step": 12970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014148181590737695,
+      "loss": 1.0009,
+      "step": 12971
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014146286493867078,
+      "loss": 0.6904,
+      "step": 12972
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014144391410667403,
+      "loss": 0.0118,
+      "step": 12973
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014142496341169013,
+      "loss": 0.0031,
+      "step": 12974
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014140601285402257,
+      "loss": 0.1046,
+      "step": 12975
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014138706243397485,
+      "loss": 0.796,
+      "step": 12976
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014136811215185034,
+      "loss": 0.0488,
+      "step": 12977
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014134916200795263,
+      "loss": 0.7749,
+      "step": 12978
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001413302120025851,
+      "loss": 0.1936,
+      "step": 12979
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014131126213605115,
+      "loss": 0.4693,
+      "step": 12980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014129231240865437,
+      "loss": 0.0573,
+      "step": 12981
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014127336282069807,
+      "loss": 0.6878,
+      "step": 12982
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001412544133724858,
+      "loss": 0.0027,
+      "step": 12983
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014123546406432101,
+      "loss": 1.1858,
+      "step": 12984
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014121651489650702,
+      "loss": 0.5052,
+      "step": 12985
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014119756586934742,
+      "loss": 0.591,
+      "step": 12986
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001411786169831456,
+      "loss": 1.9404,
+      "step": 12987
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014115966823820497,
+      "loss": 0.8514,
+      "step": 12988
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014114071963482902,
+      "loss": 0.1926,
+      "step": 12989
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014112177117332113,
+      "loss": 0.0182,
+      "step": 12990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001411028228539847,
+      "loss": 0.4772,
+      "step": 12991
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014108387467712328,
+      "loss": 0.3694,
+      "step": 12992
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001410649266430402,
+      "loss": 0.0045,
+      "step": 12993
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014104597875203885,
+      "loss": 0.0136,
+      "step": 12994
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014102703100442283,
+      "loss": 0.9746,
+      "step": 12995
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014100808340049536,
+      "loss": 0.2168,
+      "step": 12996
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014098913594055994,
+      "loss": 0.3938,
+      "step": 12997
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014097018862492007,
+      "loss": 0.2363,
+      "step": 12998
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014095124145387897,
+      "loss": 0.6797,
+      "step": 12999
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014093229442774022,
+      "loss": 0.0563,
+      "step": 13000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014091334754680717,
+      "loss": 0.6316,
+      "step": 13001
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014089440081138316,
+      "loss": 0.1015,
+      "step": 13002
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001408754542217717,
+      "loss": 0.7045,
+      "step": 13003
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014085650777827615,
+      "loss": 0.0019,
+      "step": 13004
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014083756148119985,
+      "loss": 0.6804,
+      "step": 13005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014081861533084635,
+      "loss": 0.6374,
+      "step": 13006
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014079966932751888,
+      "loss": 0.9463,
+      "step": 13007
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014078072347152094,
+      "loss": 2.8186,
+      "step": 13008
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014076177776315582,
+      "loss": 0.0045,
+      "step": 13009
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000140742832202727,
+      "loss": 0.4734,
+      "step": 13010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014072388679053785,
+      "loss": 0.2244,
+      "step": 13011
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014070494152689168,
+      "loss": 0.6542,
+      "step": 13012
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014070494152689168,
+      "loss": 0.9839,
+      "step": 13013
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014068599641209193,
+      "loss": 0.331,
+      "step": 13014
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014066705144644202,
+      "loss": 0.2107,
+      "step": 13015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014064810663024524,
+      "loss": 0.2838,
+      "step": 13016
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014062916196380494,
+      "loss": 0.3034,
+      "step": 13017
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014061021744742465,
+      "loss": 0.6226,
+      "step": 13018
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014059127308140758,
+      "loss": 0.143,
+      "step": 13019
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014057232886605718,
+      "loss": 1.8877,
+      "step": 13020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014055338480167676,
+      "loss": 2.3575,
+      "step": 13021
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001405344408885697,
+      "loss": 0.4784,
+      "step": 13022
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001405154971270394,
+      "loss": 0.5465,
+      "step": 13023
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014049655351738911,
+      "loss": 1.3777,
+      "step": 13024
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001404776100599223,
+      "loss": 0.6439,
+      "step": 13025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001404586667549423,
+      "loss": 0.4573,
+      "step": 13026
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014043972360275236,
+      "loss": 0.5717,
+      "step": 13027
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014042078060365597,
+      "loss": 0.2462,
+      "step": 13028
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014040183775795636,
+      "loss": 0.6796,
+      "step": 13029
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001403828950659569,
+      "loss": 0.2536,
+      "step": 13030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014036395252796103,
+      "loss": 0.3503,
+      "step": 13031
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001403450101442719,
+      "loss": 0.0016,
+      "step": 13032
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000140326067915193,
+      "loss": 0.2402,
+      "step": 13033
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001403071258410276,
+      "loss": 0.3342,
+      "step": 13034
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014028818392207898,
+      "loss": 0.2576,
+      "step": 13035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014026924215865055,
+      "loss": 0.2723,
+      "step": 13036
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014025030055104563,
+      "loss": 0.0105,
+      "step": 13037
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014023135909956746,
+      "loss": 0.0737,
+      "step": 13038
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001402124178045195,
+      "loss": 0.7665,
+      "step": 13039
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014019347666620492,
+      "loss": 0.0028,
+      "step": 13040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014017453568492707,
+      "loss": 2.041,
+      "step": 13041
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014015559486098934,
+      "loss": 0.2361,
+      "step": 13042
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014013665419469496,
+      "loss": 0.1656,
+      "step": 13043
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001401177136863472,
+      "loss": 0.2563,
+      "step": 13044
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014009877333624952,
+      "loss": 0.7101,
+      "step": 13045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001400798331447051,
+      "loss": 1.2143,
+      "step": 13046
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014006089311201721,
+      "loss": 0.0029,
+      "step": 13047
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001400419532384893,
+      "loss": 0.7943,
+      "step": 13048
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014002301352442447,
+      "loss": 0.3303,
+      "step": 13049
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00014000407397012616,
+      "loss": 0.0022,
+      "step": 13050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001399851345758976,
+      "loss": 1.035,
+      "step": 13051
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013996619534204202,
+      "loss": 0.3179,
+      "step": 13052
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013994725626886284,
+      "loss": 0.0236,
+      "step": 13053
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013992831735666324,
+      "loss": 0.5316,
+      "step": 13054
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001399093786057465,
+      "loss": 0.2596,
+      "step": 13055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013989044001641593,
+      "loss": 0.8552,
+      "step": 13056
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001398715015889748,
+      "loss": 0.945,
+      "step": 13057
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013985256332372639,
+      "loss": 0.1248,
+      "step": 13058
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013983362522097388,
+      "loss": 0.4293,
+      "step": 13059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013981468728102066,
+      "loss": 0.3175,
+      "step": 13060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013979574950416995,
+      "loss": 0.197,
+      "step": 13061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013977681189072494,
+      "loss": 0.3318,
+      "step": 13062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013975787444098898,
+      "loss": 0.2801,
+      "step": 13063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001397389371552653,
+      "loss": 0.5079,
+      "step": 13064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013972000003385708,
+      "loss": 0.4395,
+      "step": 13065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013970106307706768,
+      "loss": 0.3882,
+      "step": 13066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013968212628520034,
+      "loss": 0.2342,
+      "step": 13067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001396631896585582,
+      "loss": 0.1848,
+      "step": 13068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013964425319744464,
+      "loss": 0.0048,
+      "step": 13069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001396253169021627,
+      "loss": 0.4166,
+      "step": 13070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013960638077301584,
+      "loss": 0.0186,
+      "step": 13071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013958744481030718,
+      "loss": 0.796,
+      "step": 13072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001395685090143399,
+      "loss": 0.2711,
+      "step": 13073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013954957338541734,
+      "loss": 0.506,
+      "step": 13074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001395306379238427,
+      "loss": 0.4516,
+      "step": 13075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001395117026299191,
+      "loss": 0.5052,
+      "step": 13076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013949276750394995,
+      "loss": 0.0154,
+      "step": 13077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013947383254623831,
+      "loss": 0.9378,
+      "step": 13078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001394548977570874,
+      "loss": 0.5135,
+      "step": 13079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013943596313680055,
+      "loss": 0.153,
+      "step": 13080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013941702868568086,
+      "loss": 0.007,
+      "step": 13081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013939809440403154,
+      "loss": 0.6807,
+      "step": 13082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013937916029215593,
+      "loss": 0.5811,
+      "step": 13083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013936022635035702,
+      "loss": 0.3021,
+      "step": 13084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013934129257893817,
+      "loss": 0.2857,
+      "step": 13085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00013932235897820254,
+      "loss": 0.4496,
+      "step": 13086
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013930342554845324,
+      "loss": 0.1397,
+      "step": 13087
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001392844922899936,
+      "loss": 0.0565,
+      "step": 13088
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001392655592031267,
+      "loss": 0.5691,
+      "step": 13089
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001392466262881557,
+      "loss": 0.3219,
+      "step": 13090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013922769354538395,
+      "loss": 0.8023,
+      "step": 13091
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013920876097511445,
+      "loss": 0.8918,
+      "step": 13092
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013918982857765046,
+      "loss": 0.1678,
+      "step": 13093
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013917089635329517,
+      "loss": 0.29,
+      "step": 13094
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013915196430235168,
+      "loss": 0.0046,
+      "step": 13095
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013913303242512319,
+      "loss": 2.5123,
+      "step": 13096
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013911410072191293,
+      "loss": 0.4629,
+      "step": 13097
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000139095169193024,
+      "loss": 0.5441,
+      "step": 13098
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013907623783875959,
+      "loss": 0.7822,
+      "step": 13099
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001390573066594228,
+      "loss": 0.5184,
+      "step": 13100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001390383756553168,
+      "loss": 0.1978,
+      "step": 13101
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013901944482674482,
+      "loss": 0.0075,
+      "step": 13102
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013900051417400992,
+      "loss": 0.0103,
+      "step": 13103
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013898158369741523,
+      "loss": 1.4551,
+      "step": 13104
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013896265339726402,
+      "loss": 0.2204,
+      "step": 13105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013894372327385933,
+      "loss": 0.2063,
+      "step": 13106
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013892479332750434,
+      "loss": 0.5189,
+      "step": 13107
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001389058635585021,
+      "loss": 0.3878,
+      "step": 13108
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013888693396715582,
+      "loss": 0.3423,
+      "step": 13109
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013886800455376866,
+      "loss": 0.2569,
+      "step": 13110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013884907531864365,
+      "loss": 0.7676,
+      "step": 13111
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013883014626208398,
+      "loss": 0.8993,
+      "step": 13112
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001388112173843928,
+      "loss": 2.2112,
+      "step": 13113
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013879228868587308,
+      "loss": 0.0033,
+      "step": 13114
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001387733601668281,
+      "loss": 0.1786,
+      "step": 13115
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013875443182756095,
+      "loss": 0.6173,
+      "step": 13116
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013873550366837466,
+      "loss": 0.2594,
+      "step": 13117
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001387165756895724,
+      "loss": 2.827,
+      "step": 13118
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013869764789145716,
+      "loss": 0.3425,
+      "step": 13119
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001386787202743322,
+      "loss": 0.3857,
+      "step": 13120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013865979283850054,
+      "loss": 0.8947,
+      "step": 13121
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013864086558426522,
+      "loss": 0.3668,
+      "step": 13122
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013862193851192943,
+      "loss": 0.791,
+      "step": 13123
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013860301162179625,
+      "loss": 0.2974,
+      "step": 13124
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013858408491416866,
+      "loss": 0.0012,
+      "step": 13125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001385651583893499,
+      "loss": 0.5162,
+      "step": 13126
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013854623204764293,
+      "loss": 0.6872,
+      "step": 13127
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013852730588935082,
+      "loss": 0.1727,
+      "step": 13128
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013850837991477677,
+      "loss": 0.6021,
+      "step": 13129
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001384894541242237,
+      "loss": 0.5021,
+      "step": 13130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013847052851799477,
+      "loss": 0.0075,
+      "step": 13131
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013845160309639308,
+      "loss": 0.3533,
+      "step": 13132
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013843267785972155,
+      "loss": 2.7562,
+      "step": 13133
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013841375280828337,
+      "loss": 0.5054,
+      "step": 13134
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013839482794238155,
+      "loss": 0.3037,
+      "step": 13135
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001383759032623191,
+      "loss": 0.3295,
+      "step": 13136
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013835697876839916,
+      "loss": 0.8555,
+      "step": 13137
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013833805446092473,
+      "loss": 0.2171,
+      "step": 13138
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013831913034019881,
+      "loss": 0.0247,
+      "step": 13139
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013830020640652455,
+      "loss": 0.707,
+      "step": 13140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001382812826602049,
+      "loss": 0.4579,
+      "step": 13141
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013826235910154285,
+      "loss": 0.2789,
+      "step": 13142
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013824343573084162,
+      "loss": 0.3967,
+      "step": 13143
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013822451254840405,
+      "loss": 0.2322,
+      "step": 13144
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013820558955453325,
+      "loss": 0.5919,
+      "step": 13145
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013818666674953226,
+      "loss": 0.2325,
+      "step": 13146
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013816774413370406,
+      "loss": 0.0237,
+      "step": 13147
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001381488217073517,
+      "loss": 0.3518,
+      "step": 13148
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013812989947077812,
+      "loss": 0.1941,
+      "step": 13149
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001381109774242864,
+      "loss": 0.2633,
+      "step": 13150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001380920555681796,
+      "loss": 0.0142,
+      "step": 13151
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001380731339027606,
+      "loss": 0.0039,
+      "step": 13152
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013805421242833244,
+      "loss": 0.0184,
+      "step": 13153
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001380352911451982,
+      "loss": 0.3703,
+      "step": 13154
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013801637005366078,
+      "loss": 0.4018,
+      "step": 13155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013799744915402325,
+      "loss": 0.0181,
+      "step": 13156
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001379785284465885,
+      "loss": 0.768,
+      "step": 13157
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013795960793165957,
+      "loss": 0.9165,
+      "step": 13158
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001379406876095395,
+      "loss": 0.7705,
+      "step": 13159
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013792176748053116,
+      "loss": 0.1909,
+      "step": 13160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013790284754493763,
+      "loss": 0.4523,
+      "step": 13161
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013788392780306186,
+      "loss": 0.3232,
+      "step": 13162
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013786500825520675,
+      "loss": 0.0015,
+      "step": 13163
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013784608890167534,
+      "loss": 0.2914,
+      "step": 13164
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013782716974277063,
+      "loss": 0.1753,
+      "step": 13165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013780825077879544,
+      "loss": 0.3186,
+      "step": 13166
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013778933201005292,
+      "loss": 0.4088,
+      "step": 13167
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001377704134368458,
+      "loss": 0.8276,
+      "step": 13168
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013775149505947722,
+      "loss": 1.2245,
+      "step": 13169
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001377325768782501,
+      "loss": 1.9639,
+      "step": 13170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013771365889346727,
+      "loss": 0.5703,
+      "step": 13171
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013769474110543177,
+      "loss": 2.2867,
+      "step": 13172
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013767582351444658,
+      "loss": 1.7749,
+      "step": 13173
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013765690612081453,
+      "loss": 0.0024,
+      "step": 13174
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013763798892483864,
+      "loss": 0.2941,
+      "step": 13175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013761907192682177,
+      "loss": 0.3521,
+      "step": 13176
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013760015512706688,
+      "loss": 0.0721,
+      "step": 13177
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013758123852587693,
+      "loss": 0.1875,
+      "step": 13178
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013756232212355478,
+      "loss": 0.0003,
+      "step": 13179
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013754340592040336,
+      "loss": 0.1922,
+      "step": 13180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013752448991672566,
+      "loss": 0.0058,
+      "step": 13181
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013750557411282448,
+      "loss": 0.0004,
+      "step": 13182
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013748665850900276,
+      "loss": 0.001,
+      "step": 13183
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013746774310556352,
+      "loss": 0.0007,
+      "step": 13184
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001374488279028095,
+      "loss": 0.1788,
+      "step": 13185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001374299129010437,
+      "loss": 0.3537,
+      "step": 13186
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001374109981005689,
+      "loss": 1.5125,
+      "step": 13187
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001373920835016881,
+      "loss": 0.6365,
+      "step": 13188
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001373731691047042,
+      "loss": 0.7489,
+      "step": 13189
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013735425490992004,
+      "loss": 0.0011,
+      "step": 13190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013733534091763848,
+      "loss": 0.379,
+      "step": 13191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013731642712816247,
+      "loss": 0.3482,
+      "step": 13192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013729751354179482,
+      "loss": 4.4492,
+      "step": 13193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013727860015883838,
+      "loss": 0.0038,
+      "step": 13194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013725968697959616,
+      "loss": 0.2826,
+      "step": 13195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013724077400437086,
+      "loss": 0.3071,
+      "step": 13196
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001372218612334655,
+      "loss": 4.1271,
+      "step": 13197
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013720294866718274,
+      "loss": 0.0034,
+      "step": 13198
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013718403630582558,
+      "loss": 2.7232,
+      "step": 13199
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013716512414969692,
+      "loss": 0.2382,
+      "step": 13200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013714621219909948,
+      "loss": 0.2966,
+      "step": 13201
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013712730045433612,
+      "loss": 0.8974,
+      "step": 13202
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013710838891570982,
+      "loss": 0.0014,
+      "step": 13203
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013708947758352325,
+      "loss": 0.8679,
+      "step": 13204
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013707056645807938,
+      "loss": 1.459,
+      "step": 13205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013705165553968093,
+      "loss": 0.8385,
+      "step": 13206
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013703274482863078,
+      "loss": 2.4539,
+      "step": 13207
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013701383432523182,
+      "loss": 0.8796,
+      "step": 13208
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013699492402978675,
+      "loss": 1.4951,
+      "step": 13209
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013697601394259848,
+      "loss": 0.7011,
+      "step": 13210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013695710406396984,
+      "loss": 2.3698,
+      "step": 13211
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001369381943942035,
+      "loss": 0.3453,
+      "step": 13212
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013691928493360245,
+      "loss": 0.7915,
+      "step": 13213
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013690037568246942,
+      "loss": 0.3211,
+      "step": 13214
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013688146664110718,
+      "loss": 0.5585,
+      "step": 13215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013686255780981862,
+      "loss": 0.1905,
+      "step": 13216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013684364918890645,
+      "loss": 0.3186,
+      "step": 13217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013682474077867345,
+      "loss": 0.021,
+      "step": 13218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013680583257942255,
+      "loss": 0.7535,
+      "step": 13219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013678692459145634,
+      "loss": 0.4061,
+      "step": 13220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013676801681507773,
+      "loss": 0.2378,
+      "step": 13221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013674910925058953,
+      "loss": 0.5773,
+      "step": 13222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001367302018982944,
+      "loss": 0.2946,
+      "step": 13223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001367112947584952,
+      "loss": 0.8674,
+      "step": 13224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013669238783149465,
+      "loss": 0.733,
+      "step": 13225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013667348111759552,
+      "loss": 0.0028,
+      "step": 13226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013665457461710064,
+      "loss": 0.0014,
+      "step": 13227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001366356683303127,
+      "loss": 0.4329,
+      "step": 13228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013661676225753443,
+      "loss": 0.8291,
+      "step": 13229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013659785639906868,
+      "loss": 0.7373,
+      "step": 13230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013657895075521814,
+      "loss": 0.5275,
+      "step": 13231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013656004532628554,
+      "loss": 0.0109,
+      "step": 13232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013654114011257368,
+      "loss": 0.5119,
+      "step": 13233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013652223511438521,
+      "loss": 0.0011,
+      "step": 13234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013650333033202297,
+      "loss": 0.3889,
+      "step": 13235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013648442576578963,
+      "loss": 0.3442,
+      "step": 13236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013646552141598786,
+      "loss": 0.0018,
+      "step": 13237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013644661728292055,
+      "loss": 2.3692,
+      "step": 13238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013642771336689027,
+      "loss": 0.6372,
+      "step": 13239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013640880966819975,
+      "loss": 0.8458,
+      "step": 13240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013638990618715184,
+      "loss": 0.4195,
+      "step": 13241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013637100292404912,
+      "loss": 0.5817,
+      "step": 13242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013635209987919428,
+      "loss": 0.3738,
+      "step": 13243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013633319705289016,
+      "loss": 0.4268,
+      "step": 13244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013631429444543934,
+      "loss": 2.1872,
+      "step": 13245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013629539205714457,
+      "loss": 0.4199,
+      "step": 13246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001362764898883085,
+      "loss": 0.2232,
+      "step": 13247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013625758793923384,
+      "loss": 0.1783,
+      "step": 13248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013623868621022333,
+      "loss": 0.4009,
+      "step": 13249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013621978470157957,
+      "loss": 0.2003,
+      "step": 13250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013620088341360527,
+      "loss": 0.0068,
+      "step": 13251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013618198234660318,
+      "loss": 5.9683,
+      "step": 13252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013616308150087584,
+      "loss": 0.302,
+      "step": 13253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013614418087672605,
+      "loss": 0.2358,
+      "step": 13254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001361252804744563,
+      "loss": 0.3451,
+      "step": 13255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013610638029436944,
+      "loss": 0.2996,
+      "step": 13256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013608748033676807,
+      "loss": 0.236,
+      "step": 13257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013606858060195474,
+      "loss": 0.3907,
+      "step": 13258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001360496810902322,
+      "loss": 0.3357,
+      "step": 13259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013603078180190315,
+      "loss": 0.0083,
+      "step": 13260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001360118827372701,
+      "loss": 0.1295,
+      "step": 13261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013599298389663575,
+      "loss": 0.6837,
+      "step": 13262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001359740852803028,
+      "loss": 0.3352,
+      "step": 13263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013595518688857376,
+      "loss": 0.1114,
+      "step": 13264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001359362887217514,
+      "loss": 0.0212,
+      "step": 13265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013591739078013822,
+      "loss": 0.1964,
+      "step": 13266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013589849306403686,
+      "loss": 0.271,
+      "step": 13267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013587959557375006,
+      "loss": 0.5292,
+      "step": 13268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013586069830958026,
+      "loss": 0.4738,
+      "step": 13269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013584180127183018,
+      "loss": 0.268,
+      "step": 13270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013582290446080244,
+      "loss": 0.5238,
+      "step": 13271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013580400787679953,
+      "loss": 0.8775,
+      "step": 13272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013578511152012422,
+      "loss": 0.8669,
+      "step": 13273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000135766215391079,
+      "loss": 1.1764,
+      "step": 13274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013574731948996638,
+      "loss": 0.0093,
+      "step": 13275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013572842381708917,
+      "loss": 1.4238,
+      "step": 13276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013570952837274977,
+      "loss": 1.1877,
+      "step": 13277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001356906331572508,
+      "loss": 0.1969,
+      "step": 13278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013567173817089496,
+      "loss": 0.1731,
+      "step": 13279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013565284341398467,
+      "loss": 1.7368,
+      "step": 13280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013563394888682252,
+      "loss": 0.6938,
+      "step": 13281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001356150545897112,
+      "loss": 0.1234,
+      "step": 13282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013559616052295315,
+      "loss": 0.2585,
+      "step": 13283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013557726668685102,
+      "loss": 0.5635,
+      "step": 13284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013555837308170727,
+      "loss": 0.267,
+      "step": 13285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013553947970782447,
+      "loss": 0.5767,
+      "step": 13286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013552058656550526,
+      "loss": 0.7441,
+      "step": 13287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001355016936550521,
+      "loss": 0.1611,
+      "step": 13288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013548280097676753,
+      "loss": 1.418,
+      "step": 13289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001354639085309542,
+      "loss": 0.2312,
+      "step": 13290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013544501631791448,
+      "loss": 0.516,
+      "step": 13291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013542612433795097,
+      "loss": 0.1129,
+      "step": 13292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013540723259136626,
+      "loss": 0.0537,
+      "step": 13293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013538834107846282,
+      "loss": 1.0793,
+      "step": 13294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013536944979954316,
+      "loss": 0.196,
+      "step": 13295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013535055875490976,
+      "loss": 0.7013,
+      "step": 13296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013533166794486523,
+      "loss": 0.5852,
+      "step": 13297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.000135312777369712,
+      "loss": 0.1337,
+      "step": 13298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013529388702975258,
+      "loss": 0.0177,
+      "step": 13299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001352749969252895,
+      "loss": 0.2487,
+      "step": 13300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013525610705662528,
+      "loss": 0.0227,
+      "step": 13301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013523721742406234,
+      "loss": 0.4268,
+      "step": 13302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013521832802790328,
+      "loss": 0.259,
+      "step": 13303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013519943886845042,
+      "loss": 0.5936,
+      "step": 13304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013518054994600637,
+      "loss": 0.0015,
+      "step": 13305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013516166126087357,
+      "loss": 0.0019,
+      "step": 13306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013514277281335448,
+      "loss": 0.4729,
+      "step": 13307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001351238846037516,
+      "loss": 0.6964,
+      "step": 13308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013510499663236739,
+      "loss": 0.5861,
+      "step": 13309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013508610889950427,
+      "loss": 0.6181,
+      "step": 13310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013506722140546474,
+      "loss": 0.4954,
+      "step": 13311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001350483341505513,
+      "loss": 0.002,
+      "step": 13312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013502944713506627,
+      "loss": 0.1863,
+      "step": 13313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013501056035931228,
+      "loss": 1.2881,
+      "step": 13314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013499167382359162,
+      "loss": 0.2598,
+      "step": 13315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013497278752820673,
+      "loss": 0.4025,
+      "step": 13316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013495390147346016,
+      "loss": 0.8798,
+      "step": 13317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013493501565965425,
+      "loss": 0.0909,
+      "step": 13318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013491613008709142,
+      "loss": 0.5593,
+      "step": 13319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013489724475607425,
+      "loss": 0.3395,
+      "step": 13320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001348783596669049,
+      "loss": 1.1317,
+      "step": 13321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013485947481988602,
+      "loss": 0.3658,
+      "step": 13322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001348405902153199,
+      "loss": 0.4775,
+      "step": 13323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013482170585350892,
+      "loss": 0.2531,
+      "step": 13324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013480282173475563,
+      "loss": 0.6185,
+      "step": 13325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001347839378593623,
+      "loss": 0.4149,
+      "step": 13326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013476505422763133,
+      "loss": 1.6833,
+      "step": 13327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013474617083986522,
+      "loss": 0.632,
+      "step": 13328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013472728769636625,
+      "loss": 0.3852,
+      "step": 13329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013470840479743682,
+      "loss": 0.2648,
+      "step": 13330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001346895221433794,
+      "loss": 0.0035,
+      "step": 13331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00013467063973449632,
+      "loss": 0.184,
+      "step": 13332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001346517575710899,
+      "loss": 0.7915,
+      "step": 13333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001346328756534625,
+      "loss": 0.8225,
+      "step": 13334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001346139939819166,
+      "loss": 0.8727,
+      "step": 13335
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001345951125567545,
+      "loss": 0.7876,
+      "step": 13336
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013457623137827854,
+      "loss": 0.5933,
+      "step": 13337
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013455735044679105,
+      "loss": 0.7494,
+      "step": 13338
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013453846976259446,
+      "loss": 0.1429,
+      "step": 13339
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013451958932599107,
+      "loss": 0.2902,
+      "step": 13340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013450070913728315,
+      "loss": 0.456,
+      "step": 13341
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013448182919677322,
+      "loss": 0.1361,
+      "step": 13342
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013446294950476346,
+      "loss": 0.9235,
+      "step": 13343
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013444407006155628,
+      "loss": 0.1576,
+      "step": 13344
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013442519086745387,
+      "loss": 0.0026,
+      "step": 13345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013440631192275873,
+      "loss": 0.6288,
+      "step": 13346
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013438743322777313,
+      "loss": 0.7015,
+      "step": 13347
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001343685547827993,
+      "loss": 0.4348,
+      "step": 13348
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001343496765881396,
+      "loss": 0.7162,
+      "step": 13349
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013433079864409638,
+      "loss": 0.7645,
+      "step": 13350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013431192095097186,
+      "loss": 0.1121,
+      "step": 13351
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013429304350906848,
+      "loss": 0.5719,
+      "step": 13352
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013427416631868832,
+      "loss": 0.7033,
+      "step": 13353
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013425528938013382,
+      "loss": 0.6616,
+      "step": 13354
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013423641269370729,
+      "loss": 0.2779,
+      "step": 13355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013421753625971085,
+      "loss": 0.8945,
+      "step": 13356
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013419866007844693,
+      "loss": 0.0014,
+      "step": 13357
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001341797841502178,
+      "loss": 0.447,
+      "step": 13358
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001341609084753256,
+      "loss": 0.3204,
+      "step": 13359
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013414203305407273,
+      "loss": 0.4742,
+      "step": 13360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013412315788676143,
+      "loss": 0.6522,
+      "step": 13361
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013410428297369384,
+      "loss": 0.3384,
+      "step": 13362
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013408540831517238,
+      "loss": 0.6718,
+      "step": 13363
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001340665339114992,
+      "loss": 0.0061,
+      "step": 13364
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013404765976297652,
+      "loss": 0.9109,
+      "step": 13365
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013402878586990673,
+      "loss": 0.3085,
+      "step": 13366
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013400991223259188,
+      "loss": 2.1356,
+      "step": 13367
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013399103885133428,
+      "loss": 0.5928,
+      "step": 13368
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013397216572643623,
+      "loss": 0.2638,
+      "step": 13369
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013395329285819985,
+      "loss": 0.8122,
+      "step": 13370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013393442024692745,
+      "loss": 0.0012,
+      "step": 13371
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013391554789292116,
+      "loss": 2.1515,
+      "step": 13372
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001338966757964832,
+      "loss": 0.2756,
+      "step": 13373
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013387780395791586,
+      "loss": 0.6059,
+      "step": 13374
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013385893237752126,
+      "loss": 0.4522,
+      "step": 13375
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013384006105560163,
+      "loss": 0.6152,
+      "step": 13376
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013382118999245922,
+      "loss": 0.602,
+      "step": 13377
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013380231918839612,
+      "loss": 0.5549,
+      "step": 13378
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013378344864371458,
+      "loss": 0.0024,
+      "step": 13379
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001337645783587168,
+      "loss": 0.5636,
+      "step": 13380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013374570833370492,
+      "loss": 0.7794,
+      "step": 13381
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013372683856898115,
+      "loss": 0.9367,
+      "step": 13382
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013370796906484757,
+      "loss": 0.7548,
+      "step": 13383
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013368909982160645,
+      "loss": 0.4694,
+      "step": 13384
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013367023083955996,
+      "loss": 0.8047,
+      "step": 13385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013365136211901014,
+      "loss": 0.2827,
+      "step": 13386
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013363249366025925,
+      "loss": 0.3181,
+      "step": 13387
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013361362546360944,
+      "loss": 0.5606,
+      "step": 13388
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001335947575293628,
+      "loss": 0.0015,
+      "step": 13389
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013357588985782147,
+      "loss": 0.0214,
+      "step": 13390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013355702244928765,
+      "loss": 0.9961,
+      "step": 13391
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013353815530406344,
+      "loss": 0.6356,
+      "step": 13392
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013351928842245098,
+      "loss": 0.0109,
+      "step": 13393
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013350042180475233,
+      "loss": 0.6085,
+      "step": 13394
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001334815554512697,
+      "loss": 1.0965,
+      "step": 13395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013346268936230519,
+      "loss": 0.7241,
+      "step": 13396
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001334438235381608,
+      "loss": 0.8129,
+      "step": 13397
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001334249579791388,
+      "loss": 0.4573,
+      "step": 13398
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013340609268554124,
+      "loss": 0.003,
+      "step": 13399
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013338722765767014,
+      "loss": 4.0601,
+      "step": 13400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001333683628958277,
+      "loss": 0.2701,
+      "step": 13401
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013334949840031597,
+      "loss": 0.8523,
+      "step": 13402
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000133330634171437,
+      "loss": 0.3021,
+      "step": 13403
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000133311770209493,
+      "loss": 0.0697,
+      "step": 13404
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013329290651478584,
+      "loss": 0.8714,
+      "step": 13405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013327404308761774,
+      "loss": 0.6013,
+      "step": 13406
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013325517992829078,
+      "loss": 0.0026,
+      "step": 13407
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013323631703710692,
+      "loss": 0.0843,
+      "step": 13408
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001332174544143683,
+      "loss": 0.2786,
+      "step": 13409
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013319859206037702,
+      "loss": 0.4804,
+      "step": 13410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013317972997543498,
+      "loss": 0.2881,
+      "step": 13411
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013316086815984442,
+      "loss": 0.3723,
+      "step": 13412
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013314200661390722,
+      "loss": 0.0014,
+      "step": 13413
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013312314533792547,
+      "loss": 0.5807,
+      "step": 13414
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013310428433220128,
+      "loss": 0.9997,
+      "step": 13415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001330854235970366,
+      "loss": 0.7576,
+      "step": 13416
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001330665631327334,
+      "loss": 0.0031,
+      "step": 13417
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013304770293959386,
+      "loss": 0.3123,
+      "step": 13418
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013302884301791992,
+      "loss": 0.2929,
+      "step": 13419
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013300998336801358,
+      "loss": 0.0033,
+      "step": 13420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013299112399017685,
+      "loss": 0.6721,
+      "step": 13421
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013297226488471169,
+      "loss": 0.2805,
+      "step": 13422
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013295340605192025,
+      "loss": 0.0009,
+      "step": 13423
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013293454749210436,
+      "loss": 0.4536,
+      "step": 13424
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013291568920556607,
+      "loss": 0.3571,
+      "step": 13425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013289683119260741,
+      "loss": 0.0022,
+      "step": 13426
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013287797345353034,
+      "loss": 1.5237,
+      "step": 13427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013285911598863676,
+      "loss": 0.4946,
+      "step": 13428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013284025879822881,
+      "loss": 0.2522,
+      "step": 13429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001328214018826083,
+      "loss": 5.0442,
+      "step": 13430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013280254524207732,
+      "loss": 0.0014,
+      "step": 13431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001327836888769377,
+      "loss": 0.3947,
+      "step": 13432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013276483278749148,
+      "loss": 0.4951,
+      "step": 13433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013274597697404063,
+      "loss": 0.0073,
+      "step": 13434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000132727121436887,
+      "loss": 0.5868,
+      "step": 13435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013270826617633265,
+      "loss": 0.0043,
+      "step": 13436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013268941119267948,
+      "loss": 0.4267,
+      "step": 13437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013267055648622937,
+      "loss": 0.3415,
+      "step": 13438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013265170205728426,
+      "loss": 0.0062,
+      "step": 13439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001326328479061462,
+      "loss": 0.0021,
+      "step": 13440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013261399403311695,
+      "loss": 0.0011,
+      "step": 13441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013259514043849853,
+      "loss": 0.6509,
+      "step": 13442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013257628712259276,
+      "loss": 0.7587,
+      "step": 13443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013255743408570165,
+      "loss": 0.7944,
+      "step": 13444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013253858132812704,
+      "loss": 0.6818,
+      "step": 13445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013251972885017083,
+      "loss": 0.087,
+      "step": 13446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013250087665213494,
+      "loss": 0.9863,
+      "step": 13447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013248202473432126,
+      "loss": 0.0015,
+      "step": 13448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013246317309703164,
+      "loss": 0.0662,
+      "step": 13449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013244432174056803,
+      "loss": 0.4535,
+      "step": 13450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013242547066523222,
+      "loss": 0.4657,
+      "step": 13451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001324066198713261,
+      "loss": 0.4775,
+      "step": 13452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001323877693591516,
+      "loss": 0.245,
+      "step": 13453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001323689191290105,
+      "loss": 1.4091,
+      "step": 13454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013235006918120473,
+      "loss": 0.003,
+      "step": 13455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013233121951603617,
+      "loss": 0.0023,
+      "step": 13456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013231237013380651,
+      "loss": 0.4118,
+      "step": 13457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013229352103481776,
+      "loss": 0.4259,
+      "step": 13458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013227467221937168,
+      "loss": 1.2246,
+      "step": 13459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001322558236877701,
+      "loss": 0.7026,
+      "step": 13460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013223697544031493,
+      "loss": 0.003,
+      "step": 13461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001322181274773079,
+      "loss": 0.0007,
+      "step": 13462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013219927979905082,
+      "loss": 0.0099,
+      "step": 13463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013218043240584564,
+      "loss": 0.3361,
+      "step": 13464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013216158529799407,
+      "loss": 0.5761,
+      "step": 13465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001321427384757979,
+      "loss": 1.8545,
+      "step": 13466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013212389193955904,
+      "loss": 0.0018,
+      "step": 13467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001321050456895792,
+      "loss": 2.209,
+      "step": 13468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001320861997261602,
+      "loss": 0.0976,
+      "step": 13469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013206735404960377,
+      "loss": 0.0045,
+      "step": 13470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001320485086602118,
+      "loss": 0.0031,
+      "step": 13471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013202966355828607,
+      "loss": 1.5102,
+      "step": 13472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013201081874412825,
+      "loss": 0.5353,
+      "step": 13473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013199197421804015,
+      "loss": 0.87,
+      "step": 13474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001319731299803236,
+      "loss": 0.0024,
+      "step": 13475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001319542860312803,
+      "loss": 0.2226,
+      "step": 13476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000131935442371212,
+      "loss": 1.2555,
+      "step": 13477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013191659900042052,
+      "loss": 0.0033,
+      "step": 13478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013189775591920756,
+      "loss": 0.908,
+      "step": 13479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001318789131278749,
+      "loss": 0.0022,
+      "step": 13480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013186007062672417,
+      "loss": 0.5396,
+      "step": 13481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013184122841605723,
+      "loss": 0.1093,
+      "step": 13482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001318223864961758,
+      "loss": 0.5745,
+      "step": 13483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001318035448673815,
+      "loss": 0.2325,
+      "step": 13484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013178470352997616,
+      "loss": 0.3678,
+      "step": 13485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013176586248426147,
+      "loss": 1.0903,
+      "step": 13486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001317470217305391,
+      "loss": 0.9973,
+      "step": 13487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013172818126911076,
+      "loss": 0.3685,
+      "step": 13488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013170934110027824,
+      "loss": 0.1256,
+      "step": 13489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013169050122434313,
+      "loss": 2.1953,
+      "step": 13490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001316716616416072,
+      "loss": 0.0945,
+      "step": 13491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013165282235237205,
+      "loss": 1.651,
+      "step": 13492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013163398335693942,
+      "loss": 3.6296,
+      "step": 13493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013161514465561105,
+      "loss": 0.0039,
+      "step": 13494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013159630624868846,
+      "loss": 0.1545,
+      "step": 13495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013157746813647342,
+      "loss": 0.0862,
+      "step": 13496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013155863031926766,
+      "loss": 0.5668,
+      "step": 13497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013153979279737266,
+      "loss": 0.661,
+      "step": 13498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013152095557109023,
+      "loss": 0.2439,
+      "step": 13499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013150211864072195,
+      "loss": 0.3905,
+      "step": 13500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013148328200656944,
+      "loss": 0.4029,
+      "step": 13501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013146444566893446,
+      "loss": 0.8412,
+      "step": 13502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001314456096281185,
+      "loss": 0.3137,
+      "step": 13503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013142677388442326,
+      "loss": 0.0018,
+      "step": 13504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013140793843815043,
+      "loss": 0.0009,
+      "step": 13505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013138910328960147,
+      "loss": 0.3801,
+      "step": 13506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013137026843907813,
+      "loss": 0.4546,
+      "step": 13507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000131351433886882,
+      "loss": 1.4769,
+      "step": 13508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013133259963331464,
+      "loss": 2.238,
+      "step": 13509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013131376567867772,
+      "loss": 0.0034,
+      "step": 13510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013129493202327275,
+      "loss": 0.3471,
+      "step": 13511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013127609866740137,
+      "loss": 0.0015,
+      "step": 13512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013125726561136523,
+      "loss": 1.2647,
+      "step": 13513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001312384328554658,
+      "loss": 0.3314,
+      "step": 13514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001312196004000047,
+      "loss": 0.3941,
+      "step": 13515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013120076824528354,
+      "loss": 0.9097,
+      "step": 13516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013118193639160387,
+      "loss": 0.9843,
+      "step": 13517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013116310483926726,
+      "loss": 0.4838,
+      "step": 13518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001311442735885752,
+      "loss": 0.1907,
+      "step": 13519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013112544263982933,
+      "loss": 0.2949,
+      "step": 13520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001311066119933312,
+      "loss": 0.3467,
+      "step": 13521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013108778164938226,
+      "loss": 0.8274,
+      "step": 13522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013106895160828417,
+      "loss": 0.6187,
+      "step": 13523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001310501218703384,
+      "loss": 0.4229,
+      "step": 13524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001310312924358465,
+      "loss": 0.0014,
+      "step": 13525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001310124633051099,
+      "loss": 0.6041,
+      "step": 13526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001309936344784303,
+      "loss": 0.4717,
+      "step": 13527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013097480595610907,
+      "loss": 0.6512,
+      "step": 13528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013095597773844783,
+      "loss": 0.1017,
+      "step": 13529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013093714982574794,
+      "loss": 0.7469,
+      "step": 13530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013091832221831103,
+      "loss": 0.5825,
+      "step": 13531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013089949491643856,
+      "loss": 0.3905,
+      "step": 13532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013088066792043198,
+      "loss": 0.9673,
+      "step": 13533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013086184123059283,
+      "loss": 0.22,
+      "step": 13534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013084301484722258,
+      "loss": 0.0135,
+      "step": 13535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013082418877062267,
+      "loss": 0.2952,
+      "step": 13536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001308053630010946,
+      "loss": 0.0014,
+      "step": 13537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013078653753893988,
+      "loss": 0.8406,
+      "step": 13538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013076771238445986,
+      "loss": 0.61,
+      "step": 13539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013074888753795616,
+      "loss": 0.0755,
+      "step": 13540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013073006299973004,
+      "loss": 0.4274,
+      "step": 13541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013071123877008307,
+      "loss": 0.5837,
+      "step": 13542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001306924148493167,
+      "loss": 0.7782,
+      "step": 13543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013067359123773226,
+      "loss": 0.635,
+      "step": 13544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013065476793563128,
+      "loss": 0.4055,
+      "step": 13545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013063594494331518,
+      "loss": 0.3076,
+      "step": 13546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013061712226108532,
+      "loss": 0.3737,
+      "step": 13547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001305982998892432,
+      "loss": 0.9793,
+      "step": 13548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001305794778280902,
+      "loss": 0.0012,
+      "step": 13549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013056065607792763,
+      "loss": 0.1248,
+      "step": 13550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013054183463905708,
+      "loss": 0.2627,
+      "step": 13551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001305230135117798,
+      "loss": 0.3803,
+      "step": 13552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013050419269639718,
+      "loss": 1.0617,
+      "step": 13553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013048537219321075,
+      "loss": 0.3111,
+      "step": 13554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013046655200252174,
+      "loss": 3.7688,
+      "step": 13555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013044773212463157,
+      "loss": 0.1595,
+      "step": 13556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013042891255984172,
+      "loss": 0.5625,
+      "step": 13557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013041009330845335,
+      "loss": 0.3325,
+      "step": 13558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013039127437076801,
+      "loss": 0.0855,
+      "step": 13559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013037245574708695,
+      "loss": 0.0959,
+      "step": 13560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013035363743771152,
+      "loss": 0.5005,
+      "step": 13561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013033481944294318,
+      "loss": 0.0041,
+      "step": 13562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013031600176308315,
+      "loss": 0.1786,
+      "step": 13563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013029718439843275,
+      "loss": 0.0023,
+      "step": 13564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013027836734929348,
+      "loss": 0.663,
+      "step": 13565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001302595506159665,
+      "loss": 0.2504,
+      "step": 13566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013024073419875325,
+      "loss": 0.6617,
+      "step": 13567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001302219180979549,
+      "loss": 0.4735,
+      "step": 13568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001302031023138729,
+      "loss": 0.2443,
+      "step": 13569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013018428684680853,
+      "loss": 0.0028,
+      "step": 13570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013016547169706299,
+      "loss": 0.2385,
+      "step": 13571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001301466568649377,
+      "loss": 0.7098,
+      "step": 13572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013012784235073393,
+      "loss": 0.4092,
+      "step": 13573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001301090281547529,
+      "loss": 0.4283,
+      "step": 13574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001300902142772959,
+      "loss": 0.0026,
+      "step": 13575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013007140071866433,
+      "loss": 0.3549,
+      "step": 13576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001300525874791593,
+      "loss": 0.3063,
+      "step": 13577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001300337745590822,
+      "loss": 0.7117,
+      "step": 13578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00013001496195873418,
+      "loss": 0.4865,
+      "step": 13579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012999614967841656,
+      "loss": 0.3071,
+      "step": 13580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012997733771843062,
+      "loss": 0.2175,
+      "step": 13581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001299585260790775,
+      "loss": 0.3842,
+      "step": 13582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00012993971476065854,
+      "loss": 0.1875,
+      "step": 13583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.000129920903763475,
+      "loss": 0.1928,
+      "step": 13584
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012990209308782794,
+      "loss": 0.0167,
+      "step": 13585
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012988328273401878,
+      "loss": 0.504,
+      "step": 13586
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012986447270234866,
+      "loss": 1.2491,
+      "step": 13587
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012984566299311873,
+      "loss": 0.2473,
+      "step": 13588
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012982685360663033,
+      "loss": 0.4366,
+      "step": 13589
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012980804454318452,
+      "loss": 0.2513,
+      "step": 13590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001297892358030826,
+      "loss": 0.8976,
+      "step": 13591
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001297704273866258,
+      "loss": 0.096,
+      "step": 13592
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012975161929411515,
+      "loss": 0.023,
+      "step": 13593
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012973281152585194,
+      "loss": 0.8763,
+      "step": 13594
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012971400408213742,
+      "loss": 0.5903,
+      "step": 13595
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012969519696327258,
+      "loss": 0.159,
+      "step": 13596
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012967639016955875,
+      "loss": 0.0006,
+      "step": 13597
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012965758370129702,
+      "loss": 0.0587,
+      "step": 13598
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012963877755878852,
+      "loss": 0.001,
+      "step": 13599
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001296199717423345,
+      "loss": 0.384,
+      "step": 13600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000129601166252236,
+      "loss": 0.2194,
+      "step": 13601
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001295823610887942,
+      "loss": 2.1086,
+      "step": 13602
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012956355625231032,
+      "loss": 0.9628,
+      "step": 13603
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012954475174308536,
+      "loss": 0.154,
+      "step": 13604
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012952594756142046,
+      "loss": 0.5042,
+      "step": 13605
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012950714370761686,
+      "loss": 0.9864,
+      "step": 13606
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012948834018197556,
+      "loss": 0.8651,
+      "step": 13607
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012946953698479775,
+      "loss": 0.1692,
+      "step": 13608
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012945073411638444,
+      "loss": 0.6577,
+      "step": 13609
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012943193157703679,
+      "loss": 0.1613,
+      "step": 13610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001294131293670559,
+      "loss": 0.0004,
+      "step": 13611
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012939432748674284,
+      "loss": 0.3344,
+      "step": 13612
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012937552593639867,
+      "loss": 0.3139,
+      "step": 13613
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012935672471632456,
+      "loss": 0.3092,
+      "step": 13614
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012933792382682153,
+      "loss": 0.7907,
+      "step": 13615
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001293191232681906,
+      "loss": 0.0022,
+      "step": 13616
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012930032304073286,
+      "loss": 0.6371,
+      "step": 13617
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001292815231447494,
+      "loss": 1.1549,
+      "step": 13618
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001292627235805413,
+      "loss": 0.307,
+      "step": 13619
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012924392434840947,
+      "loss": 0.8009,
+      "step": 13620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012922512544865508,
+      "loss": 0.5072,
+      "step": 13621
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012920632688157916,
+      "loss": 0.5265,
+      "step": 13622
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012918752864748265,
+      "loss": 0.3659,
+      "step": 13623
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012916873074666664,
+      "loss": 0.8808,
+      "step": 13624
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001291499331794322,
+      "loss": 1.0463,
+      "step": 13625
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012913113594608026,
+      "loss": 0.4828,
+      "step": 13626
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012911233904691188,
+      "loss": 0.0013,
+      "step": 13627
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000129093542482228,
+      "loss": 0.3086,
+      "step": 13628
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012907474625232967,
+      "loss": 0.0022,
+      "step": 13629
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001290559503575179,
+      "loss": 0.2436,
+      "step": 13630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012903715479809362,
+      "loss": 0.7484,
+      "step": 13631
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012901835957435784,
+      "loss": 0.481,
+      "step": 13632
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001289995646866116,
+      "loss": 0.0019,
+      "step": 13633
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012898077013515574,
+      "loss": 0.2719,
+      "step": 13634
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012896197592029136,
+      "loss": 0.0204,
+      "step": 13635
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012894318204231936,
+      "loss": 0.1763,
+      "step": 13636
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001289243885015406,
+      "loss": 0.0006,
+      "step": 13637
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012890559529825626,
+      "loss": 2.0502,
+      "step": 13638
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012888680243276708,
+      "loss": 0.6128,
+      "step": 13639
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012886800990537405,
+      "loss": 0.3538,
+      "step": 13640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012884921771637824,
+      "loss": 0.0647,
+      "step": 13641
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012883042586608035,
+      "loss": 0.7301,
+      "step": 13642
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012881163435478146,
+      "loss": 0.5972,
+      "step": 13643
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001287928431827825,
+      "loss": 0.4602,
+      "step": 13644
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001287740523503842,
+      "loss": 0.6801,
+      "step": 13645
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012875526185788773,
+      "loss": 0.3561,
+      "step": 13646
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012873647170559378,
+      "loss": 0.2531,
+      "step": 13647
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001287176818938033,
+      "loss": 0.2148,
+      "step": 13648
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001286988924228173,
+      "loss": 0.1906,
+      "step": 13649
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001286801032929365,
+      "loss": 0.2958,
+      "step": 13650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012866131450446186,
+      "loss": 0.1426,
+      "step": 13651
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012864252605769425,
+      "loss": 0.2436,
+      "step": 13652
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012862373795293454,
+      "loss": 0.1677,
+      "step": 13653
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012860495019048352,
+      "loss": 0.0021,
+      "step": 13654
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001285861627706422,
+      "loss": 0.3862,
+      "step": 13655
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012856737569371133,
+      "loss": 0.3381,
+      "step": 13656
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001285485889599918,
+      "loss": 0.0016,
+      "step": 13657
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012852980256978437,
+      "loss": 0.4768,
+      "step": 13658
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001285110165233899,
+      "loss": 0.0406,
+      "step": 13659
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012849223082110934,
+      "loss": 0.6094,
+      "step": 13660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012847344546324335,
+      "loss": 0.9158,
+      "step": 13661
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012845466045009282,
+      "loss": 1.0772,
+      "step": 13662
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012843587578195863,
+      "loss": 0.4265,
+      "step": 13663
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001284170914591415,
+      "loss": 0.4675,
+      "step": 13664
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012839830748194228,
+      "loss": 0.2729,
+      "step": 13665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012837952385066169,
+      "loss": 0.0013,
+      "step": 13666
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001283607405656006,
+      "loss": 0.3538,
+      "step": 13667
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012834195762705979,
+      "loss": 0.5014,
+      "step": 13668
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012832317503534,
+      "loss": 0.0733,
+      "step": 13669
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012830439279074203,
+      "loss": 0.6628,
+      "step": 13670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001282856108935667,
+      "loss": 1.5972,
+      "step": 13671
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012826682934411466,
+      "loss": 0.776,
+      "step": 13672
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012824804814268675,
+      "loss": 0.5205,
+      "step": 13673
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012822926728958374,
+      "loss": 0.0176,
+      "step": 13674
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012821048678510628,
+      "loss": 0.4248,
+      "step": 13675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012819170662955526,
+      "loss": 0.3018,
+      "step": 13676
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012817292682323125,
+      "loss": 0.9813,
+      "step": 13677
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001281541473664351,
+      "loss": 0.2682,
+      "step": 13678
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001281353682594675,
+      "loss": 0.7014,
+      "step": 13679
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001281165895026291,
+      "loss": 0.3881,
+      "step": 13680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001280978110962207,
+      "loss": 0.5129,
+      "step": 13681
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012807903304054303,
+      "loss": 0.1514,
+      "step": 13682
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012806025533589668,
+      "loss": 0.8463,
+      "step": 13683
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012804147798258249,
+      "loss": 0.3823,
+      "step": 13684
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012802270098090103,
+      "loss": 0.2445,
+      "step": 13685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000128003924331153,
+      "loss": 0.4517,
+      "step": 13686
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012798514803363917,
+      "loss": 0.4176,
+      "step": 13687
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012796637208866015,
+      "loss": 0.343,
+      "step": 13688
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012794759649651654,
+      "loss": 0.7124,
+      "step": 13689
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012792882125750918,
+      "loss": 1.0196,
+      "step": 13690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012791004637193856,
+      "loss": 0.2795,
+      "step": 13691
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012789127184010537,
+      "loss": 0.0034,
+      "step": 13692
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012787249766231038,
+      "loss": 0.7857,
+      "step": 13693
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012785372383885405,
+      "loss": 0.7371,
+      "step": 13694
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012783495037003715,
+      "loss": 0.3136,
+      "step": 13695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012781617725616021,
+      "loss": 0.2781,
+      "step": 13696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012779740449752388,
+      "loss": 0.3914,
+      "step": 13697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012777863209442885,
+      "loss": 0.9404,
+      "step": 13698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012775986004717564,
+      "loss": 0.4318,
+      "step": 13699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012774108835606488,
+      "loss": 0.7011,
+      "step": 13700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012772231702139723,
+      "loss": 0.457,
+      "step": 13701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012770354604347325,
+      "loss": 0.5352,
+      "step": 13702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012768477542259344,
+      "loss": 0.5517,
+      "step": 13703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012766600515905853,
+      "loss": 0.4874,
+      "step": 13704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012764723525316902,
+      "loss": 0.6851,
+      "step": 13705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001276284657052255,
+      "loss": 0.001,
+      "step": 13706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012760969651552846,
+      "loss": 0.743,
+      "step": 13707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012759092768437858,
+      "loss": 0.2342,
+      "step": 13708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001275721592120764,
+      "loss": 0.3821,
+      "step": 13709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012755339109892238,
+      "loss": 0.7508,
+      "step": 13710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012753462334521708,
+      "loss": 0.6255,
+      "step": 13711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012751585595126112,
+      "loss": 0.5999,
+      "step": 13712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012749708891735497,
+      "loss": 3.209,
+      "step": 13713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001274783222437992,
+      "loss": 0.1946,
+      "step": 13714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012745955593089424,
+      "loss": 0.2902,
+      "step": 13715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001274407899789407,
+      "loss": 0.6561,
+      "step": 13716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012742202438823907,
+      "loss": 0.2566,
+      "step": 13717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012740325915908978,
+      "loss": 0.0008,
+      "step": 13718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001273844942917934,
+      "loss": 0.7971,
+      "step": 13719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012736572978665043,
+      "loss": 0.3328,
+      "step": 13720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001273469656439613,
+      "loss": 0.298,
+      "step": 13721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001273282018640265,
+      "loss": 0.4741,
+      "step": 13722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001273094384471466,
+      "loss": 0.3852,
+      "step": 13723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012729067539362193,
+      "loss": 0.0477,
+      "step": 13724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012727191270375308,
+      "loss": 0.3176,
+      "step": 13725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012725315037784033,
+      "loss": 0.4133,
+      "step": 13726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001272343884161843,
+      "loss": 0.1719,
+      "step": 13727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001272156268190854,
+      "loss": 0.0015,
+      "step": 13728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012719686558684395,
+      "loss": 0.2154,
+      "step": 13729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012717810471976054,
+      "loss": 0.1749,
+      "step": 13730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012715934421813556,
+      "loss": 0.1737,
+      "step": 13731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012714058408226932,
+      "loss": 0.0053,
+      "step": 13732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001271218243124624,
+      "loss": 0.1875,
+      "step": 13733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012710306490901509,
+      "loss": 0.0018,
+      "step": 13734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001270843058722278,
+      "loss": 0.2257,
+      "step": 13735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000127065547202401,
+      "loss": 0.1101,
+      "step": 13736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012704678889983504,
+      "loss": 0.7271,
+      "step": 13737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012702803096483029,
+      "loss": 0.7235,
+      "step": 13738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012700927339768717,
+      "loss": 0.2125,
+      "step": 13739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012699051619870602,
+      "loss": 0.7317,
+      "step": 13740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001269717593681872,
+      "loss": 0.4425,
+      "step": 13741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001269530029064312,
+      "loss": 0.2975,
+      "step": 13742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012693424681373814,
+      "loss": 0.5316,
+      "step": 13743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012691549109040856,
+      "loss": 0.3715,
+      "step": 13744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012689673573674273,
+      "loss": 0.2227,
+      "step": 13745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000126877980753041,
+      "loss": 2.4356,
+      "step": 13746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012685922613960373,
+      "loss": 0.5057,
+      "step": 13747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001268404718967312,
+      "loss": 0.445,
+      "step": 13748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012682171802472374,
+      "loss": 0.5923,
+      "step": 13749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001268029645238817,
+      "loss": 0.036,
+      "step": 13750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001267842113945054,
+      "loss": 0.3058,
+      "step": 13751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012676545863689505,
+      "loss": 0.2353,
+      "step": 13752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012674670625135106,
+      "loss": 0.3667,
+      "step": 13753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012672795423817368,
+      "loss": 0.1083,
+      "step": 13754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012670920259766317,
+      "loss": 0.0058,
+      "step": 13755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012669045133011978,
+      "loss": 0.0025,
+      "step": 13756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012667170043584387,
+      "loss": 0.0038,
+      "step": 13757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001266529499151357,
+      "loss": 0.4472,
+      "step": 13758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012663419976829542,
+      "loss": 0.0096,
+      "step": 13759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001266154499956234,
+      "loss": 0.1313,
+      "step": 13760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012659670059741987,
+      "loss": 0.2493,
+      "step": 13761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012657795157398506,
+      "loss": 0.0946,
+      "step": 13762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012655920292561923,
+      "loss": 0.5293,
+      "step": 13763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012654045465262248,
+      "loss": 3.1152,
+      "step": 13764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012652170675529523,
+      "loss": 0.3606,
+      "step": 13765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001265029592339376,
+      "loss": 0.4968,
+      "step": 13766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012648421208884977,
+      "loss": 1.2083,
+      "step": 13767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012646546532033202,
+      "loss": 0.0033,
+      "step": 13768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012644671892868457,
+      "loss": 0.6984,
+      "step": 13769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012642797291420748,
+      "loss": 0.7964,
+      "step": 13770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012640922727720106,
+      "loss": 3.8477,
+      "step": 13771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001263904820179655,
+      "loss": 0.3649,
+      "step": 13772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012637173713680086,
+      "loss": 0.0038,
+      "step": 13773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012635299263400747,
+      "loss": 0.7185,
+      "step": 13774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012633424850988539,
+      "loss": 0.0635,
+      "step": 13775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012631550476473473,
+      "loss": 0.3664,
+      "step": 13776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012629676139885585,
+      "loss": 0.0186,
+      "step": 13777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012627801841254865,
+      "loss": 0.8837,
+      "step": 13778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012625927580611342,
+      "loss": 0.5111,
+      "step": 13779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001262405335798503,
+      "loss": 0.0055,
+      "step": 13780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001262217917340593,
+      "loss": 0.2387,
+      "step": 13781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001262030502690407,
+      "loss": 0.936,
+      "step": 13782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012618430918509446,
+      "loss": 0.6256,
+      "step": 13783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012616556848252076,
+      "loss": 0.0025,
+      "step": 13784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012614682816161982,
+      "loss": 0.3633,
+      "step": 13785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012612808822269153,
+      "loss": 0.4808,
+      "step": 13786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012610934866603607,
+      "loss": 0.7642,
+      "step": 13787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001260906094919536,
+      "loss": 0.4793,
+      "step": 13788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012607187070074412,
+      "loss": 0.1545,
+      "step": 13789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012605313229270768,
+      "loss": 0.3727,
+      "step": 13790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012603439426814444,
+      "loss": 0.0998,
+      "step": 13791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001260156566273544,
+      "loss": 0.007,
+      "step": 13792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012599691937063763,
+      "loss": 0.6046,
+      "step": 13793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012597818249829412,
+      "loss": 2.0361,
+      "step": 13794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012595944601062397,
+      "loss": 0.7671,
+      "step": 13795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001259407099079273,
+      "loss": 0.2269,
+      "step": 13796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012592197419050399,
+      "loss": 0.1853,
+      "step": 13797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001259032388586541,
+      "loss": 1.8424,
+      "step": 13798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012588450391267771,
+      "loss": 0.5109,
+      "step": 13799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001258657693528748,
+      "loss": 0.4636,
+      "step": 13800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012584703517954532,
+      "loss": 0.6185,
+      "step": 13801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001258283013929894,
+      "loss": 0.5518,
+      "step": 13802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012580956799350694,
+      "loss": 0.0035,
+      "step": 13803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012579083498139793,
+      "loss": 0.2173,
+      "step": 13804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012577210235696236,
+      "loss": 0.2131,
+      "step": 13805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001257533701205002,
+      "loss": 0.1823,
+      "step": 13806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012573463827231148,
+      "loss": 0.6915,
+      "step": 13807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012571590681269604,
+      "loss": 0.3518,
+      "step": 13808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012569717574195395,
+      "loss": 0.4868,
+      "step": 13809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001256784450603851,
+      "loss": 0.9761,
+      "step": 13810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012565971476828948,
+      "loss": 0.3885,
+      "step": 13811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012564098486596702,
+      "loss": 0.0025,
+      "step": 13812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012562225535371754,
+      "loss": 0.1181,
+      "step": 13813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001256035262318411,
+      "loss": 0.0014,
+      "step": 13814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001255847975006376,
+      "loss": 1.9092,
+      "step": 13815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012556606916040686,
+      "loss": 2.471,
+      "step": 13816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001255473412114489,
+      "loss": 2.7149,
+      "step": 13817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001255286136540636,
+      "loss": 0.4921,
+      "step": 13818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012550988648855075,
+      "loss": 1.3742,
+      "step": 13819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012549115971521033,
+      "loss": 0.0077,
+      "step": 13820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001254724333343423,
+      "loss": 0.2059,
+      "step": 13821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001254537073462463,
+      "loss": 0.5719,
+      "step": 13822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012543498175122245,
+      "loss": 0.7555,
+      "step": 13823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012541625654957045,
+      "loss": 1.0396,
+      "step": 13824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012539753174159018,
+      "loss": 0.6414,
+      "step": 13825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012537880732758157,
+      "loss": 0.2746,
+      "step": 13826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0001253600833078444,
+      "loss": 0.3966,
+      "step": 13827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012534135968267848,
+      "loss": 0.406,
+      "step": 13828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012532263645238375,
+      "loss": 0.1454,
+      "step": 13829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012530391361725987,
+      "loss": 0.5591,
+      "step": 13830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012528519117760683,
+      "loss": 0.7133,
+      "step": 13831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012526646913372432,
+      "loss": 0.9752,
+      "step": 13832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00012524774748591216,
+      "loss": 0.4024,
+      "step": 13833
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012522902623447025,
+      "loss": 0.1854,
+      "step": 13834
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012521030537969826,
+      "loss": 0.3296,
+      "step": 13835
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000125191584921896,
+      "loss": 0.0025,
+      "step": 13836
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012517286486136333,
+      "loss": 0.6517,
+      "step": 13837
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012515414519839995,
+      "loss": 0.3881,
+      "step": 13838
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001251354259333056,
+      "loss": 0.4339,
+      "step": 13839
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012511670706638015,
+      "loss": 0.6035,
+      "step": 13840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012509798859792326,
+      "loss": 0.0063,
+      "step": 13841
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012507927052823477,
+      "loss": 0.6451,
+      "step": 13842
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001250605528576143,
+      "loss": 0.0142,
+      "step": 13843
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012504183558636163,
+      "loss": 0.4897,
+      "step": 13844
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012502311871477657,
+      "loss": 1.4215,
+      "step": 13845
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012500440224315874,
+      "loss": 0.0026,
+      "step": 13846
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012498568617180786,
+      "loss": 0.0054,
+      "step": 13847
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012496697050102374,
+      "loss": 0.6413,
+      "step": 13848
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000124948255231106,
+      "loss": 0.3398,
+      "step": 13849
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001249295403623543,
+      "loss": 0.7147,
+      "step": 13850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012491082589506848,
+      "loss": 0.4304,
+      "step": 13851
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012489211182954807,
+      "loss": 0.2871,
+      "step": 13852
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012487339816609285,
+      "loss": 0.2867,
+      "step": 13853
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001248546849050024,
+      "loss": 0.4732,
+      "step": 13854
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012483597204657646,
+      "loss": 0.5125,
+      "step": 13855
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012481725959111466,
+      "loss": 0.5157,
+      "step": 13856
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012479854753891662,
+      "loss": 0.0244,
+      "step": 13857
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012477983589028206,
+      "loss": 0.6146,
+      "step": 13858
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012476112464551058,
+      "loss": 0.0024,
+      "step": 13859
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012474241380490175,
+      "loss": 0.3458,
+      "step": 13860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012472370336875536,
+      "loss": 0.4635,
+      "step": 13861
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012470499333737086,
+      "loss": 0.424,
+      "step": 13862
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012468628371104792,
+      "loss": 0.0618,
+      "step": 13863
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012466757449008622,
+      "loss": 0.8713,
+      "step": 13864
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001246488656747852,
+      "loss": 0.3934,
+      "step": 13865
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001246301572654446,
+      "loss": 0.4068,
+      "step": 13866
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000124611449262364,
+      "loss": 0.0543,
+      "step": 13867
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012459274166584284,
+      "loss": 0.2495,
+      "step": 13868
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012457403447618085,
+      "loss": 0.4119,
+      "step": 13869
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012455532769367755,
+      "loss": 0.2936,
+      "step": 13870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012453662131863244,
+      "loss": 0.5651,
+      "step": 13871
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001245179153513452,
+      "loss": 0.2973,
+      "step": 13872
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012449920979211524,
+      "loss": 0.9219,
+      "step": 13873
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012448050464124215,
+      "loss": 0.2669,
+      "step": 13874
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012446179989902554,
+      "loss": 0.2021,
+      "step": 13875
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012444309556576487,
+      "loss": 0.8055,
+      "step": 13876
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001244243916417596,
+      "loss": 0.2426,
+      "step": 13877
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012440568812730943,
+      "loss": 0.0029,
+      "step": 13878
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012438698502271363,
+      "loss": 1.3795,
+      "step": 13879
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012436828232827193,
+      "loss": 0.4544,
+      "step": 13880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012434958004428365,
+      "loss": 0.5647,
+      "step": 13881
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012433087817104836,
+      "loss": 0.1825,
+      "step": 13882
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012431217670886557,
+      "loss": 0.3364,
+      "step": 13883
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001242934756580347,
+      "loss": 0.6763,
+      "step": 13884
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012427477501885517,
+      "loss": 0.4885,
+      "step": 13885
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001242560747916266,
+      "loss": 1.0593,
+      "step": 13886
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012423737497664833,
+      "loss": 0.1254,
+      "step": 13887
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012421867557421977,
+      "loss": 0.004,
+      "step": 13888
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012419997658464054,
+      "loss": 0.1783,
+      "step": 13889
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012418127800820991,
+      "loss": 1.1179,
+      "step": 13890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001241625798452274,
+      "loss": 1.0805,
+      "step": 13891
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001241438820959923,
+      "loss": 0.181,
+      "step": 13892
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012412518476080423,
+      "loss": 0.0028,
+      "step": 13893
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012410648783996246,
+      "loss": 0.0276,
+      "step": 13894
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001240877913337664,
+      "loss": 0.1471,
+      "step": 13895
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001240690952425155,
+      "loss": 0.7024,
+      "step": 13896
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012405039956650912,
+      "loss": 1.0902,
+      "step": 13897
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012403170430604665,
+      "loss": 0.4073,
+      "step": 13898
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001240130094614274,
+      "loss": 0.5811,
+      "step": 13899
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012399431503295086,
+      "loss": 0.5267,
+      "step": 13900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012397562102091635,
+      "loss": 0.7416,
+      "step": 13901
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012395692742562324,
+      "loss": 0.0068,
+      "step": 13902
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012393823424737075,
+      "loss": 0.603,
+      "step": 13903
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012391954148645838,
+      "loss": 0.2837,
+      "step": 13904
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012390084914318546,
+      "loss": 0.8266,
+      "step": 13905
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012388215721785117,
+      "loss": 0.0509,
+      "step": 13906
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012386346571075497,
+      "loss": 0.8215,
+      "step": 13907
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001238447746221962,
+      "loss": 0.3883,
+      "step": 13908
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012382608395247404,
+      "loss": 0.7274,
+      "step": 13909
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012380739370188793,
+      "loss": 0.1743,
+      "step": 13910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012378870387073706,
+      "loss": 0.355,
+      "step": 13911
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012377001445932072,
+      "loss": 0.3781,
+      "step": 13912
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012375132546793837,
+      "loss": 0.3001,
+      "step": 13913
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000123732636896889,
+      "loss": 0.2303,
+      "step": 13914
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012371394874647208,
+      "loss": 0.0193,
+      "step": 13915
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012369526101698683,
+      "loss": 0.606,
+      "step": 13916
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012367657370873248,
+      "loss": 0.6936,
+      "step": 13917
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001236578868220083,
+      "loss": 0.5733,
+      "step": 13918
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012363920035711356,
+      "loss": 0.0055,
+      "step": 13919
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012362051431434742,
+      "loss": 0.5926,
+      "step": 13920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001236018286940092,
+      "loss": 0.3518,
+      "step": 13921
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012358314349639806,
+      "loss": 0.7903,
+      "step": 13922
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001235644587218132,
+      "loss": 2.0188,
+      "step": 13923
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001235457743705539,
+      "loss": 0.7005,
+      "step": 13924
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012352709044291928,
+      "loss": 0.7003,
+      "step": 13925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012350840693920855,
+      "loss": 0.0044,
+      "step": 13926
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000123489723859721,
+      "loss": 0.4642,
+      "step": 13927
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012347104120475572,
+      "loss": 0.243,
+      "step": 13928
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012345235897461194,
+      "loss": 0.5949,
+      "step": 13929
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012343367716958874,
+      "loss": 0.2771,
+      "step": 13930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012341499578998527,
+      "loss": 0.2874,
+      "step": 13931
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012339631483610083,
+      "loss": 0.2651,
+      "step": 13932
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012337763430823444,
+      "loss": 0.4339,
+      "step": 13933
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012335895420668527,
+      "loss": 0.0814,
+      "step": 13934
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012334027453175252,
+      "loss": 0.2328,
+      "step": 13935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012332159528373522,
+      "loss": 0.323,
+      "step": 13936
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012330291646293253,
+      "loss": 0.6226,
+      "step": 13937
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001232842380696436,
+      "loss": 0.432,
+      "step": 13938
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012326556010416748,
+      "loss": 0.0119,
+      "step": 13939
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012324688256680334,
+      "loss": 0.4803,
+      "step": 13940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012322820545785013,
+      "loss": 0.0152,
+      "step": 13941
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001232095287776071,
+      "loss": 0.4175,
+      "step": 13942
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012319085252637324,
+      "loss": 0.3565,
+      "step": 13943
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012317217670444763,
+      "loss": 0.0124,
+      "step": 13944
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012315350131212935,
+      "loss": 0.3329,
+      "step": 13945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001231348263497175,
+      "loss": 0.3318,
+      "step": 13946
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012311615181751106,
+      "loss": 0.9422,
+      "step": 13947
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012309747771580904,
+      "loss": 0.6541,
+      "step": 13948
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012307880404491064,
+      "loss": 0.0542,
+      "step": 13949
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012306013080511474,
+      "loss": 0.8103,
+      "step": 13950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012304145799672044,
+      "loss": 0.2756,
+      "step": 13951
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012302278562002666,
+      "loss": 1.0101,
+      "step": 13952
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012300411367533254,
+      "loss": 0.9121,
+      "step": 13953
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012298544216293706,
+      "loss": 0.001,
+      "step": 13954
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012296677108313909,
+      "loss": 0.0211,
+      "step": 13955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012294810043623775,
+      "loss": 0.7402,
+      "step": 13956
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012292943022253203,
+      "loss": 0.1101,
+      "step": 13957
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012291076044232077,
+      "loss": 0.1154,
+      "step": 13958
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001228920910959031,
+      "loss": 0.2311,
+      "step": 13959
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012287342218357788,
+      "loss": 0.6997,
+      "step": 13960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012285475370564408,
+      "loss": 0.0149,
+      "step": 13961
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001228360856624007,
+      "loss": 0.8119,
+      "step": 13962
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001228174180541466,
+      "loss": 0.5116,
+      "step": 13963
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012279875088118075,
+      "loss": 0.4868,
+      "step": 13964
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012278008414380216,
+      "loss": 0.2708,
+      "step": 13965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012276141784230954,
+      "loss": 1.2129,
+      "step": 13966
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000122742751977002,
+      "loss": 0.0104,
+      "step": 13967
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012272408654817843,
+      "loss": 0.0043,
+      "step": 13968
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001227054215561376,
+      "loss": 0.0874,
+      "step": 13969
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012268675700117854,
+      "loss": 0.0022,
+      "step": 13970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012266809288360002,
+      "loss": 1.834,
+      "step": 13971
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000122649429203701,
+      "loss": 0.6147,
+      "step": 13972
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012263076596178034,
+      "loss": 0.0094,
+      "step": 13973
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012261210315813687,
+      "loss": 0.5851,
+      "step": 13974
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012259344079306944,
+      "loss": 0.3475,
+      "step": 13975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012257477886687697,
+      "loss": 0.5731,
+      "step": 13976
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012255611737985823,
+      "loss": 0.1914,
+      "step": 13977
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012253745633231214,
+      "loss": 0.0909,
+      "step": 13978
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012251879572453738,
+      "loss": 0.0014,
+      "step": 13979
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001225001355568329,
+      "loss": 0.2881,
+      "step": 13980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012248147582949752,
+      "loss": 0.8618,
+      "step": 13981
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012246281654282995,
+      "loss": 0.013,
+      "step": 13982
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012244415769712905,
+      "loss": 0.1421,
+      "step": 13983
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012242549929269362,
+      "loss": 0.0026,
+      "step": 13984
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012240684132982244,
+      "loss": 0.2664,
+      "step": 13985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012238818380881425,
+      "loss": 0.6055,
+      "step": 13986
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012236952672996793,
+      "loss": 0.4313,
+      "step": 13987
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001223508700935821,
+      "loss": 0.611,
+      "step": 13988
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012233221389995567,
+      "loss": 0.2853,
+      "step": 13989
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012231355814938723,
+      "loss": 0.4261,
+      "step": 13990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012229490284217562,
+      "loss": 0.0673,
+      "step": 13991
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012227624797861962,
+      "loss": 0.4181,
+      "step": 13992
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012225759355901782,
+      "loss": 0.4,
+      "step": 13993
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012223893958366908,
+      "loss": 0.4972,
+      "step": 13994
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012222028605287207,
+      "loss": 0.2312,
+      "step": 13995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001222016329669254,
+      "loss": 0.7816,
+      "step": 13996
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012218298032612792,
+      "loss": 0.5385,
+      "step": 13997
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001221643281307783,
+      "loss": 2.2604,
+      "step": 13998
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012214567638117513,
+      "loss": 0.0179,
+      "step": 13999
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001221270250776172,
+      "loss": 0.0204,
+      "step": 14000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012210837422040307,
+      "loss": 2.4883,
+      "step": 14001
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001220897238098315,
+      "loss": 0.3265,
+      "step": 14002
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012207107384620112,
+      "loss": 0.5656,
+      "step": 14003
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012205242432981055,
+      "loss": 0.2807,
+      "step": 14004
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012203377526095849,
+      "loss": 0.2693,
+      "step": 14005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012201512663994356,
+      "loss": 0.5444,
+      "step": 14006
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012199647846706432,
+      "loss": 0.4627,
+      "step": 14007
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012197783074261953,
+      "loss": 0.2583,
+      "step": 14008
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012195918346690767,
+      "loss": 1.0485,
+      "step": 14009
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012194053664022737,
+      "loss": 0.4891,
+      "step": 14010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012192189026287735,
+      "loss": 0.7213,
+      "step": 14011
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012190324433515606,
+      "loss": 0.1891,
+      "step": 14012
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012188459885736214,
+      "loss": 0.2436,
+      "step": 14013
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012186595382979425,
+      "loss": 0.2784,
+      "step": 14014
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001218473092527508,
+      "loss": 0.1803,
+      "step": 14015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012182866512653047,
+      "loss": 0.2042,
+      "step": 14016
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012181002145143183,
+      "loss": 0.8358,
+      "step": 14017
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001217913782277533,
+      "loss": 0.6867,
+      "step": 14018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012177273545579361,
+      "loss": 0.2939,
+      "step": 14019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012175409313585115,
+      "loss": 1.0797,
+      "step": 14020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012173545126822446,
+      "loss": 0.6789,
+      "step": 14021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012171680985321217,
+      "loss": 0.0015,
+      "step": 14022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001216981688911127,
+      "loss": 0.218,
+      "step": 14023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012167952838222451,
+      "loss": 0.0051,
+      "step": 14024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012166088832684628,
+      "loss": 2.5332,
+      "step": 14025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012164224872527634,
+      "loss": 0.5094,
+      "step": 14026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012162360957781326,
+      "loss": 0.0007,
+      "step": 14027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012160497088475541,
+      "loss": 0.2475,
+      "step": 14028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012158633264640139,
+      "loss": 1.0331,
+      "step": 14029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012156769486304964,
+      "loss": 0.3247,
+      "step": 14030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012154905753499853,
+      "loss": 0.1803,
+      "step": 14031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012153042066254655,
+      "loss": 0.1897,
+      "step": 14032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012151178424599221,
+      "loss": 0.0012,
+      "step": 14033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012149314828563389,
+      "loss": 0.3304,
+      "step": 14034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012147451278176995,
+      "loss": 0.3153,
+      "step": 14035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012145587773469896,
+      "loss": 0.5767,
+      "step": 14036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012143724314471921,
+      "loss": 0.0925,
+      "step": 14037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012141860901212919,
+      "loss": 1.6807,
+      "step": 14038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012139997533722717,
+      "loss": 0.2112,
+      "step": 14039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012138134212031165,
+      "loss": 0.7626,
+      "step": 14040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012136270936168104,
+      "loss": 0.5509,
+      "step": 14041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012134407706163359,
+      "loss": 0.1693,
+      "step": 14042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012132544522046775,
+      "loss": 0.3336,
+      "step": 14043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012130681383848191,
+      "loss": 0.8762,
+      "step": 14044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012128818291597434,
+      "loss": 0.4117,
+      "step": 14045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012126955245324345,
+      "loss": 0.0014,
+      "step": 14046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012125092245058759,
+      "loss": 0.3257,
+      "step": 14047
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012123229290830504,
+      "loss": 0.0012,
+      "step": 14048
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012121366382669417,
+      "loss": 0.4195,
+      "step": 14049
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012119503520605318,
+      "loss": 0.2942,
+      "step": 14050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012117640704668053,
+      "loss": 0.5865,
+      "step": 14051
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001211577793488745,
+      "loss": 0.7244,
+      "step": 14052
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012113915211293329,
+      "loss": 0.0013,
+      "step": 14053
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012112052533915525,
+      "loss": 0.0021,
+      "step": 14054
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012110189902783868,
+      "loss": 0.1542,
+      "step": 14055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012108327317928179,
+      "loss": 0.3951,
+      "step": 14056
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012106464779378295,
+      "loss": 0.2147,
+      "step": 14057
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001210460228716403,
+      "loss": 0.004,
+      "step": 14058
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012102739841315211,
+      "loss": 0.5688,
+      "step": 14059
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012100877441861674,
+      "loss": 0.8705,
+      "step": 14060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012099015088833228,
+      "loss": 0.1821,
+      "step": 14061
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.000120971527822597,
+      "loss": 1.0541,
+      "step": 14062
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012095290522170917,
+      "loss": 0.3629,
+      "step": 14063
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012093428308596697,
+      "loss": 0.0742,
+      "step": 14064
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012091566141566856,
+      "loss": 1.9994,
+      "step": 14065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012089704021111229,
+      "loss": 0.2992,
+      "step": 14066
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012087841947259613,
+      "loss": 0.5206,
+      "step": 14067
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012085979920041844,
+      "loss": 0.3158,
+      "step": 14068
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012084117939487731,
+      "loss": 0.2955,
+      "step": 14069
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001208225600562709,
+      "loss": 1.0396,
+      "step": 14070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012080394118489746,
+      "loss": 0.7827,
+      "step": 14071
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012078532278105506,
+      "loss": 0.6924,
+      "step": 14072
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012076670484504183,
+      "loss": 0.436,
+      "step": 14073
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012074808737715602,
+      "loss": 2.7362,
+      "step": 14074
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012072947037769566,
+      "loss": 0.002,
+      "step": 14075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012071085384695892,
+      "loss": 0.0015,
+      "step": 14076
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012069223778524386,
+      "loss": 0.2882,
+      "step": 14077
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012067362219284864,
+      "loss": 1.2165,
+      "step": 14078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012065500707007139,
+      "loss": 0.4889,
+      "step": 14079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012063639241721007,
+      "loss": 1.0898,
+      "step": 14080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012061777823456292,
+      "loss": 0.0009,
+      "step": 14081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012059916452242798,
+      "loss": 0.4964,
+      "step": 14082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00012058055128110326,
+      "loss": 1.0355,
+      "step": 14083
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012056193851088683,
+      "loss": 0.2179,
+      "step": 14084
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012054332621207683,
+      "loss": 0.1625,
+      "step": 14085
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012052471438497122,
+      "loss": 0.3116,
+      "step": 14086
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012050610302986811,
+      "loss": 0.3248,
+      "step": 14087
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012048749214706544,
+      "loss": 0.0019,
+      "step": 14088
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012046888173686133,
+      "loss": 1.7051,
+      "step": 14089
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012045027179955376,
+      "loss": 0.9524,
+      "step": 14090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001204316623354407,
+      "loss": 0.2757,
+      "step": 14091
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012041305334482022,
+      "loss": 0.5332,
+      "step": 14092
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001203944448279903,
+      "loss": 0.2866,
+      "step": 14093
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012037583678524888,
+      "loss": 0.7887,
+      "step": 14094
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012035722921689399,
+      "loss": 0.0023,
+      "step": 14095
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012033862212322362,
+      "loss": 0.1642,
+      "step": 14096
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012032001550453563,
+      "loss": 0.4408,
+      "step": 14097
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012030140936112815,
+      "loss": 0.5978,
+      "step": 14098
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012028280369329894,
+      "loss": 0.7587,
+      "step": 14099
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012026419850134606,
+      "loss": 0.2828,
+      "step": 14100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012024559378556744,
+      "loss": 2.5811,
+      "step": 14101
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012022698954626092,
+      "loss": 0.1809,
+      "step": 14102
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012020838578372449,
+      "loss": 0.8046,
+      "step": 14103
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012018978249825612,
+      "loss": 0.1816,
+      "step": 14104
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012017117969015354,
+      "loss": 0.1657,
+      "step": 14105
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012015257735971484,
+      "loss": 0.2567,
+      "step": 14106
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012013397550723778,
+      "loss": 0.2523,
+      "step": 14107
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012011537413302024,
+      "loss": 0.4221,
+      "step": 14108
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012009677323736023,
+      "loss": 0.451,
+      "step": 14109
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012007817282055546,
+      "loss": 1.089,
+      "step": 14110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012005957288290382,
+      "loss": 0.2778,
+      "step": 14111
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012004097342470324,
+      "loss": 0.5825,
+      "step": 14112
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012002237444625148,
+      "loss": 0.2654,
+      "step": 14113
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00012000377594784638,
+      "loss": 0.5173,
+      "step": 14114
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011998517792978589,
+      "loss": 0.7534,
+      "step": 14115
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011996658039236763,
+      "loss": 0.3477,
+      "step": 14116
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011994798333588956,
+      "loss": 2.327,
+      "step": 14117
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011992938676064941,
+      "loss": 0.5642,
+      "step": 14118
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011991079066694497,
+      "loss": 0.4693,
+      "step": 14119
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011989219505507413,
+      "loss": 1.1091,
+      "step": 14120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011987359992533456,
+      "loss": 1.0812,
+      "step": 14121
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011985500527802404,
+      "loss": 0.8588,
+      "step": 14122
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011983641111344045,
+      "loss": 1.5293,
+      "step": 14123
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011981781743188142,
+      "loss": 0.0011,
+      "step": 14124
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011979922423364479,
+      "loss": 0.3154,
+      "step": 14125
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011978063151902819,
+      "loss": 0.2094,
+      "step": 14126
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011976203928832944,
+      "loss": 0.7364,
+      "step": 14127
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001197434475418463,
+      "loss": 0.2652,
+      "step": 14128
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011972485627987637,
+      "loss": 2.6063,
+      "step": 14129
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011970626550271746,
+      "loss": 0.2444,
+      "step": 14130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011968767521066729,
+      "loss": 0.4804,
+      "step": 14131
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011966908540402346,
+      "loss": 0.3466,
+      "step": 14132
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011965049608308368,
+      "loss": 0.0023,
+      "step": 14133
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011963190724814572,
+      "loss": 0.3525,
+      "step": 14134
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011961331889950716,
+      "loss": 0.0019,
+      "step": 14135
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011959473103746573,
+      "loss": 0.166,
+      "step": 14136
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011957614366231897,
+      "loss": 0.0075,
+      "step": 14137
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011955755677436466,
+      "loss": 0.7025,
+      "step": 14138
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001195389703739004,
+      "loss": 0.5295,
+      "step": 14139
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011952038446122377,
+      "loss": 0.3953,
+      "step": 14140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011950179903663247,
+      "loss": 0.3002,
+      "step": 14141
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011948321410042411,
+      "loss": 0.7787,
+      "step": 14142
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011946462965289622,
+      "loss": 0.6876,
+      "step": 14143
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011944604569434648,
+      "loss": 1.4109,
+      "step": 14144
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001194274622250725,
+      "loss": 3.3389,
+      "step": 14145
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011940887924537177,
+      "loss": 0.8344,
+      "step": 14146
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011939029675554201,
+      "loss": 0.5419,
+      "step": 14147
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011937171475588065,
+      "loss": 0.3251,
+      "step": 14148
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011935313324668529,
+      "loss": 0.2157,
+      "step": 14149
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011933455222825358,
+      "loss": 0.8986,
+      "step": 14150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011931597170088293,
+      "loss": 0.2712,
+      "step": 14151
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011929739166487097,
+      "loss": 0.2292,
+      "step": 14152
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011927881212051524,
+      "loss": 0.2628,
+      "step": 14153
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011926023306811314,
+      "loss": 0.4793,
+      "step": 14154
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011924165450796235,
+      "loss": 0.0682,
+      "step": 14155
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011922307644036027,
+      "loss": 0.2341,
+      "step": 14156
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001192044988656044,
+      "loss": 0.2953,
+      "step": 14157
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011918592178399232,
+      "loss": 0.9861,
+      "step": 14158
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011916734519582142,
+      "loss": 0.0005,
+      "step": 14159
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001191487691013892,
+      "loss": 0.1117,
+      "step": 14160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001191301935009932,
+      "loss": 0.5344,
+      "step": 14161
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001191116183949308,
+      "loss": 0.0143,
+      "step": 14162
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011909304378349943,
+      "loss": 0.6017,
+      "step": 14163
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011907446966699664,
+      "loss": 0.4992,
+      "step": 14164
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011905589604571978,
+      "loss": 0.5352,
+      "step": 14165
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011903732291996631,
+      "loss": 0.2753,
+      "step": 14166
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011901875029003367,
+      "loss": 0.2549,
+      "step": 14167
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011900017815621917,
+      "loss": 0.1838,
+      "step": 14168
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011898160651882036,
+      "loss": 0.7814,
+      "step": 14169
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011896303537813455,
+      "loss": 0.47,
+      "step": 14170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001189444647344591,
+      "loss": 0.0529,
+      "step": 14171
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011892589458809152,
+      "loss": 0.6328,
+      "step": 14172
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011890732493932907,
+      "loss": 0.7144,
+      "step": 14173
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011888875578846918,
+      "loss": 0.3342,
+      "step": 14174
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001188701871358091,
+      "loss": 0.4267,
+      "step": 14175
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001188516189816463,
+      "loss": 0.3249,
+      "step": 14176
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011883305132627809,
+      "loss": 0.7666,
+      "step": 14177
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011881448417000174,
+      "loss": 0.0037,
+      "step": 14178
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011879591751311464,
+      "loss": 0.2762,
+      "step": 14179
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011877735135591414,
+      "loss": 0.4599,
+      "step": 14180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011875878569869742,
+      "loss": 0.191,
+      "step": 14181
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001187402205417619,
+      "loss": 0.1835,
+      "step": 14182
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011872165588540487,
+      "loss": 0.0013,
+      "step": 14183
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011870309172992353,
+      "loss": 0.0017,
+      "step": 14184
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011868452807561525,
+      "loss": 0.6391,
+      "step": 14185
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001186659649227772,
+      "loss": 5.1563,
+      "step": 14186
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001186474022717067,
+      "loss": 0.5952,
+      "step": 14187
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011862884012270105,
+      "loss": 0.0267,
+      "step": 14188
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011861027847605738,
+      "loss": 0.3952,
+      "step": 14189
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011859171733207303,
+      "loss": 0.0055,
+      "step": 14190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011857315669104521,
+      "loss": 0.5839,
+      "step": 14191
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011855459655327106,
+      "loss": 0.4247,
+      "step": 14192
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011853603691904789,
+      "loss": 0.5061,
+      "step": 14193
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001185174777886729,
+      "loss": 0.504,
+      "step": 14194
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011849891916244317,
+      "loss": 0.6561,
+      "step": 14195
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011848036104065605,
+      "loss": 0.9577,
+      "step": 14196
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011846180342360861,
+      "loss": 1.8152,
+      "step": 14197
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011844324631159803,
+      "loss": 0.1107,
+      "step": 14198
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011842468970492161,
+      "loss": 0.4995,
+      "step": 14199
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011840613360387627,
+      "loss": 1.0409,
+      "step": 14200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011838757800875935,
+      "loss": 0.4356,
+      "step": 14201
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011836902291986793,
+      "loss": 0.8861,
+      "step": 14202
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011835046833749908,
+      "loss": 0.1057,
+      "step": 14203
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011833191426195003,
+      "loss": 0.2758,
+      "step": 14204
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011831336069351784,
+      "loss": 1.4238,
+      "step": 14205
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011829480763249958,
+      "loss": 0.2522,
+      "step": 14206
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011827625507919246,
+      "loss": 0.334,
+      "step": 14207
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011825770303389347,
+      "loss": 0.2093,
+      "step": 14208
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011823915149689969,
+      "loss": 0.7298,
+      "step": 14209
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001182206004685083,
+      "loss": 0.2478,
+      "step": 14210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011820204994901627,
+      "loss": 0.5952,
+      "step": 14211
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011818349993872066,
+      "loss": 0.1133,
+      "step": 14212
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001181649504379186,
+      "loss": 0.2722,
+      "step": 14213
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011814640144690705,
+      "loss": 0.4668,
+      "step": 14214
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011812785296598311,
+      "loss": 0.0011,
+      "step": 14215
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011810930499544371,
+      "loss": 0.0481,
+      "step": 14216
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011809075753558597,
+      "loss": 0.8443,
+      "step": 14217
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011807221058670688,
+      "loss": 0.0014,
+      "step": 14218
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011805366414910339,
+      "loss": 0.441,
+      "step": 14219
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011803511822307249,
+      "loss": 0.292,
+      "step": 14220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011801657280891127,
+      "loss": 0.344,
+      "step": 14221
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001179980279069166,
+      "loss": 0.7043,
+      "step": 14222
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001179794835173855,
+      "loss": 0.0212,
+      "step": 14223
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011796093964061488,
+      "loss": 0.5061,
+      "step": 14224
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011794239627690176,
+      "loss": 0.3619,
+      "step": 14225
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011792385342654306,
+      "loss": 0.6386,
+      "step": 14226
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011790531108983567,
+      "loss": 0.3212,
+      "step": 14227
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011788676926707657,
+      "loss": 0.2653,
+      "step": 14228
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011786822795856269,
+      "loss": 0.0478,
+      "step": 14229
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011784968716459087,
+      "loss": 0.4815,
+      "step": 14230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011783114688545809,
+      "loss": 0.6548,
+      "step": 14231
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011781260712146123,
+      "loss": 0.3564,
+      "step": 14232
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001177940678728971,
+      "loss": 0.0012,
+      "step": 14233
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011777552914006276,
+      "loss": 2.3133,
+      "step": 14234
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011775699092325483,
+      "loss": 1.0532,
+      "step": 14235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011773845322277035,
+      "loss": 0.2495,
+      "step": 14236
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011771991603890615,
+      "loss": 0.0085,
+      "step": 14237
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011770137937195898,
+      "loss": 0.6366,
+      "step": 14238
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001176828432222258,
+      "loss": 0.0078,
+      "step": 14239
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001176643075900034,
+      "loss": 0.911,
+      "step": 14240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011764577247558852,
+      "loss": 0.5545,
+      "step": 14241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011762723787927806,
+      "loss": 0.9462,
+      "step": 14242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011760870380136884,
+      "loss": 0.703,
+      "step": 14243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011759017024215756,
+      "loss": 0.4492,
+      "step": 14244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011757163720194113,
+      "loss": 0.6415,
+      "step": 14245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011755310468101622,
+      "loss": 0.8045,
+      "step": 14246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011753457267967963,
+      "loss": 0.2432,
+      "step": 14247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011751604119822817,
+      "loss": 0.7338,
+      "step": 14248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011749751023695854,
+      "loss": 0.1991,
+      "step": 14249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011747897979616746,
+      "loss": 0.0096,
+      "step": 14250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011746044987615182,
+      "loss": 0.2065,
+      "step": 14251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011744192047720816,
+      "loss": 0.6025,
+      "step": 14252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011742339159963333,
+      "loss": 0.7256,
+      "step": 14253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011740486324372394,
+      "loss": 0.849,
+      "step": 14254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011738633540977674,
+      "loss": 0.3516,
+      "step": 14255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011736780809808849,
+      "loss": 0.4449,
+      "step": 14256
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011734928130895578,
+      "loss": 0.068,
+      "step": 14257
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001173307550426753,
+      "loss": 2.0587,
+      "step": 14258
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011731222929954382,
+      "loss": 2.7053,
+      "step": 14259
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001172937040798579,
+      "loss": 0.8685,
+      "step": 14260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011727517938391418,
+      "loss": 0.476,
+      "step": 14261
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011725665521200942,
+      "loss": 2.3955,
+      "step": 14262
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011723813156444015,
+      "loss": 0.1743,
+      "step": 14263
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011721960844150308,
+      "loss": 2.7637,
+      "step": 14264
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011720108584349468,
+      "loss": 2.3457,
+      "step": 14265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011718256377071173,
+      "loss": 0.5165,
+      "step": 14266
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001171640422234508,
+      "loss": 0.6133,
+      "step": 14267
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011714552120200844,
+      "loss": 0.3223,
+      "step": 14268
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011712700070668119,
+      "loss": 0.1007,
+      "step": 14269
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011710848073776576,
+      "loss": 0.514,
+      "step": 14270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011708996129555862,
+      "loss": 0.5151,
+      "step": 14271
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011707144238035638,
+      "loss": 0.4084,
+      "step": 14272
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011705292399245552,
+      "loss": 0.3279,
+      "step": 14273
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011703440613215265,
+      "loss": 0.825,
+      "step": 14274
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011701588879974435,
+      "loss": 0.783,
+      "step": 14275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011699737199552701,
+      "loss": 0.2373,
+      "step": 14276
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011697885571979726,
+      "loss": 0.2269,
+      "step": 14277
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001169603399728516,
+      "loss": 0.0013,
+      "step": 14278
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011694182475498645,
+      "loss": 0.004,
+      "step": 14279
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001169233100664984,
+      "loss": 0.8305,
+      "step": 14280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011690479590768393,
+      "loss": 0.5598,
+      "step": 14281
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011688628227883941,
+      "loss": 0.4515,
+      "step": 14282
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011686776918026147,
+      "loss": 0.002,
+      "step": 14283
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011684925661224645,
+      "loss": 0.9202,
+      "step": 14284
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011683074457509081,
+      "loss": 1.1182,
+      "step": 14285
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011681223306909105,
+      "loss": 0.6239,
+      "step": 14286
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011679372209454353,
+      "loss": 1.5549,
+      "step": 14287
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011677521165174472,
+      "loss": 0.451,
+      "step": 14288
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011675670174099107,
+      "loss": 0.0791,
+      "step": 14289
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011673819236257889,
+      "loss": 0.4114,
+      "step": 14290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011671968351680467,
+      "loss": 0.1998,
+      "step": 14291
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011670117520396481,
+      "loss": 0.9276,
+      "step": 14292
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011668266742435559,
+      "loss": 0.4961,
+      "step": 14293
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011666416017827352,
+      "loss": 0.8324,
+      "step": 14294
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011664565346601485,
+      "loss": 0.8018,
+      "step": 14295
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011662714728787594,
+      "loss": 1.0769,
+      "step": 14296
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011660864164415326,
+      "loss": 0.0063,
+      "step": 14297
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011659013653514305,
+      "loss": 0.7105,
+      "step": 14298
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011657163196114161,
+      "loss": 0.0048,
+      "step": 14299
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011655312792244539,
+      "loss": 0.8802,
+      "step": 14300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001165346244193506,
+      "loss": 0.2404,
+      "step": 14301
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011651612145215359,
+      "loss": 2.9607,
+      "step": 14302
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011649761902115062,
+      "loss": 0.2596,
+      "step": 14303
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011647911712663797,
+      "loss": 0.4759,
+      "step": 14304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011646061576891201,
+      "loss": 0.8362,
+      "step": 14305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011644211494826894,
+      "loss": 1.0367,
+      "step": 14306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011642361466500499,
+      "loss": 0.1962,
+      "step": 14307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001164051149194165,
+      "loss": 0.1383,
+      "step": 14308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011638661571179968,
+      "loss": 0.4377,
+      "step": 14309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011636811704245072,
+      "loss": 0.2998,
+      "step": 14310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011634961891166596,
+      "loss": 1.169,
+      "step": 14311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001163311213197415,
+      "loss": 0.01,
+      "step": 14312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011631262426697367,
+      "loss": 0.3451,
+      "step": 14313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011629412775365852,
+      "loss": 0.4494,
+      "step": 14314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011627563178009235,
+      "loss": 0.481,
+      "step": 14315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011625713634657136,
+      "loss": 0.0028,
+      "step": 14316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011623864145339163,
+      "loss": 0.2005,
+      "step": 14317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011622014710084941,
+      "loss": 0.0128,
+      "step": 14318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001162016532892409,
+      "loss": 0.006,
+      "step": 14319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011618316001886213,
+      "loss": 0.5919,
+      "step": 14320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011616466729000935,
+      "loss": 3.0505,
+      "step": 14321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011614617510297857,
+      "loss": 0.4555,
+      "step": 14322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011612768345806602,
+      "loss": 0.0053,
+      "step": 14323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001161091923555678,
+      "loss": 0.7796,
+      "step": 14324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011609070179577994,
+      "loss": 0.5442,
+      "step": 14325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011607221177899866,
+      "loss": 0.3847,
+      "step": 14326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011605372230552001,
+      "loss": 0.2084,
+      "step": 14327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011603523337563998,
+      "loss": 0.7174,
+      "step": 14328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011601674498965476,
+      "loss": 0.1575,
+      "step": 14329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0001159982571478604,
+      "loss": 0.2749,
+      "step": 14330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011597976985055284,
+      "loss": 0.5971,
+      "step": 14331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00011596128309802827,
+      "loss": 0.0084,
+      "step": 14332
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011594279689058265,
+      "loss": 0.8225,
+      "step": 14333
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011592431122851198,
+      "loss": 0.6179,
+      "step": 14334
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011590582611211243,
+      "loss": 0.1234,
+      "step": 14335
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011588734154167982,
+      "loss": 0.6096,
+      "step": 14336
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011586885751751026,
+      "loss": 0.1271,
+      "step": 14337
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011585037403989974,
+      "loss": 0.281,
+      "step": 14338
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001158318911091442,
+      "loss": 0.2846,
+      "step": 14339
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011581340872553964,
+      "loss": 0.0052,
+      "step": 14340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011579492688938208,
+      "loss": 0.6898,
+      "step": 14341
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011577644560096738,
+      "loss": 1.1254,
+      "step": 14342
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011575796486059158,
+      "loss": 2.0305,
+      "step": 14343
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011573948466855057,
+      "loss": 0.8496,
+      "step": 14344
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011572100502514026,
+      "loss": 1.0051,
+      "step": 14345
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011570252593065668,
+      "loss": 0.0869,
+      "step": 14346
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011568404738539563,
+      "loss": 0.1663,
+      "step": 14347
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011566556938965302,
+      "loss": 0.6907,
+      "step": 14348
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011564709194372486,
+      "loss": 0.0238,
+      "step": 14349
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011562861504790693,
+      "loss": 0.3505,
+      "step": 14350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011561013870249521,
+      "loss": 0.14,
+      "step": 14351
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011559166290778544,
+      "loss": 0.1288,
+      "step": 14352
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011557318766407352,
+      "loss": 0.0049,
+      "step": 14353
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001155547129716554,
+      "loss": 0.5012,
+      "step": 14354
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011553623883082683,
+      "loss": 0.2818,
+      "step": 14355
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011551776524188363,
+      "loss": 0.5693,
+      "step": 14356
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011549929220512174,
+      "loss": 0.9925,
+      "step": 14357
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011548081972083688,
+      "loss": 0.4537,
+      "step": 14358
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011546234778932485,
+      "loss": 3.0427,
+      "step": 14359
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011544387641088154,
+      "loss": 0.5302,
+      "step": 14360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011542540558580265,
+      "loss": 0.3133,
+      "step": 14361
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011540693531438407,
+      "loss": 0.6897,
+      "step": 14362
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011538846559692141,
+      "loss": 0.0779,
+      "step": 14363
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011536999643371058,
+      "loss": 0.0945,
+      "step": 14364
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011535152782504729,
+      "loss": 0.1794,
+      "step": 14365
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011533305977122724,
+      "loss": 0.5851,
+      "step": 14366
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011531459227254623,
+      "loss": 0.2119,
+      "step": 14367
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011529612532930001,
+      "loss": 0.0013,
+      "step": 14368
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011527765894178418,
+      "loss": 0.3378,
+      "step": 14369
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011525919311029466,
+      "loss": 0.5142,
+      "step": 14370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011524072783512689,
+      "loss": 0.2193,
+      "step": 14371
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011522226311657675,
+      "loss": 0.4818,
+      "step": 14372
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001152037989549399,
+      "loss": 0.0006,
+      "step": 14373
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011518533535051192,
+      "loss": 1.1163,
+      "step": 14374
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011516687230358858,
+      "loss": 0.2742,
+      "step": 14375
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011514840981446553,
+      "loss": 0.1201,
+      "step": 14376
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011512994788343833,
+      "loss": 0.6104,
+      "step": 14377
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011511148651080272,
+      "loss": 0.5179,
+      "step": 14378
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011509302569685433,
+      "loss": 0.1847,
+      "step": 14379
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011507456544188868,
+      "loss": 0.7856,
+      "step": 14380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001150561057462015,
+      "loss": 0.2978,
+      "step": 14381
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011503764661008831,
+      "loss": 0.0012,
+      "step": 14382
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011501918803384471,
+      "loss": 0.0056,
+      "step": 14383
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011500073001776639,
+      "loss": 0.3726,
+      "step": 14384
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011498227256214881,
+      "loss": 0.3322,
+      "step": 14385
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011496381566728755,
+      "loss": 0.2124,
+      "step": 14386
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011494535933347831,
+      "loss": 0.2383,
+      "step": 14387
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011492690356101642,
+      "loss": 0.435,
+      "step": 14388
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011490844835019756,
+      "loss": 0.0227,
+      "step": 14389
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011488999370131728,
+      "loss": 0.7045,
+      "step": 14390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011487153961467099,
+      "loss": 0.4901,
+      "step": 14391
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011485308609055434,
+      "loss": 0.9131,
+      "step": 14392
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011483463312926271,
+      "loss": 0.0026,
+      "step": 14393
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011481618073109165,
+      "loss": 0.0005,
+      "step": 14394
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001147977288963367,
+      "loss": 1.0459,
+      "step": 14395
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011477927762529329,
+      "loss": 0.732,
+      "step": 14396
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011476082691825683,
+      "loss": 0.2452,
+      "step": 14397
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001147423767755229,
+      "loss": 0.037,
+      "step": 14398
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011472392719738687,
+      "loss": 0.0015,
+      "step": 14399
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011470547818414423,
+      "loss": 0.0078,
+      "step": 14400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011468702973609033,
+      "loss": 0.2785,
+      "step": 14401
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011466858185352068,
+      "loss": 0.6033,
+      "step": 14402
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011465013453673071,
+      "loss": 0.4345,
+      "step": 14403
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011463168778601573,
+      "loss": 0.2949,
+      "step": 14404
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011461324160167117,
+      "loss": 0.2008,
+      "step": 14405
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001145947959839925,
+      "loss": 0.9104,
+      "step": 14406
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000114576350933275,
+      "loss": 1.8603,
+      "step": 14407
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011455790644981405,
+      "loss": 0.0012,
+      "step": 14408
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001145394625339051,
+      "loss": 0.0056,
+      "step": 14409
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011452101918584342,
+      "loss": 0.3236,
+      "step": 14410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011450257640592438,
+      "loss": 0.2892,
+      "step": 14411
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011448413419444326,
+      "loss": 0.3272,
+      "step": 14412
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011446569255169544,
+      "loss": 0.3126,
+      "step": 14413
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011444725147797628,
+      "loss": 2.1598,
+      "step": 14414
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011442881097358096,
+      "loss": 0.0029,
+      "step": 14415
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011441037103880486,
+      "loss": 0.0036,
+      "step": 14416
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011439193167394331,
+      "loss": 0.2235,
+      "step": 14417
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011437349287929146,
+      "loss": 0.4329,
+      "step": 14418
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011435505465514476,
+      "loss": 0.3295,
+      "step": 14419
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011433661700179824,
+      "loss": 0.2893,
+      "step": 14420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011431817991954733,
+      "loss": 0.001,
+      "step": 14421
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011429974340868725,
+      "loss": 0.1829,
+      "step": 14422
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011428130746951313,
+      "loss": 0.0014,
+      "step": 14423
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001142628721023203,
+      "loss": 0.0412,
+      "step": 14424
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011424443730740398,
+      "loss": 0.5058,
+      "step": 14425
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011422600308505927,
+      "loss": 1.0164,
+      "step": 14426
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011420756943558144,
+      "loss": 0.4798,
+      "step": 14427
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001141891363592657,
+      "loss": 0.1983,
+      "step": 14428
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011417070385640716,
+      "loss": 0.132,
+      "step": 14429
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011415227192730108,
+      "loss": 1.1882,
+      "step": 14430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011413384057224253,
+      "loss": 0.2229,
+      "step": 14431
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011411540979152667,
+      "loss": 0.3698,
+      "step": 14432
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011409697958544873,
+      "loss": 0.3827,
+      "step": 14433
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011407854995430374,
+      "loss": 1.0788,
+      "step": 14434
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011406012089838683,
+      "loss": 0.0786,
+      "step": 14435
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011404169241799322,
+      "loss": 1.0557,
+      "step": 14436
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001140232645134179,
+      "loss": 0.2665,
+      "step": 14437
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.000114004837184956,
+      "loss": 0.2582,
+      "step": 14438
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011398641043290265,
+      "loss": 0.6406,
+      "step": 14439
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011396798425755282,
+      "loss": 1.4318,
+      "step": 14440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011394955865920172,
+      "loss": 0.1224,
+      "step": 14441
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011393113363814428,
+      "loss": 0.7063,
+      "step": 14442
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001139127091946756,
+      "loss": 0.5362,
+      "step": 14443
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011389428532909076,
+      "loss": 0.4157,
+      "step": 14444
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011387586204168474,
+      "loss": 0.7994,
+      "step": 14445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011385743933275253,
+      "loss": 1.2671,
+      "step": 14446
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011383901720258925,
+      "loss": 0.2462,
+      "step": 14447
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011382059565148982,
+      "loss": 0.0102,
+      "step": 14448
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001138021746797493,
+      "loss": 0.2177,
+      "step": 14449
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011378375428766253,
+      "loss": 0.0952,
+      "step": 14450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011376533447552462,
+      "loss": 0.2757,
+      "step": 14451
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011374691524363055,
+      "loss": 0.0398,
+      "step": 14452
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011372849659227515,
+      "loss": 0.1725,
+      "step": 14453
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011371007852175345,
+      "loss": 0.0205,
+      "step": 14454
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011369166103236045,
+      "loss": 1.6143,
+      "step": 14455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011367324412439094,
+      "loss": 0.2509,
+      "step": 14456
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001136548277981399,
+      "loss": 0.2217,
+      "step": 14457
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001136364120539023,
+      "loss": 0.0026,
+      "step": 14458
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011361799689197295,
+      "loss": 0.8857,
+      "step": 14459
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011359958231264683,
+      "loss": 0.3517,
+      "step": 14460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011358116831621869,
+      "loss": 0.0029,
+      "step": 14461
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011356275490298352,
+      "loss": 0.3323,
+      "step": 14462
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011354434207323617,
+      "loss": 0.246,
+      "step": 14463
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011352592982727141,
+      "loss": 0.0018,
+      "step": 14464
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011350751816538416,
+      "loss": 0.532,
+      "step": 14465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011348910708786929,
+      "loss": 0.8082,
+      "step": 14466
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011347069659502152,
+      "loss": 0.6771,
+      "step": 14467
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011345228668713577,
+      "loss": 0.7005,
+      "step": 14468
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011343387736450675,
+      "loss": 0.6925,
+      "step": 14469
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001134154686274293,
+      "loss": 0.3512,
+      "step": 14470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011339706047619831,
+      "loss": 0.3644,
+      "step": 14471
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011337865291110836,
+      "loss": 0.0015,
+      "step": 14472
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011336024593245435,
+      "loss": 0.0127,
+      "step": 14473
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011334183954053105,
+      "loss": 0.322,
+      "step": 14474
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011332343373563312,
+      "loss": 0.7685,
+      "step": 14475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011330502851805536,
+      "loss": 0.175,
+      "step": 14476
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011328662388809255,
+      "loss": 0.7478,
+      "step": 14477
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001132682198460393,
+      "loss": 0.7614,
+      "step": 14478
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011324981639219042,
+      "loss": 0.006,
+      "step": 14479
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011323141352684056,
+      "loss": 1.1813,
+      "step": 14480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001132130112502844,
+      "loss": 0.0024,
+      "step": 14481
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011319460956281673,
+      "loss": 0.4793,
+      "step": 14482
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001131762084647321,
+      "loss": 0.0038,
+      "step": 14483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001131578079563252,
+      "loss": 0.247,
+      "step": 14484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011313940803789079,
+      "loss": 1.0818,
+      "step": 14485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011312100870972339,
+      "loss": 0.2863,
+      "step": 14486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011310260997211766,
+      "loss": 0.1365,
+      "step": 14487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011308421182536837,
+      "loss": 0.0029,
+      "step": 14488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001130658142697699,
+      "loss": 0.2206,
+      "step": 14489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011304741730561707,
+      "loss": 0.5626,
+      "step": 14490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011302902093320434,
+      "loss": 0.465,
+      "step": 14491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011301062515282633,
+      "loss": 0.0012,
+      "step": 14492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011299222996477773,
+      "loss": 0.6814,
+      "step": 14493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011297383536935299,
+      "loss": 0.3113,
+      "step": 14494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011295544136684664,
+      "loss": 0.147,
+      "step": 14495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001129370479575534,
+      "loss": 0.3181,
+      "step": 14496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011291865514176766,
+      "loss": 0.4442,
+      "step": 14497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011290026291978405,
+      "loss": 0.0005,
+      "step": 14498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011288187129189701,
+      "loss": 0.0109,
+      "step": 14499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011286348025840109,
+      "loss": 0.002,
+      "step": 14500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011284508981959085,
+      "loss": 0.358,
+      "step": 14501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011282669997576067,
+      "loss": 0.8324,
+      "step": 14502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011280831072720511,
+      "loss": 0.0007,
+      "step": 14503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001127899220742187,
+      "loss": 0.7764,
+      "step": 14504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001127715340170958,
+      "loss": 0.311,
+      "step": 14505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011275314655613086,
+      "loss": 0.0006,
+      "step": 14506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011273475969161847,
+      "loss": 0.2368,
+      "step": 14507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011271637342385293,
+      "loss": 0.9777,
+      "step": 14508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011269798775312874,
+      "loss": 0.2949,
+      "step": 14509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011267960267974024,
+      "loss": 0.2728,
+      "step": 14510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011266121820398193,
+      "loss": 0.0003,
+      "step": 14511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011264283432614819,
+      "loss": 0.3245,
+      "step": 14512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011262445104653335,
+      "loss": 1.2422,
+      "step": 14513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011260606836543184,
+      "loss": 0.0006,
+      "step": 14514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011258768628313808,
+      "loss": 0.6836,
+      "step": 14515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011256930479994631,
+      "loss": 0.4203,
+      "step": 14516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011255092391615101,
+      "loss": 0.2923,
+      "step": 14517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011253254363204642,
+      "loss": 0.2447,
+      "step": 14518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011251416394792688,
+      "loss": 0.1977,
+      "step": 14519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011249578486408681,
+      "loss": 0.6098,
+      "step": 14520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011247740638082043,
+      "loss": 1.8114,
+      "step": 14521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011245902849842205,
+      "loss": 2.8125,
+      "step": 14522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011244065121718607,
+      "loss": 0.8623,
+      "step": 14523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001124222745374066,
+      "loss": 0.6842,
+      "step": 14524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011240389845937805,
+      "loss": 0.3781,
+      "step": 14525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011238552298339466,
+      "loss": 0.4224,
+      "step": 14526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011236714810975061,
+      "loss": 0.3595,
+      "step": 14527
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011234877383874027,
+      "loss": 4.326,
+      "step": 14528
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011233040017065776,
+      "loss": 0.3906,
+      "step": 14529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011231202710579735,
+      "loss": 0.0037,
+      "step": 14530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011229365464445331,
+      "loss": 0.3563,
+      "step": 14531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001122752827869198,
+      "loss": 0.1403,
+      "step": 14532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011225691153349094,
+      "loss": 0.7792,
+      "step": 14533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011223854088446109,
+      "loss": 0.0065,
+      "step": 14534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011222017084012431,
+      "loss": 1.2579,
+      "step": 14535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011220180140077475,
+      "loss": 0.2462,
+      "step": 14536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011218343256670668,
+      "loss": 0.0721,
+      "step": 14537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011216506433821415,
+      "loss": 0.7054,
+      "step": 14538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011214669671559137,
+      "loss": 0.9184,
+      "step": 14539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001121283296991324,
+      "loss": 0.1951,
+      "step": 14540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011210996328913136,
+      "loss": 0.6133,
+      "step": 14541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011209159748588245,
+      "loss": 1.1302,
+      "step": 14542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011207323228967969,
+      "loss": 1.0533,
+      "step": 14543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011205486770081716,
+      "loss": 0.8035,
+      "step": 14544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011203650371958903,
+      "loss": 0.5435,
+      "step": 14545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011201814034628929,
+      "loss": 0.0028,
+      "step": 14546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011199977758121208,
+      "loss": 0.1935,
+      "step": 14547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011198141542465131,
+      "loss": 0.0024,
+      "step": 14548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011196305387690116,
+      "loss": 0.0308,
+      "step": 14549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011194469293825566,
+      "loss": 0.2662,
+      "step": 14550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001119263326090087,
+      "loss": 0.8299,
+      "step": 14551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011190797288945443,
+      "loss": 0.216,
+      "step": 14552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011188961377988683,
+      "loss": 0.1798,
+      "step": 14553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011187125528059983,
+      "loss": 1.0297,
+      "step": 14554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011185289739188746,
+      "loss": 0.6976,
+      "step": 14555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011183454011404371,
+      "loss": 0.2201,
+      "step": 14556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011181618344736253,
+      "loss": 0.3225,
+      "step": 14557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011179782739213786,
+      "loss": 0.8786,
+      "step": 14558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011177947194866361,
+      "loss": 0.3082,
+      "step": 14559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011176111711723377,
+      "loss": 0.4775,
+      "step": 14560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001117427628981423,
+      "loss": 2.6309,
+      "step": 14561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011172440929168299,
+      "loss": 0.0479,
+      "step": 14562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011170605629814986,
+      "loss": 0.3168,
+      "step": 14563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001116877039178368,
+      "loss": 0.2849,
+      "step": 14564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011166935215103761,
+      "loss": 0.5572,
+      "step": 14565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011165100099804626,
+      "loss": 0.2986,
+      "step": 14566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011163265045915654,
+      "loss": 0.6042,
+      "step": 14567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011161430053466233,
+      "loss": 0.2676,
+      "step": 14568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011159595122485754,
+      "loss": 0.3584,
+      "step": 14569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011157760253003593,
+      "loss": 0.2691,
+      "step": 14570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011155925445049131,
+      "loss": 0.8913,
+      "step": 14571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011154090698651765,
+      "loss": 0.6346,
+      "step": 14572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011152256013840854,
+      "loss": 0.2197,
+      "step": 14573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001115042139064579,
+      "loss": 0.317,
+      "step": 14574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011148586829095957,
+      "loss": 0.7089,
+      "step": 14575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011146752329220716,
+      "loss": 0.835,
+      "step": 14576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011144917891049462,
+      "loss": 0.1539,
+      "step": 14577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011143083514611557,
+      "loss": 0.7145,
+      "step": 14578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011141249199936379,
+      "loss": 0.9752,
+      "step": 14579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0001113941494705331,
+      "loss": 0.2651,
+      "step": 14580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00011137580755991715,
+      "loss": 0.6076,
+      "step": 14581
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011135746626780962,
+      "loss": 0.425,
+      "step": 14582
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011133912559450435,
+      "loss": 0.6532,
+      "step": 14583
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011132078554029493,
+      "loss": 0.0029,
+      "step": 14584
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011130244610547504,
+      "loss": 0.3159,
+      "step": 14585
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011128410729033848,
+      "loss": 0.3869,
+      "step": 14586
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001112657690951788,
+      "loss": 0.0083,
+      "step": 14587
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001112474315202897,
+      "loss": 0.108,
+      "step": 14588
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011122909456596483,
+      "loss": 0.2775,
+      "step": 14589
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011121075823249779,
+      "loss": 0.003,
+      "step": 14590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001111924225201823,
+      "loss": 0.3041,
+      "step": 14591
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011117408742931185,
+      "loss": 2.2514,
+      "step": 14592
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011115575296018014,
+      "loss": 0.0437,
+      "step": 14593
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011113741911308079,
+      "loss": 0.1455,
+      "step": 14594
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011111908588830732,
+      "loss": 0.0006,
+      "step": 14595
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011110075328615336,
+      "loss": 0.4173,
+      "step": 14596
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011108242130691242,
+      "loss": 0.0024,
+      "step": 14597
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001110640899508781,
+      "loss": 0.0038,
+      "step": 14598
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000111045759218344,
+      "loss": 0.3905,
+      "step": 14599
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011102742910960352,
+      "loss": 1.0615,
+      "step": 14600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001110090996249503,
+      "loss": 0.0642,
+      "step": 14601
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011099077076467789,
+      "loss": 0.5538,
+      "step": 14602
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011097244252907966,
+      "loss": 0.0016,
+      "step": 14603
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011095411491844922,
+      "loss": 0.3462,
+      "step": 14604
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011093578793308008,
+      "loss": 2.0087,
+      "step": 14605
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011091746157326559,
+      "loss": 0.9555,
+      "step": 14606
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001108991358392994,
+      "loss": 0.4786,
+      "step": 14607
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011088081073147477,
+      "loss": 0.8291,
+      "step": 14608
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011086248625008528,
+      "loss": 0.2547,
+      "step": 14609
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011084416239542438,
+      "loss": 0.3632,
+      "step": 14610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011082583916778537,
+      "loss": 0.7783,
+      "step": 14611
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011080751656746178,
+      "loss": 1.0357,
+      "step": 14612
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011078919459474704,
+      "loss": 0.9246,
+      "step": 14613
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011077087324993442,
+      "loss": 0.1009,
+      "step": 14614
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011075255253331748,
+      "loss": 0.4196,
+      "step": 14615
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011073423244518948,
+      "loss": 0.2803,
+      "step": 14616
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011071591298584375,
+      "loss": 0.4923,
+      "step": 14617
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011069759415557381,
+      "loss": 0.3595,
+      "step": 14618
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011067927595467287,
+      "loss": 0.0019,
+      "step": 14619
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011066095838343427,
+      "loss": 1.2882,
+      "step": 14620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011064264144215147,
+      "loss": 0.6078,
+      "step": 14621
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011062432513111766,
+      "loss": 0.3587,
+      "step": 14622
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011060600945062616,
+      "loss": 0.3538,
+      "step": 14623
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011058769440097039,
+      "loss": 0.9815,
+      "step": 14624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011056937998244344,
+      "loss": 0.2757,
+      "step": 14625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011055106619533875,
+      "loss": 0.2751,
+      "step": 14626
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011053275303994952,
+      "loss": 0.1542,
+      "step": 14627
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011051444051656899,
+      "loss": 0.5919,
+      "step": 14628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011049612862549047,
+      "loss": 0.0037,
+      "step": 14629
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011047781736700717,
+      "loss": 0.2248,
+      "step": 14630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011045950674141224,
+      "loss": 0.1284,
+      "step": 14631
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011044119674899905,
+      "loss": 0.3052,
+      "step": 14632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001104228873900607,
+      "loss": 0.0056,
+      "step": 14633
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011040457866489037,
+      "loss": 0.2647,
+      "step": 14634
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011038627057378134,
+      "loss": 0.6636,
+      "step": 14635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011036796311702673,
+      "loss": 0.002,
+      "step": 14636
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011034965629491973,
+      "loss": 0.0015,
+      "step": 14637
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011033135010775341,
+      "loss": 0.7823,
+      "step": 14638
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011031304455582101,
+      "loss": 0.3668,
+      "step": 14639
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011029473963941567,
+      "loss": 0.5466,
+      "step": 14640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011027643535883047,
+      "loss": 0.639,
+      "step": 14641
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011025813171435851,
+      "loss": 0.5732,
+      "step": 14642
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011023982870629299,
+      "loss": 2.1934,
+      "step": 14643
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001102215263349269,
+      "loss": 1.0088,
+      "step": 14644
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011020322460055338,
+      "loss": 1.1127,
+      "step": 14645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011018492350346547,
+      "loss": 0.2314,
+      "step": 14646
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011016662304395626,
+      "loss": 0.0024,
+      "step": 14647
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011014832322231884,
+      "loss": 2.1118,
+      "step": 14648
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011013002403884615,
+      "loss": 0.1993,
+      "step": 14649
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011011172549383131,
+      "loss": 0.4448,
+      "step": 14650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011009342758756738,
+      "loss": 0.4761,
+      "step": 14651
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011007513032034724,
+      "loss": 0.2666,
+      "step": 14652
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000110056833692464,
+      "loss": 0.0014,
+      "step": 14653
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011003853770421064,
+      "loss": 0.0021,
+      "step": 14654
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011002024235588005,
+      "loss": 0.6279,
+      "step": 14655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00011000194764776541,
+      "loss": 0.225,
+      "step": 14656
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010998365358015944,
+      "loss": 0.3566,
+      "step": 14657
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010996536015335524,
+      "loss": 0.0723,
+      "step": 14658
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010994706736764575,
+      "loss": 0.7227,
+      "step": 14659
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010992877522332378,
+      "loss": 0.5405,
+      "step": 14660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010991048372068239,
+      "loss": 0.0033,
+      "step": 14661
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010989219286001446,
+      "loss": 0.5541,
+      "step": 14662
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010987390264161279,
+      "loss": 0.0247,
+      "step": 14663
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010985561306577042,
+      "loss": 0.0031,
+      "step": 14664
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010983732413278013,
+      "loss": 0.3942,
+      "step": 14665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010981903584293478,
+      "loss": 0.985,
+      "step": 14666
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010980074819652736,
+      "loss": 0.4088,
+      "step": 14667
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010978246119385055,
+      "loss": 0.0023,
+      "step": 14668
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010976417483519727,
+      "loss": 1.8337,
+      "step": 14669
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010974588912086039,
+      "loss": 0.1972,
+      "step": 14670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010972760405113266,
+      "loss": 0.2399,
+      "step": 14671
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010970931962630688,
+      "loss": 0.0007,
+      "step": 14672
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010969103584667593,
+      "loss": 0.1943,
+      "step": 14673
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010967275271253252,
+      "loss": 0.3321,
+      "step": 14674
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010965447022416949,
+      "loss": 0.2095,
+      "step": 14675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010963618838187952,
+      "loss": 0.507,
+      "step": 14676
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010961790718595542,
+      "loss": 0.2105,
+      "step": 14677
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010959962663668996,
+      "loss": 0.8946,
+      "step": 14678
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010958134673437582,
+      "loss": 0.2174,
+      "step": 14679
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010956306747930573,
+      "loss": 0.3689,
+      "step": 14680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010954478887177248,
+      "loss": 0.0064,
+      "step": 14681
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010952651091206871,
+      "loss": 0.8042,
+      "step": 14682
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010950823360048708,
+      "loss": 0.4541,
+      "step": 14683
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010948995693732037,
+      "loss": 0.1999,
+      "step": 14684
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001094716809228612,
+      "loss": 0.3047,
+      "step": 14685
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010945340555740225,
+      "loss": 0.2754,
+      "step": 14686
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010943513084123609,
+      "loss": 0.0079,
+      "step": 14687
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010941685677465546,
+      "loss": 0.1025,
+      "step": 14688
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000109398583357953,
+      "loss": 0.3962,
+      "step": 14689
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010938031059142125,
+      "loss": 0.4546,
+      "step": 14690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010936203847535287,
+      "loss": 0.0015,
+      "step": 14691
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010934376701004051,
+      "loss": 0.0026,
+      "step": 14692
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010932549619577665,
+      "loss": 0.4004,
+      "step": 14693
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010930722603285396,
+      "loss": 1.6665,
+      "step": 14694
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010928895652156492,
+      "loss": 0.5954,
+      "step": 14695
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010927068766220217,
+      "loss": 0.4231,
+      "step": 14696
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010925241945505824,
+      "loss": 0.4881,
+      "step": 14697
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010923415190042563,
+      "loss": 0.4735,
+      "step": 14698
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010921588499859693,
+      "loss": 0.2099,
+      "step": 14699
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010919761874986461,
+      "loss": 0.2556,
+      "step": 14700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010917935315452118,
+      "loss": 0.4194,
+      "step": 14701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010916108821285914,
+      "loss": 0.2808,
+      "step": 14702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010914282392517103,
+      "loss": 0.0015,
+      "step": 14703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010912456029174922,
+      "loss": 0.0016,
+      "step": 14704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010910629731288629,
+      "loss": 0.1824,
+      "step": 14705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010908803498887459,
+      "loss": 0.6569,
+      "step": 14706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010906977332000659,
+      "loss": 1.6807,
+      "step": 14707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010905151230657484,
+      "loss": 0.0087,
+      "step": 14708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010903325194887155,
+      "loss": 0.8658,
+      "step": 14709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001090149922471893,
+      "loss": 0.999,
+      "step": 14710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010899673320182048,
+      "loss": 0.1044,
+      "step": 14711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010897847481305733,
+      "loss": 0.3172,
+      "step": 14712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010896021708119247,
+      "loss": 0.1995,
+      "step": 14713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010894196000651807,
+      "loss": 0.1848,
+      "step": 14714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010892370358932654,
+      "loss": 0.517,
+      "step": 14715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010890544782991031,
+      "loss": 0.0794,
+      "step": 14716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010888719272856161,
+      "loss": 0.419,
+      "step": 14717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010886893828557282,
+      "loss": 0.2948,
+      "step": 14718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010885068450123628,
+      "loss": 0.7661,
+      "step": 14719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010883243137584427,
+      "loss": 0.6813,
+      "step": 14720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010881417890968903,
+      "loss": 0.5229,
+      "step": 14721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010879592710306298,
+      "loss": 0.0049,
+      "step": 14722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010877767595625831,
+      "loss": 0.1615,
+      "step": 14723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010875942546956731,
+      "loss": 0.3583,
+      "step": 14724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010874117564328217,
+      "loss": 0.2663,
+      "step": 14725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010872292647769515,
+      "loss": 0.2676,
+      "step": 14726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001087046779730986,
+      "loss": 0.3578,
+      "step": 14727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010868643012978461,
+      "loss": 1.0452,
+      "step": 14728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010866818294804541,
+      "loss": 0.1867,
+      "step": 14729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010864993642817328,
+      "loss": 0.9662,
+      "step": 14730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010863169057046034,
+      "loss": 0.3177,
+      "step": 14731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010861344537519876,
+      "loss": 0.2664,
+      "step": 14732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010859520084268081,
+      "loss": 0.0015,
+      "step": 14733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010857695697319855,
+      "loss": 0.5272,
+      "step": 14734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010855871376704417,
+      "loss": 0.0011,
+      "step": 14735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010854047122450974,
+      "loss": 0.2158,
+      "step": 14736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010852222934588746,
+      "loss": 1.5274,
+      "step": 14737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010850398813146949,
+      "loss": 0.1593,
+      "step": 14738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010848574758154779,
+      "loss": 0.3618,
+      "step": 14739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010846750769641456,
+      "loss": 0.4072,
+      "step": 14740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001084492684763619,
+      "loss": 0.6762,
+      "step": 14741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010843102992168182,
+      "loss": 0.185,
+      "step": 14742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010841279203266644,
+      "loss": 0.2401,
+      "step": 14743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010839455480960773,
+      "loss": 0.659,
+      "step": 14744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010837631825279781,
+      "loss": 0.3709,
+      "step": 14745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001083580823625287,
+      "loss": 0.4316,
+      "step": 14746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010833984713909235,
+      "loss": 0.5276,
+      "step": 14747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010832161258278087,
+      "loss": 0.5378,
+      "step": 14748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010830337869388624,
+      "loss": 0.4571,
+      "step": 14749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010828514547270036,
+      "loss": 1.043,
+      "step": 14750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010826691291951528,
+      "loss": 0.003,
+      "step": 14751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010824868103462301,
+      "loss": 0.5531,
+      "step": 14752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001082304498183154,
+      "loss": 1.1473,
+      "step": 14753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010821221927088449,
+      "loss": 0.4743,
+      "step": 14754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010819398939262215,
+      "loss": 0.6154,
+      "step": 14755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001081757601838203,
+      "loss": 0.0011,
+      "step": 14756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010815753164477093,
+      "loss": 0.2416,
+      "step": 14757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010813930377576587,
+      "loss": 0.0089,
+      "step": 14758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010812107657709702,
+      "loss": 0.4052,
+      "step": 14759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010810285004905636,
+      "loss": 0.2727,
+      "step": 14760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010808462419193557,
+      "loss": 0.4298,
+      "step": 14761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010806639900602671,
+      "loss": 0.5781,
+      "step": 14762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010804817449162149,
+      "loss": 1.3631,
+      "step": 14763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010802995064901176,
+      "loss": 1.0791,
+      "step": 14764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010801172747848944,
+      "loss": 0.0025,
+      "step": 14765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010799350498034624,
+      "loss": 0.1974,
+      "step": 14766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010797528315487401,
+      "loss": 0.3088,
+      "step": 14767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010795706200236461,
+      "loss": 0.5072,
+      "step": 14768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010793884152310974,
+      "loss": 0.3434,
+      "step": 14769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010792062171740115,
+      "loss": 0.0036,
+      "step": 14770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010790240258553072,
+      "loss": 0.3309,
+      "step": 14771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010788418412779011,
+      "loss": 0.0042,
+      "step": 14772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010786596634447112,
+      "loss": 0.696,
+      "step": 14773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010784774923586537,
+      "loss": 1.1383,
+      "step": 14774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010782953280226467,
+      "loss": 0.0367,
+      "step": 14775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010781131704396076,
+      "loss": 0.1414,
+      "step": 14776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010779310196124528,
+      "loss": 0.0017,
+      "step": 14777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010777488755440987,
+      "loss": 0.1906,
+      "step": 14778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010775667382374636,
+      "loss": 0.0022,
+      "step": 14779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010773846076954625,
+      "loss": 0.8626,
+      "step": 14780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010772024839210123,
+      "loss": 0.3651,
+      "step": 14781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010770203669170308,
+      "loss": 0.5507,
+      "step": 14782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010768382566864328,
+      "loss": 0.001,
+      "step": 14783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010766561532321354,
+      "loss": 0.001,
+      "step": 14784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010764740565570536,
+      "loss": 0.6408,
+      "step": 14785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010762919666641046,
+      "loss": 0.394,
+      "step": 14786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001076109883556204,
+      "loss": 0.0011,
+      "step": 14787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010759278072362671,
+      "loss": 0.2614,
+      "step": 14788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010757457377072101,
+      "loss": 0.9424,
+      "step": 14789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010755636749719488,
+      "loss": 0.4955,
+      "step": 14790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010753816190333975,
+      "loss": 0.9434,
+      "step": 14791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010751995698944732,
+      "loss": 2.2417,
+      "step": 14792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010750175275580894,
+      "loss": 0.6512,
+      "step": 14793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010748354920271624,
+      "loss": 0.1261,
+      "step": 14794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010746534633046072,
+      "loss": 0.814,
+      "step": 14795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001074471441393338,
+      "loss": 0.6855,
+      "step": 14796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010742894262962701,
+      "loss": 0.6734,
+      "step": 14797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010741074180163184,
+      "loss": 0.0162,
+      "step": 14798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010739254165563967,
+      "loss": 0.7668,
+      "step": 14799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010737434219194204,
+      "loss": 0.5906,
+      "step": 14800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010735614341083038,
+      "loss": 0.3931,
+      "step": 14801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010735614341083038,
+      "loss": 0.8901,
+      "step": 14802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000107337945312596,
+      "loss": 2.5762,
+      "step": 14803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0001073197478975305,
+      "loss": 0.3137,
+      "step": 14804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010730155116592511,
+      "loss": 0.5889,
+      "step": 14805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010728335511807128,
+      "loss": 0.3298,
+      "step": 14806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010726515975426049,
+      "loss": 0.4318,
+      "step": 14807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010724696507478396,
+      "loss": 0.5659,
+      "step": 14808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010722877107993312,
+      "loss": 0.4792,
+      "step": 14809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010721057776999942,
+      "loss": 0.5168,
+      "step": 14810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010719238514527399,
+      "loss": 0.1048,
+      "step": 14811
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010717419320604833,
+      "loss": 0.0645,
+      "step": 14812
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010715600195261365,
+      "loss": 0.3304,
+      "step": 14813
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010713781138526131,
+      "loss": 0.0061,
+      "step": 14814
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010711962150428261,
+      "loss": 0.0086,
+      "step": 14815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010710143230996882,
+      "loss": 0.8167,
+      "step": 14816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010708324380261116,
+      "loss": 0.2917,
+      "step": 14817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010706505598250099,
+      "loss": 0.8571,
+      "step": 14818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010704686884992951,
+      "loss": 0.0034,
+      "step": 14819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010702868240518792,
+      "loss": 0.0008,
+      "step": 14820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010701049664856755,
+      "loss": 0.6386,
+      "step": 14821
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010699231158035953,
+      "loss": 2.9579,
+      "step": 14822
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010697412720085511,
+      "loss": 0.3718,
+      "step": 14823
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010695594351034542,
+      "loss": 0.0187,
+      "step": 14824
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010693776050912172,
+      "loss": 0.4073,
+      "step": 14825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010691957819747522,
+      "loss": 0.3226,
+      "step": 14826
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010690139657569694,
+      "loss": 0.1905,
+      "step": 14827
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010688321564407813,
+      "loss": 1.1498,
+      "step": 14828
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010686503540290995,
+      "loss": 0.2189,
+      "step": 14829
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00010684685585248346,
+      "loss": 0.2917,
+      "step": 14830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010682867699308978,
+      "loss": 0.0019,
+      "step": 14831
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010681049882502011,
+      "loss": 0.1929,
+      "step": 14832
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010679232134856546,
+      "loss": 0.0029,
+      "step": 14833
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010677414456401694,
+      "loss": 0.9451,
+      "step": 14834
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010675596847166559,
+      "loss": 0.0016,
+      "step": 14835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010673779307180252,
+      "loss": 0.6539,
+      "step": 14836
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001067196183647188,
+      "loss": 0.29,
+      "step": 14837
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010670144435070538,
+      "loss": 0.3922,
+      "step": 14838
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010668327103005339,
+      "loss": 0.4498,
+      "step": 14839
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010666509840305381,
+      "loss": 0.3895,
+      "step": 14840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001066469264699976,
+      "loss": 1.8682,
+      "step": 14841
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010662875523117585,
+      "loss": 0.0037,
+      "step": 14842
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010661058468687945,
+      "loss": 0.0011,
+      "step": 14843
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010659241483739942,
+      "loss": 0.432,
+      "step": 14844
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010657424568302679,
+      "loss": 0.4111,
+      "step": 14845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010655607722405235,
+      "loss": 0.393,
+      "step": 14846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010653790946076718,
+      "loss": 0.5895,
+      "step": 14847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010651974239346218,
+      "loss": 0.2185,
+      "step": 14848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010650157602242818,
+      "loss": 0.3414,
+      "step": 14849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010648341034795622,
+      "loss": 0.6071,
+      "step": 14850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010646524537033713,
+      "loss": 0.003,
+      "step": 14851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010644708108986176,
+      "loss": 0.0031,
+      "step": 14852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010642891750682107,
+      "loss": 0.0013,
+      "step": 14853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010641075462150586,
+      "loss": 0.0043,
+      "step": 14854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010639259243420695,
+      "loss": 0.0054,
+      "step": 14855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001063744309452153,
+      "loss": 0.8861,
+      "step": 14856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010635627015482162,
+      "loss": 0.5553,
+      "step": 14857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010633811006331676,
+      "loss": 0.2403,
+      "step": 14858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010631995067099159,
+      "loss": 0.0222,
+      "step": 14859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010630179197813684,
+      "loss": 0.6209,
+      "step": 14860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010628363398504332,
+      "loss": 0.7984,
+      "step": 14861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010626547669200179,
+      "loss": 0.0016,
+      "step": 14862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010624732009930295,
+      "loss": 0.9644,
+      "step": 14863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001062291642072377,
+      "loss": 1.0411,
+      "step": 14864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010621100901609666,
+      "loss": 0.3929,
+      "step": 14865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010619285452617056,
+      "loss": 0.8289,
+      "step": 14866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001061747007377502,
+      "loss": 0.2791,
+      "step": 14867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010615654765112621,
+      "loss": 0.367,
+      "step": 14868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010613839526658927,
+      "loss": 0.0033,
+      "step": 14869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010612024358443018,
+      "loss": 1.0115,
+      "step": 14870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010610209260493951,
+      "loss": 0.111,
+      "step": 14871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010608394232840795,
+      "loss": 0.2543,
+      "step": 14872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001060657927551261,
+      "loss": 1.0665,
+      "step": 14873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010604764388538469,
+      "loss": 0.4869,
+      "step": 14874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001060294957194743,
+      "loss": 0.4393,
+      "step": 14875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001060113482576855,
+      "loss": 0.0063,
+      "step": 14876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010599320150030896,
+      "loss": 0.6751,
+      "step": 14877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010597505544763531,
+      "loss": 0.1086,
+      "step": 14878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010595691009995502,
+      "loss": 2.1172,
+      "step": 14879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001059387654575587,
+      "loss": 0.1384,
+      "step": 14880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010592062152073696,
+      "loss": 1.8427,
+      "step": 14881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010590247828978032,
+      "loss": 0.9363,
+      "step": 14882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001058843357649793,
+      "loss": 0.5318,
+      "step": 14883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010586619394662439,
+      "loss": 0.3708,
+      "step": 14884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010584805283500618,
+      "loss": 0.3907,
+      "step": 14885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010582991243041518,
+      "loss": 0.0216,
+      "step": 14886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010581177273314178,
+      "loss": 0.4161,
+      "step": 14887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010579363374347654,
+      "loss": 0.5877,
+      "step": 14888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010577549546170996,
+      "loss": 0.1838,
+      "step": 14889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010575735788813236,
+      "loss": 0.2557,
+      "step": 14890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010573922102303436,
+      "loss": 0.1676,
+      "step": 14891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010572108486670627,
+      "loss": 0.4233,
+      "step": 14892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010570294941943852,
+      "loss": 0.4342,
+      "step": 14893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010568481468152165,
+      "loss": 0.5038,
+      "step": 14894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010566668065324584,
+      "loss": 0.0016,
+      "step": 14895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010564854733490169,
+      "loss": 0.0314,
+      "step": 14896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001056304147267795,
+      "loss": 2.1222,
+      "step": 14897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010561228282916957,
+      "loss": 2.6582,
+      "step": 14898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010559415164236233,
+      "loss": 0.326,
+      "step": 14899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010557602116664817,
+      "loss": 0.4624,
+      "step": 14900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010555789140231727,
+      "loss": 0.3336,
+      "step": 14901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001055397623496601,
+      "loss": 0.2278,
+      "step": 14902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010552163400896689,
+      "loss": 0.204,
+      "step": 14903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010550350638052792,
+      "loss": 0.3612,
+      "step": 14904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010548537946463358,
+      "loss": 0.2657,
+      "step": 14905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010546725326157406,
+      "loss": 0.2218,
+      "step": 14906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010544912777163961,
+      "loss": 0.1669,
+      "step": 14907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010543100299512058,
+      "loss": 0.0885,
+      "step": 14908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010541287893230712,
+      "loss": 0.5833,
+      "step": 14909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010539475558348951,
+      "loss": 0.4207,
+      "step": 14910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001053766329489579,
+      "loss": 0.0085,
+      "step": 14911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010535851102900255,
+      "loss": 0.867,
+      "step": 14912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001053403898239137,
+      "loss": 0.0021,
+      "step": 14913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010532226933398145,
+      "loss": 0.8899,
+      "step": 14914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010530414955949595,
+      "loss": 0.6312,
+      "step": 14915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010528603050074748,
+      "loss": 0.4561,
+      "step": 14916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010526791215802611,
+      "loss": 0.2724,
+      "step": 14917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010524979453162198,
+      "loss": 0.0869,
+      "step": 14918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010523167762182527,
+      "loss": 0.0043,
+      "step": 14919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010521356142892602,
+      "loss": 0.3243,
+      "step": 14920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010519544595321442,
+      "loss": 0.8552,
+      "step": 14921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010517733119498042,
+      "loss": 0.7812,
+      "step": 14922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010515921715451423,
+      "loss": 0.0064,
+      "step": 14923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010514110383210592,
+      "loss": 0.0019,
+      "step": 14924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010512299122804546,
+      "loss": 0.3148,
+      "step": 14925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010510487934262294,
+      "loss": 0.1771,
+      "step": 14926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010508676817612848,
+      "loss": 0.1738,
+      "step": 14927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010506865772885191,
+      "loss": 0.8983,
+      "step": 14928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010505054800108343,
+      "loss": 0.1719,
+      "step": 14929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010503243899311296,
+      "loss": 0.0029,
+      "step": 14930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010501433070523047,
+      "loss": 1.9268,
+      "step": 14931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010499622313772598,
+      "loss": 0.5559,
+      "step": 14932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001049781162908894,
+      "loss": 0.0757,
+      "step": 14933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010496001016501076,
+      "loss": 0.3481,
+      "step": 14934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010494190476037995,
+      "loss": 0.1541,
+      "step": 14935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010492380007728686,
+      "loss": 2.2539,
+      "step": 14936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010490569611602151,
+      "loss": 0.2782,
+      "step": 14937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001048875928768738,
+      "loss": 0.3821,
+      "step": 14938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001048694903601335,
+      "loss": 0.0189,
+      "step": 14939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010485138856609064,
+      "loss": 0.3426,
+      "step": 14940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010483328749503504,
+      "loss": 0.5379,
+      "step": 14941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010481518714725648,
+      "loss": 1.3063,
+      "step": 14942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010479708752304497,
+      "loss": 0.4701,
+      "step": 14943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010477898862269023,
+      "loss": 1.2295,
+      "step": 14944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010476089044648207,
+      "loss": 0.4057,
+      "step": 14945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010474279299471047,
+      "loss": 0.0041,
+      "step": 14946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010472469626766503,
+      "loss": 0.8083,
+      "step": 14947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010470660026563566,
+      "loss": 0.3596,
+      "step": 14948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010468850498891212,
+      "loss": 1.7074,
+      "step": 14949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010467041043778411,
+      "loss": 0.4551,
+      "step": 14950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010465231661254152,
+      "loss": 0.1329,
+      "step": 14951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010463422351347399,
+      "loss": 1.9924,
+      "step": 14952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010461613114087126,
+      "loss": 0.0011,
+      "step": 14953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010459803949502313,
+      "loss": 0.0135,
+      "step": 14954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010457994857621923,
+      "loss": 0.0944,
+      "step": 14955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010456185838474926,
+      "loss": 0.2865,
+      "step": 14956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000104543768920903,
+      "loss": 0.7851,
+      "step": 14957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010452568018497001,
+      "loss": 0.0352,
+      "step": 14958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010450759217724006,
+      "loss": 0.4597,
+      "step": 14959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010448950489800267,
+      "loss": 0.043,
+      "step": 14960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010447141834754761,
+      "loss": 0.2319,
+      "step": 14961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010445333252616447,
+      "loss": 0.3139,
+      "step": 14962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010443524743414282,
+      "loss": 0.2771,
+      "step": 14963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010441716307177229,
+      "loss": 0.0021,
+      "step": 14964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010439907943934252,
+      "loss": 0.4412,
+      "step": 14965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010438099653714302,
+      "loss": 1.3482,
+      "step": 14966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010436291436546338,
+      "loss": 0.3353,
+      "step": 14967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010434483292459325,
+      "loss": 1.6041,
+      "step": 14968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010432675221482202,
+      "loss": 0.3701,
+      "step": 14969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001043086722364394,
+      "loss": 3.6387,
+      "step": 14970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010429059298973473,
+      "loss": 0.0747,
+      "step": 14971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010427251447499765,
+      "loss": 0.338,
+      "step": 14972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010425443669251765,
+      "loss": 0.0109,
+      "step": 14973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010423635964258413,
+      "loss": 0.731,
+      "step": 14974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010421828332548666,
+      "loss": 0.9165,
+      "step": 14975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010420020774151469,
+      "loss": 0.0041,
+      "step": 14976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010418213289095762,
+      "loss": 0.1656,
+      "step": 14977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010416405877410495,
+      "loss": 0.4938,
+      "step": 14978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010414598539124611,
+      "loss": 0.7864,
+      "step": 14979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010412791274267044,
+      "loss": 0.7506,
+      "step": 14980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001041098408286675,
+      "loss": 0.0613,
+      "step": 14981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001040917696495265,
+      "loss": 0.4414,
+      "step": 14982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010407369920553694,
+      "loss": 0.192,
+      "step": 14983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010405562949698819,
+      "loss": 0.3572,
+      "step": 14984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010403756052416951,
+      "loss": 0.261,
+      "step": 14985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010401949228737035,
+      "loss": 1.8898,
+      "step": 14986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010400142478688008,
+      "loss": 0.041,
+      "step": 14987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010398335802298784,
+      "loss": 0.4363,
+      "step": 14988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010396529199598316,
+      "loss": 0.5263,
+      "step": 14989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010394722670615519,
+      "loss": 0.0139,
+      "step": 14990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010392916215379326,
+      "loss": 0.4527,
+      "step": 14991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001039110983391867,
+      "loss": 0.5144,
+      "step": 14992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010389303526262472,
+      "loss": 0.2024,
+      "step": 14993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010387497292439651,
+      "loss": 0.3273,
+      "step": 14994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010385691132479149,
+      "loss": 0.0048,
+      "step": 14995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010383885046409873,
+      "loss": 0.8354,
+      "step": 14996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010382079034260747,
+      "loss": 0.0045,
+      "step": 14997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010380273096060705,
+      "loss": 0.2048,
+      "step": 14998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010378467231838646,
+      "loss": 0.9668,
+      "step": 14999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010376661441623505,
+      "loss": 0.26,
+      "step": 15000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001037485572544419,
+      "loss": 0.7271,
+      "step": 15001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010373050083329614,
+      "loss": 0.3593,
+      "step": 15002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010371244515308705,
+      "loss": 0.4067,
+      "step": 15003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010369439021410365,
+      "loss": 0.6024,
+      "step": 15004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010367633601663505,
+      "loss": 2.1353,
+      "step": 15005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010365828256097047,
+      "loss": 0.4924,
+      "step": 15006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010364022984739894,
+      "loss": 0.0017,
+      "step": 15007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010362217787620958,
+      "loss": 1.8284,
+      "step": 15008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010360412664769135,
+      "loss": 0.2511,
+      "step": 15009
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010358607616213345,
+      "loss": 0.3351,
+      "step": 15010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001035680264198249,
+      "loss": 0.9037,
+      "step": 15011
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010354997742105466,
+      "loss": 0.1964,
+      "step": 15012
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010353192916611187,
+      "loss": 0.4293,
+      "step": 15013
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010351388165528552,
+      "loss": 0.4332,
+      "step": 15014
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010349583488886456,
+      "loss": 0.2311,
+      "step": 15015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010347778886713798,
+      "loss": 0.3352,
+      "step": 15016
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010345974359039484,
+      "loss": 0.4768,
+      "step": 15017
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010344169905892404,
+      "loss": 0.9189,
+      "step": 15018
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010342365527301458,
+      "loss": 0.0044,
+      "step": 15019
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010340561223295532,
+      "loss": 1.8391,
+      "step": 15020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010338756993903528,
+      "loss": 0.0004,
+      "step": 15021
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010336952839154339,
+      "loss": 0.9086,
+      "step": 15022
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010335148759076847,
+      "loss": 0.475,
+      "step": 15023
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010333344753699947,
+      "loss": 0.5417,
+      "step": 15024
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010331540823052532,
+      "loss": 0.5606,
+      "step": 15025
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010329736967163477,
+      "loss": 0.4931,
+      "step": 15026
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010327933186061678,
+      "loss": 0.3604,
+      "step": 15027
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010326129479776022,
+      "loss": 0.0022,
+      "step": 15028
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010324325848335382,
+      "loss": 0.0257,
+      "step": 15029
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010322522291768654,
+      "loss": 0.6104,
+      "step": 15030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010320718810104703,
+      "loss": 0.0195,
+      "step": 15031
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010318915403372422,
+      "loss": 0.4051,
+      "step": 15032
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010317112071600687,
+      "loss": 0.31,
+      "step": 15033
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010315308814818368,
+      "loss": 0.4558,
+      "step": 15034
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010313505633054351,
+      "loss": 0.3133,
+      "step": 15035
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001031170252633751,
+      "loss": 0.3435,
+      "step": 15036
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010309899494696713,
+      "loss": 0.3996,
+      "step": 15037
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010308096538160842,
+      "loss": 0.8772,
+      "step": 15038
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001030629365675876,
+      "loss": 0.0006,
+      "step": 15039
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010304490850519337,
+      "loss": 0.4172,
+      "step": 15040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010302688119471453,
+      "loss": 0.0013,
+      "step": 15041
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010300885463643965,
+      "loss": 0.868,
+      "step": 15042
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010299082883065742,
+      "loss": 0.1764,
+      "step": 15043
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010297280377765656,
+      "loss": 0.2159,
+      "step": 15044
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010295477947772567,
+      "loss": 0.0104,
+      "step": 15045
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010293675593115333,
+      "loss": 0.0021,
+      "step": 15046
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010291873313822829,
+      "loss": 1.3589,
+      "step": 15047
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.000102900711099239,
+      "loss": 0.2467,
+      "step": 15048
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001028826898144742,
+      "loss": 2.318,
+      "step": 15049
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010286466928422239,
+      "loss": 0.6882,
+      "step": 15050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010284664950877212,
+      "loss": 0.5095,
+      "step": 15051
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010282863048841206,
+      "loss": 0.6412,
+      "step": 15052
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010281061222343064,
+      "loss": 0.0003,
+      "step": 15053
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010279259471411643,
+      "loss": 1.3248,
+      "step": 15054
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010277457796075803,
+      "loss": 0.0071,
+      "step": 15055
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010275656196364386,
+      "loss": 0.001,
+      "step": 15056
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010273854672306246,
+      "loss": 0.001,
+      "step": 15057
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010272053223930225,
+      "loss": 0.1427,
+      "step": 15058
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001027025185126518,
+      "loss": 0.3531,
+      "step": 15059
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010268450554339954,
+      "loss": 1.8921,
+      "step": 15060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010266649333183386,
+      "loss": 0.5074,
+      "step": 15061
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001026484818782433,
+      "loss": 0.1196,
+      "step": 15062
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010263047118291626,
+      "loss": 3.7598,
+      "step": 15063
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010261246124614108,
+      "loss": 0.0052,
+      "step": 15064
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010259445206820621,
+      "loss": 0.3087,
+      "step": 15065
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010257644364940011,
+      "loss": 0.5047,
+      "step": 15066
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010255843599001106,
+      "loss": 0.4072,
+      "step": 15067
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010254042909032747,
+      "loss": 0.3455,
+      "step": 15068
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010252242295063764,
+      "loss": 0.0114,
+      "step": 15069
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010250441757122996,
+      "loss": 0.5522,
+      "step": 15070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001024864129523928,
+      "loss": 0.078,
+      "step": 15071
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010246840909441436,
+      "loss": 0.5119,
+      "step": 15072
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010245040599758306,
+      "loss": 0.0035,
+      "step": 15073
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010243240366218716,
+      "loss": 0.9856,
+      "step": 15074
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010241440208851487,
+      "loss": 0.3299,
+      "step": 15075
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010239640127685452,
+      "loss": 0.3492,
+      "step": 15076
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010237840122749443,
+      "loss": 0.6933,
+      "step": 15077
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010236040194072269,
+      "loss": 0.2525,
+      "step": 15078
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001023424034168277,
+      "loss": 0.6484,
+      "step": 15079
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00010232440565609756,
+      "loss": 0.4563,
+      "step": 15080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010230640865882045,
+      "loss": 0.004,
+      "step": 15081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010228841242528474,
+      "loss": 0.4074,
+      "step": 15082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010227041695577842,
+      "loss": 0.267,
+      "step": 15083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010225242225058975,
+      "loss": 0.0024,
+      "step": 15084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010223442831000691,
+      "loss": 0.2917,
+      "step": 15085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010221643513431798,
+      "loss": 0.6113,
+      "step": 15086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010219844272381116,
+      "loss": 0.9182,
+      "step": 15087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010218045107877453,
+      "loss": 1.5018,
+      "step": 15088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010216246019949615,
+      "loss": 0.653,
+      "step": 15089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010214447008626427,
+      "loss": 0.4585,
+      "step": 15090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010212648073936682,
+      "loss": 0.2736,
+      "step": 15091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010210849215909194,
+      "loss": 0.4165,
+      "step": 15092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010209050434572772,
+      "loss": 0.295,
+      "step": 15093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010207251729956214,
+      "loss": 0.5363,
+      "step": 15094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010205453102088324,
+      "loss": 0.2959,
+      "step": 15095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010203654550997912,
+      "loss": 0.1168,
+      "step": 15096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010201856076713772,
+      "loss": 0.3529,
+      "step": 15097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010200057679264709,
+      "loss": 0.0025,
+      "step": 15098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010198259358679514,
+      "loss": 0.6693,
+      "step": 15099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010196461114986989,
+      "loss": 0.002,
+      "step": 15100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010194662948215934,
+      "loss": 0.5174,
+      "step": 15101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010192864858395137,
+      "loss": 0.4882,
+      "step": 15102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010191066845553392,
+      "loss": 0.0123,
+      "step": 15103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010189268909719502,
+      "loss": 0.0241,
+      "step": 15104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010187471050922244,
+      "loss": 0.0033,
+      "step": 15105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010185673269190421,
+      "loss": 1.3949,
+      "step": 15106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010183875564552808,
+      "loss": 0.7972,
+      "step": 15107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010182077937038203,
+      "loss": 0.5079,
+      "step": 15108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010180280386675391,
+      "loss": 1.05,
+      "step": 15109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010178482913493152,
+      "loss": 0.6394,
+      "step": 15110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010176685517520273,
+      "loss": 0.4749,
+      "step": 15111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010174888198785542,
+      "loss": 0.0009,
+      "step": 15112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010173090957317728,
+      "loss": 0.0009,
+      "step": 15113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001017129379314562,
+      "loss": 0.358,
+      "step": 15114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010169496706298001,
+      "loss": 0.6922,
+      "step": 15115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010167699696803635,
+      "loss": 0.2027,
+      "step": 15116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010165902764691314,
+      "loss": 0.5388,
+      "step": 15117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010164105909989797,
+      "loss": 0.8345,
+      "step": 15118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010162309132727868,
+      "loss": 0.0008,
+      "step": 15119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010160512432934303,
+      "loss": 0.254,
+      "step": 15120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010158715810637862,
+      "loss": 0.4651,
+      "step": 15121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010156919265867324,
+      "loss": 0.5897,
+      "step": 15122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010155122798651457,
+      "loss": 0.881,
+      "step": 15123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010153326409019023,
+      "loss": 0.208,
+      "step": 15124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010151530096998797,
+      "loss": 0.5386,
+      "step": 15125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001014973386261954,
+      "loss": 0.2147,
+      "step": 15126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010147937705910011,
+      "loss": 0.287,
+      "step": 15127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010146141626898986,
+      "loss": 2.0026,
+      "step": 15128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010144345625615212,
+      "loss": 0.7656,
+      "step": 15129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010142549702087453,
+      "loss": 0.5163,
+      "step": 15130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010140753856344479,
+      "loss": 0.4159,
+      "step": 15131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010138958088415035,
+      "loss": 0.582,
+      "step": 15132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001013716239832788,
+      "loss": 0.4867,
+      "step": 15133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010135366786111777,
+      "loss": 0.775,
+      "step": 15134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010133571251795466,
+      "loss": 0.2464,
+      "step": 15135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010131775795407714,
+      "loss": 0.6584,
+      "step": 15136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010129980416977266,
+      "loss": 0.2527,
+      "step": 15137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010128185116532867,
+      "loss": 0.3244,
+      "step": 15138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001012638989410328,
+      "loss": 0.385,
+      "step": 15139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010124594749717238,
+      "loss": 0.4338,
+      "step": 15140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010122799683403492,
+      "loss": 0.0041,
+      "step": 15141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000101210046951908,
+      "loss": 0.2257,
+      "step": 15142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010119209785107887,
+      "loss": 0.0108,
+      "step": 15143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010117414953183505,
+      "loss": 0.5151,
+      "step": 15144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010115620199446398,
+      "loss": 0.6945,
+      "step": 15145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010113825523925301,
+      "loss": 0.7273,
+      "step": 15146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001011203092664896,
+      "loss": 0.0195,
+      "step": 15147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010110236407646099,
+      "loss": 0.3683,
+      "step": 15148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010108441966945467,
+      "loss": 0.7068,
+      "step": 15149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000101066476045758,
+      "loss": 1.6039,
+      "step": 15150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010104853320565825,
+      "loss": 0.0372,
+      "step": 15151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010103059114944274,
+      "loss": 0.1996,
+      "step": 15152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010101264987739887,
+      "loss": 0.3089,
+      "step": 15153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010099470938981387,
+      "loss": 0.4875,
+      "step": 15154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010097676968697509,
+      "loss": 0.0009,
+      "step": 15155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010095883076916969,
+      "loss": 0.3388,
+      "step": 15156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010094089263668506,
+      "loss": 0.0019,
+      "step": 15157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010092295528980843,
+      "loss": 0.5501,
+      "step": 15158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010090501872882696,
+      "loss": 0.0094,
+      "step": 15159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010088708295402796,
+      "loss": 0.3001,
+      "step": 15160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010086914796569867,
+      "loss": 0.1666,
+      "step": 15161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010085121376412615,
+      "loss": 0.941,
+      "step": 15162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010083328034959772,
+      "loss": 1.9141,
+      "step": 15163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010081534772240056,
+      "loss": 0.001,
+      "step": 15164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010079741588282171,
+      "loss": 0.2036,
+      "step": 15165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001007794848311485,
+      "loss": 0.103,
+      "step": 15166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010076155456766788,
+      "loss": 0.0032,
+      "step": 15167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001007436250926671,
+      "loss": 0.0031,
+      "step": 15168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010072569640643328,
+      "loss": 0.3072,
+      "step": 15169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010070776850925341,
+      "loss": 0.6019,
+      "step": 15170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010068984140141466,
+      "loss": 0.1681,
+      "step": 15171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010067191508320414,
+      "loss": 0.0009,
+      "step": 15172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010065398955490883,
+      "loss": 1.6729,
+      "step": 15173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010063606481681582,
+      "loss": 0.7373,
+      "step": 15174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001006181408692122,
+      "loss": 0.2629,
+      "step": 15175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010060021771238487,
+      "loss": 2.1768,
+      "step": 15176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010058229534662097,
+      "loss": 1.6939,
+      "step": 15177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010056437377220743,
+      "loss": 0.2534,
+      "step": 15178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010054645298943121,
+      "loss": 0.001,
+      "step": 15179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001005285329985794,
+      "loss": 0.9043,
+      "step": 15180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010051061379993885,
+      "loss": 1.1169,
+      "step": 15181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010049269539379651,
+      "loss": 0.5593,
+      "step": 15182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010047477778043946,
+      "loss": 0.7514,
+      "step": 15183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001004568609601544,
+      "loss": 0.7823,
+      "step": 15184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010043894493322844,
+      "loss": 4.1577,
+      "step": 15185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010042102969994836,
+      "loss": 0.3502,
+      "step": 15186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010040311526060104,
+      "loss": 0.284,
+      "step": 15187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010038520161547347,
+      "loss": 2.7325,
+      "step": 15188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001003672887648524,
+      "loss": 0.0008,
+      "step": 15189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001003493767090247,
+      "loss": 0.4104,
+      "step": 15190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010033146544827724,
+      "loss": 0.3155,
+      "step": 15191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010031355498289682,
+      "loss": 0.0026,
+      "step": 15192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001002956453131702,
+      "loss": 0.9568,
+      "step": 15193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010027773643938429,
+      "loss": 0.1915,
+      "step": 15194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010025982836182576,
+      "loss": 0.0024,
+      "step": 15195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010024192108078149,
+      "loss": 1.9105,
+      "step": 15196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010022401459653811,
+      "loss": 1.0469,
+      "step": 15197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010020610890938247,
+      "loss": 0.1785,
+      "step": 15198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010018820401960129,
+      "loss": 0.0159,
+      "step": 15199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010017029992748122,
+      "loss": 0.7079,
+      "step": 15200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.000100152396633309,
+      "loss": 0.7727,
+      "step": 15201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010013449413737137,
+      "loss": 0.9996,
+      "step": 15202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010011659243995498,
+      "loss": 0.1823,
+      "step": 15203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010009869154134652,
+      "loss": 0.6667,
+      "step": 15204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010008079144183255,
+      "loss": 0.9579,
+      "step": 15205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010006289214169982,
+      "loss": 0.146,
+      "step": 15206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010004499364123499,
+      "loss": 2.5352,
+      "step": 15207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001000270959407245,
+      "loss": 0.21,
+      "step": 15208
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00010000919904045515,
+      "loss": 0.2336,
+      "step": 15209
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.999130294071346e-05,
+      "loss": 0.3374,
+      "step": 15210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.997340764178591e-05,
+      "loss": 0.8533,
+      "step": 15211
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.995551314395924e-05,
+      "loss": 1.0357,
+      "step": 15212
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.993761944751993e-05,
+      "loss": 0.1676,
+      "step": 15213
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.991972655275443e-05,
+      "loss": 1.6841,
+      "step": 15214
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.990183445994944e-05,
+      "loss": 0.4966,
+      "step": 15215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.988394316939136e-05,
+      "loss": 0.3198,
+      "step": 15216
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.986605268136666e-05,
+      "loss": 0.0049,
+      "step": 15217
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.984816299616197e-05,
+      "loss": 0.208,
+      "step": 15218
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.983027411406361e-05,
+      "loss": 0.749,
+      "step": 15219
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.981238603535815e-05,
+      "loss": 0.4625,
+      "step": 15220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.979449876033204e-05,
+      "loss": 0.647,
+      "step": 15221
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.977661228927161e-05,
+      "loss": 0.1302,
+      "step": 15222
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.975872662246339e-05,
+      "loss": 0.6632,
+      "step": 15223
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.974084176019379e-05,
+      "loss": 0.01,
+      "step": 15224
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.972295770274914e-05,
+      "loss": 0.3458,
+      "step": 15225
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.97050744504159e-05,
+      "loss": 0.3361,
+      "step": 15226
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.96871920034804e-05,
+      "loss": 0.4045,
+      "step": 15227
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.966931036222898e-05,
+      "loss": 0.664,
+      "step": 15228
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.965142952694808e-05,
+      "loss": 0.2946,
+      "step": 15229
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.963354949792393e-05,
+      "loss": 1.3533,
+      "step": 15230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.961567027544286e-05,
+      "loss": 0.3796,
+      "step": 15231
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.95977918597913e-05,
+      "loss": 0.3939,
+      "step": 15232
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.957991425125541e-05,
+      "loss": 0.1985,
+      "step": 15233
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.956203745012157e-05,
+      "loss": 1.066,
+      "step": 15234
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.954416145667597e-05,
+      "loss": 0.883,
+      "step": 15235
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.952628627120487e-05,
+      "loss": 0.0277,
+      "step": 15236
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.950841189399459e-05,
+      "loss": 0.0012,
+      "step": 15237
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.94905383253313e-05,
+      "loss": 0.3524,
+      "step": 15238
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.947266556550119e-05,
+      "loss": 0.507,
+      "step": 15239
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.945479361479058e-05,
+      "loss": 0.0058,
+      "step": 15240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.943692247348557e-05,
+      "loss": 0.7576,
+      "step": 15241
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.941905214187231e-05,
+      "loss": 0.3202,
+      "step": 15242
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.940118262023711e-05,
+      "loss": 0.2083,
+      "step": 15243
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.938331390886599e-05,
+      "loss": 0.9917,
+      "step": 15244
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.936544600804515e-05,
+      "loss": 1.2861,
+      "step": 15245
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.934757891806065e-05,
+      "loss": 1.1914,
+      "step": 15246
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.93297126391987e-05,
+      "loss": 0.1096,
+      "step": 15247
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.931184717174535e-05,
+      "loss": 0.9787,
+      "step": 15248
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.929398251598667e-05,
+      "loss": 0.2939,
+      "step": 15249
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.927611867220878e-05,
+      "loss": 0.0607,
+      "step": 15250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.925825564069774e-05,
+      "loss": 0.773,
+      "step": 15251
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.924039342173957e-05,
+      "loss": 0.1665,
+      "step": 15252
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.922253201562034e-05,
+      "loss": 0.3162,
+      "step": 15253
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.9204671422626e-05,
+      "loss": 0.003,
+      "step": 15254
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.918681164304265e-05,
+      "loss": 0.3717,
+      "step": 15255
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.916895267715626e-05,
+      "loss": 0.0037,
+      "step": 15256
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.915109452525276e-05,
+      "loss": 0.4493,
+      "step": 15257
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.91332371876182e-05,
+      "loss": 0.0023,
+      "step": 15258
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.911538066453852e-05,
+      "loss": 0.6817,
+      "step": 15259
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.909752495629961e-05,
+      "loss": 0.6605,
+      "step": 15260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.907967006318746e-05,
+      "loss": 0.494,
+      "step": 15261
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.906181598548802e-05,
+      "loss": 0.004,
+      "step": 15262
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.904396272348709e-05,
+      "loss": 0.3537,
+      "step": 15263
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.902611027747068e-05,
+      "loss": 0.6904,
+      "step": 15264
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.900825864772457e-05,
+      "loss": 0.5523,
+      "step": 15265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.899040783453464e-05,
+      "loss": 0.8265,
+      "step": 15266
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.897255783818687e-05,
+      "loss": 0.5193,
+      "step": 15267
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.895470865896692e-05,
+      "loss": 0.2961,
+      "step": 15268
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.893686029716072e-05,
+      "loss": 0.0015,
+      "step": 15269
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.891901275305411e-05,
+      "loss": 0.0643,
+      "step": 15270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.890116602693278e-05,
+      "loss": 0.0213,
+      "step": 15271
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.888332011908262e-05,
+      "loss": 2.1163,
+      "step": 15272
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.88654750297894e-05,
+      "loss": 0.2328,
+      "step": 15273
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.88476307593388e-05,
+      "loss": 0.9557,
+      "step": 15274
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.882978730801669e-05,
+      "loss": 0.4678,
+      "step": 15275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.881194467610871e-05,
+      "loss": 3.2971,
+      "step": 15276
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.879410286390056e-05,
+      "loss": 0.3885,
+      "step": 15277
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.877626187167809e-05,
+      "loss": 0.5736,
+      "step": 15278
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.875842169972686e-05,
+      "loss": 1.1569,
+      "step": 15279
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.87405823483326e-05,
+      "loss": 1.1471,
+      "step": 15280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.872274381778103e-05,
+      "loss": 0.2466,
+      "step": 15281
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.870490610835772e-05,
+      "loss": 0.4318,
+      "step": 15282
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.868706922034842e-05,
+      "loss": 0.4884,
+      "step": 15283
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.866923315403858e-05,
+      "loss": 0.2133,
+      "step": 15284
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.865139790971401e-05,
+      "loss": 0.0263,
+      "step": 15285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.863356348766026e-05,
+      "loss": 0.3919,
+      "step": 15286
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.861572988816287e-05,
+      "loss": 0.4757,
+      "step": 15287
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.859789711150741e-05,
+      "loss": 0.3689,
+      "step": 15288
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.858006515797954e-05,
+      "loss": 0.0455,
+      "step": 15289
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.856223402786475e-05,
+      "loss": 0.5366,
+      "step": 15290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.854440372144852e-05,
+      "loss": 0.6296,
+      "step": 15291
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.852657423901654e-05,
+      "loss": 0.19,
+      "step": 15292
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.850874558085418e-05,
+      "loss": 0.4941,
+      "step": 15293
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.8490917747247e-05,
+      "loss": 0.8481,
+      "step": 15294
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.84730907384804e-05,
+      "loss": 0.6095,
+      "step": 15295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.845526455483997e-05,
+      "loss": 0.0128,
+      "step": 15296
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.843743919661113e-05,
+      "loss": 0.3682,
+      "step": 15297
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.841961466407927e-05,
+      "loss": 1.1728,
+      "step": 15298
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.840179095752991e-05,
+      "loss": 0.4509,
+      "step": 15299
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.838396807724846e-05,
+      "loss": 0.0024,
+      "step": 15300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.836614602352022e-05,
+      "loss": 0.7781,
+      "step": 15301
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.834832479663077e-05,
+      "loss": 0.1197,
+      "step": 15302
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.833050439686526e-05,
+      "loss": 0.0027,
+      "step": 15303
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.831268482450924e-05,
+      "loss": 0.2543,
+      "step": 15304
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.829486607984801e-05,
+      "loss": 0.3332,
+      "step": 15305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.827704816316682e-05,
+      "loss": 0.4363,
+      "step": 15306
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.825923107475115e-05,
+      "loss": 0.0243,
+      "step": 15307
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.824141481488625e-05,
+      "loss": 0.4367,
+      "step": 15308
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.822359938385736e-05,
+      "loss": 0.1083,
+      "step": 15309
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.820578478194986e-05,
+      "loss": 0.273,
+      "step": 15310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.8187971009449e-05,
+      "loss": 0.4569,
+      "step": 15311
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.817015806663997e-05,
+      "loss": 0.0055,
+      "step": 15312
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.815234595380813e-05,
+      "loss": 0.8091,
+      "step": 15313
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.813453467123862e-05,
+      "loss": 0.5932,
+      "step": 15314
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.811672421921668e-05,
+      "loss": 0.1529,
+      "step": 15315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.809891459802761e-05,
+      "loss": 0.3729,
+      "step": 15316
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.808110580795648e-05,
+      "loss": 0.0011,
+      "step": 15317
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.806329784928849e-05,
+      "loss": 0.4699,
+      "step": 15318
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.804549072230894e-05,
+      "loss": 0.2959,
+      "step": 15319
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.802768442730277e-05,
+      "loss": 0.5265,
+      "step": 15320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.800987896455529e-05,
+      "loss": 0.3578,
+      "step": 15321
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.79920743343516e-05,
+      "loss": 0.5146,
+      "step": 15322
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.797427053697671e-05,
+      "loss": 0.7329,
+      "step": 15323
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.795646757271588e-05,
+      "loss": 1.2967,
+      "step": 15324
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.793866544185407e-05,
+      "loss": 0.0052,
+      "step": 15325
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.792086414467636e-05,
+      "loss": 0.5699,
+      "step": 15326
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.790306368146792e-05,
+      "loss": 0.7614,
+      "step": 15327
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.78852640525137e-05,
+      "loss": 0.4517,
+      "step": 15328
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.786746525809873e-05,
+      "loss": 0.8245,
+      "step": 15329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.784966729850815e-05,
+      "loss": 0.47,
+      "step": 15330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.783187017402682e-05,
+      "loss": 0.2947,
+      "step": 15331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.781407388493984e-05,
+      "loss": 0.0085,
+      "step": 15332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.779627843153209e-05,
+      "loss": 0.4856,
+      "step": 15333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.77784838140886e-05,
+      "loss": 0.3831,
+      "step": 15334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.776069003289438e-05,
+      "loss": 0.2457,
+      "step": 15335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.774289708823426e-05,
+      "loss": 0.2393,
+      "step": 15336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.77251049803932e-05,
+      "loss": 0.4078,
+      "step": 15337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.770731370965617e-05,
+      "loss": 0.2276,
+      "step": 15338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.7689523276308e-05,
+      "loss": 0.3288,
+      "step": 15339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.76717336806336e-05,
+      "loss": 0.3498,
+      "step": 15340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.765394492291789e-05,
+      "loss": 0.3329,
+      "step": 15341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.763615700344568e-05,
+      "loss": 0.0016,
+      "step": 15342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.761836992250185e-05,
+      "loss": 0.9934,
+      "step": 15343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.760058368037115e-05,
+      "loss": 0.5988,
+      "step": 15344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.75827982773385e-05,
+      "loss": 0.4473,
+      "step": 15345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.756501371368869e-05,
+      "loss": 0.0032,
+      "step": 15346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.754722998970643e-05,
+      "loss": 0.0346,
+      "step": 15347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.752944710567659e-05,
+      "loss": 0.2942,
+      "step": 15348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.751166506188395e-05,
+      "loss": 0.0014,
+      "step": 15349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.749388385861315e-05,
+      "loss": 1.3167,
+      "step": 15350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.747610349614911e-05,
+      "loss": 0.1947,
+      "step": 15351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.745832397477631e-05,
+      "loss": 0.011,
+      "step": 15352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.744054529477967e-05,
+      "loss": 0.003,
+      "step": 15353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.742276745644383e-05,
+      "loss": 0.3294,
+      "step": 15354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.740499046005339e-05,
+      "loss": 0.715,
+      "step": 15355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.738721430589313e-05,
+      "loss": 0.0895,
+      "step": 15356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.736943899424773e-05,
+      "loss": 0.2924,
+      "step": 15357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.735166452540168e-05,
+      "loss": 0.64,
+      "step": 15358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.733389089963976e-05,
+      "loss": 0.625,
+      "step": 15359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.731611811724655e-05,
+      "loss": 0.2106,
+      "step": 15360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.72983461785066e-05,
+      "loss": 0.0271,
+      "step": 15361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.728057508370459e-05,
+      "loss": 1.7689,
+      "step": 15362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.726280483312504e-05,
+      "loss": 4.4756,
+      "step": 15363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.724503542705247e-05,
+      "loss": 0.2874,
+      "step": 15364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.722726686577154e-05,
+      "loss": 0.9713,
+      "step": 15365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.720949914956672e-05,
+      "loss": 0.266,
+      "step": 15366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.71917322787225e-05,
+      "loss": 1.3026,
+      "step": 15367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.71739662535235e-05,
+      "loss": 0.3638,
+      "step": 15368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.71562010742541e-05,
+      "loss": 2.3086,
+      "step": 15369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.713843674119883e-05,
+      "loss": 0.8681,
+      "step": 15370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.712067325464224e-05,
+      "loss": 0.0038,
+      "step": 15371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.710291061486862e-05,
+      "loss": 0.0008,
+      "step": 15372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.708514882216256e-05,
+      "loss": 0.3173,
+      "step": 15373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.706738787680837e-05,
+      "loss": 0.3926,
+      "step": 15374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.704962777909054e-05,
+      "loss": 1.1298,
+      "step": 15375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.703186852929349e-05,
+      "loss": 0.1709,
+      "step": 15376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.701411012770153e-05,
+      "loss": 4.4463,
+      "step": 15377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.699635257459905e-05,
+      "loss": 0.3857,
+      "step": 15378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.697859587027052e-05,
+      "loss": 0.0226,
+      "step": 15379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.696084001500015e-05,
+      "loss": 1.3316,
+      "step": 15380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.694308500907237e-05,
+      "loss": 1.1218,
+      "step": 15381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.692533085277139e-05,
+      "loss": 0.6548,
+      "step": 15382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.690757754638161e-05,
+      "loss": 0.0245,
+      "step": 15383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.688982509018731e-05,
+      "loss": 0.4967,
+      "step": 15384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.687207348447272e-05,
+      "loss": 0.0024,
+      "step": 15385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.685432272952215e-05,
+      "loss": 0.6854,
+      "step": 15386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.68365728256199e-05,
+      "loss": 0.0054,
+      "step": 15387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.681882377305008e-05,
+      "loss": 0.3443,
+      "step": 15388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.680107557209697e-05,
+      "loss": 0.0592,
+      "step": 15389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.678332822304486e-05,
+      "loss": 0.0044,
+      "step": 15390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.676558172617785e-05,
+      "loss": 0.6814,
+      "step": 15391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.67478360817802e-05,
+      "loss": 0.0543,
+      "step": 15392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.673009129013595e-05,
+      "loss": 0.2947,
+      "step": 15393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.671234735152938e-05,
+      "loss": 0.6675,
+      "step": 15394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.669460426624463e-05,
+      "loss": 0.3862,
+      "step": 15395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.667686203456573e-05,
+      "loss": 0.002,
+      "step": 15396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.665912065677687e-05,
+      "loss": 0.4902,
+      "step": 15397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.66413801331622e-05,
+      "loss": 0.0016,
+      "step": 15398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.662364046400566e-05,
+      "loss": 0.2793,
+      "step": 15399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.660590164959148e-05,
+      "loss": 0.4334,
+      "step": 15400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.658816369020362e-05,
+      "loss": 0.5245,
+      "step": 15401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.657042658612614e-05,
+      "loss": 1.1411,
+      "step": 15402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.655269033764318e-05,
+      "loss": 0.627,
+      "step": 15403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.653495494503855e-05,
+      "loss": 0.2255,
+      "step": 15404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.651722040859644e-05,
+      "loss": 3.679,
+      "step": 15405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.649948672860081e-05,
+      "loss": 3.3494,
+      "step": 15406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.648175390533554e-05,
+      "loss": 0.2574,
+      "step": 15407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.646402193908468e-05,
+      "loss": 0.8919,
+      "step": 15408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.644629083013221e-05,
+      "loss": 0.21,
+      "step": 15409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.642856057876194e-05,
+      "loss": 0.5073,
+      "step": 15410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.641083118525796e-05,
+      "loss": 0.1333,
+      "step": 15411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.639310264990404e-05,
+      "loss": 0.2649,
+      "step": 15412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.63753749729841e-05,
+      "loss": 2.3438,
+      "step": 15413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.635764815478211e-05,
+      "loss": 0.6719,
+      "step": 15414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.633992219558186e-05,
+      "loss": 0.0029,
+      "step": 15415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.632219709566717e-05,
+      "loss": 0.2671,
+      "step": 15416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.630447285532203e-05,
+      "loss": 0.2141,
+      "step": 15417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.628674947483012e-05,
+      "loss": 0.0009,
+      "step": 15418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.626902695447528e-05,
+      "loss": 0.4615,
+      "step": 15419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.625130529454142e-05,
+      "loss": 0.0052,
+      "step": 15420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.623358449531216e-05,
+      "loss": 0.517,
+      "step": 15421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.62158645570714e-05,
+      "loss": 0.4657,
+      "step": 15422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.619814548010282e-05,
+      "loss": 0.0113,
+      "step": 15423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.618042726469019e-05,
+      "loss": 0.3997,
+      "step": 15424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.616270991111727e-05,
+      "loss": 0.2425,
+      "step": 15425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.614499341966776e-05,
+      "loss": 0.9892,
+      "step": 15426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.612727779062528e-05,
+      "loss": 0.3034,
+      "step": 15427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.610956302427367e-05,
+      "loss": 0.0029,
+      "step": 15428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.609184912089652e-05,
+      "loss": 2.4805,
+      "step": 15429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.607413608077751e-05,
+      "loss": 0.1607,
+      "step": 15430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.605642390420024e-05,
+      "loss": 0.0158,
+      "step": 15431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.60387125914484e-05,
+      "loss": 0.3733,
+      "step": 15432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.602100214280561e-05,
+      "loss": 0.7686,
+      "step": 15433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.600329255855543e-05,
+      "loss": 0.2677,
+      "step": 15434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.598558383898151e-05,
+      "loss": 0.0013,
+      "step": 15435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.59678759843674e-05,
+      "loss": 0.0087,
+      "step": 15436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.595016899499661e-05,
+      "loss": 0.529,
+      "step": 15437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.59324628711528e-05,
+      "loss": 0.458,
+      "step": 15438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.591475761311949e-05,
+      "loss": 0.0015,
+      "step": 15439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.58970532211801e-05,
+      "loss": 0.1685,
+      "step": 15440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.587934969561824e-05,
+      "loss": 0.0067,
+      "step": 15441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.586164703671733e-05,
+      "loss": 0.6797,
+      "step": 15442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.584394524476093e-05,
+      "loss": 0.3326,
+      "step": 15443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.582624432003248e-05,
+      "loss": 0.3871,
+      "step": 15444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.580854426281537e-05,
+      "loss": 0.3823,
+      "step": 15445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.579084507339313e-05,
+      "loss": 0.2669,
+      "step": 15446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.577314675204917e-05,
+      "loss": 0.6558,
+      "step": 15447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.575544929906682e-05,
+      "loss": 0.2978,
+      "step": 15448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.573775271472961e-05,
+      "loss": 0.013,
+      "step": 15449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.572005699932082e-05,
+      "loss": 0.2032,
+      "step": 15450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.570236215312381e-05,
+      "loss": 0.0057,
+      "step": 15451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.568466817642207e-05,
+      "loss": 0.2792,
+      "step": 15452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.568466817642207e-05,
+      "loss": 0.1148,
+      "step": 15453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.566697506949879e-05,
+      "loss": 0.373,
+      "step": 15454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.564928283263737e-05,
+      "loss": 0.1711,
+      "step": 15455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.563159146612119e-05,
+      "loss": 0.0015,
+      "step": 15456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.561390097023338e-05,
+      "loss": 0.4272,
+      "step": 15457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.559621134525736e-05,
+      "loss": 0.0245,
+      "step": 15458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.557852259147638e-05,
+      "loss": 0.2975,
+      "step": 15459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.556083470917363e-05,
+      "loss": 0.4705,
+      "step": 15460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.55431476986325e-05,
+      "loss": 2.8106,
+      "step": 15461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.552546156013608e-05,
+      "loss": 0.5559,
+      "step": 15462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.55077762939676e-05,
+      "loss": 0.0512,
+      "step": 15463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.549009190041039e-05,
+      "loss": 1.0986,
+      "step": 15464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.547240837974748e-05,
+      "loss": 0.3238,
+      "step": 15465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.545472573226211e-05,
+      "loss": 0.0109,
+      "step": 15466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.54370439582375e-05,
+      "loss": 0.0082,
+      "step": 15467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.541936305795673e-05,
+      "loss": 0.2236,
+      "step": 15468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.540168303170288e-05,
+      "loss": 0.0136,
+      "step": 15469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.538400387975922e-05,
+      "loss": 0.6708,
+      "step": 15470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.536632560240876e-05,
+      "loss": 0.2752,
+      "step": 15471
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.53486481999346e-05,
+      "loss": 1.2382,
+      "step": 15472
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.53309716726198e-05,
+      "loss": 0.475,
+      "step": 15473
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.531329602074742e-05,
+      "loss": 0.6948,
+      "step": 15474
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.529562124460058e-05,
+      "loss": 0.2063,
+      "step": 15475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.527794734446222e-05,
+      "loss": 0.239,
+      "step": 15476
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.526027432061541e-05,
+      "loss": 0.898,
+      "step": 15477
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.524260217334319e-05,
+      "loss": 0.4331,
+      "step": 15478
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.52249309029285e-05,
+      "loss": 0.5747,
+      "step": 15479
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.520726050965436e-05,
+      "loss": 0.011,
+      "step": 15480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.518959099380366e-05,
+      "loss": 0.1846,
+      "step": 15481
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.517192235565942e-05,
+      "loss": 0.1986,
+      "step": 15482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.51542545955046e-05,
+      "loss": 1.9385,
+      "step": 15483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.513658771362201e-05,
+      "loss": 0.3029,
+      "step": 15484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.511892171029464e-05,
+      "loss": 0.2963,
+      "step": 15485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.510125658580546e-05,
+      "loss": 0.5938,
+      "step": 15486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.508359234043715e-05,
+      "loss": 0.3369,
+      "step": 15487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.506592897447274e-05,
+      "loss": 0.2919,
+      "step": 15488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.504826648819508e-05,
+      "loss": 0.6029,
+      "step": 15489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.503060488188689e-05,
+      "loss": 0.8433,
+      "step": 15490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.501294415583113e-05,
+      "loss": 0.3251,
+      "step": 15491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.499528431031048e-05,
+      "loss": 0.6429,
+      "step": 15492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.497762534560784e-05,
+      "loss": 0.005,
+      "step": 15493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.495996726200595e-05,
+      "loss": 0.2988,
+      "step": 15494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.494231005978755e-05,
+      "loss": 0.0006,
+      "step": 15495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.492465373923544e-05,
+      "loss": 0.0017,
+      "step": 15496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.490699830063238e-05,
+      "loss": 0.3863,
+      "step": 15497
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.4889343744261e-05,
+      "loss": 0.0008,
+      "step": 15498
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.487169007040413e-05,
+      "loss": 0.0062,
+      "step": 15499
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.485403727934438e-05,
+      "loss": 0.0015,
+      "step": 15500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.483638537136442e-05,
+      "loss": 0.1706,
+      "step": 15501
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.481873434674703e-05,
+      "loss": 1.6369,
+      "step": 15502
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.480108420577475e-05,
+      "loss": 0.0034,
+      "step": 15503
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.478343494873025e-05,
+      "loss": 0.0062,
+      "step": 15504
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.476578657589626e-05,
+      "loss": 0.1889,
+      "step": 15505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.47481390875552e-05,
+      "loss": 0.6143,
+      "step": 15506
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.473049248398981e-05,
+      "loss": 0.5526,
+      "step": 15507
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.471284676548267e-05,
+      "loss": 0.6245,
+      "step": 15508
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.469520193231623e-05,
+      "loss": 0.0981,
+      "step": 15509
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.467755798477323e-05,
+      "loss": 0.2495,
+      "step": 15510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.465991492313605e-05,
+      "loss": 1.0844,
+      "step": 15511
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.464227274768726e-05,
+      "loss": 0.1227,
+      "step": 15512
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.462463145870944e-05,
+      "loss": 0.8985,
+      "step": 15513
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.460699105648503e-05,
+      "loss": 0.1733,
+      "step": 15514
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.458935154129648e-05,
+      "loss": 0.3852,
+      "step": 15515
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.457171291342637e-05,
+      "loss": 0.4508,
+      "step": 15516
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.455407517315706e-05,
+      "loss": 0.7988,
+      "step": 15517
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.453643832077101e-05,
+      "loss": 1.0417,
+      "step": 15518
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.45188023565507e-05,
+      "loss": 1.5595,
+      "step": 15519
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.450116728077847e-05,
+      "loss": 0.4689,
+      "step": 15520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.448353309373682e-05,
+      "loss": 0.4226,
+      "step": 15521
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.446589979570799e-05,
+      "loss": 0.545,
+      "step": 15522
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.444826738697443e-05,
+      "loss": 0.8388,
+      "step": 15523
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.443063586781857e-05,
+      "loss": 0.3889,
+      "step": 15524
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.441300523852262e-05,
+      "loss": 0.3873,
+      "step": 15525
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.439537549936895e-05,
+      "loss": 0.2148,
+      "step": 15526
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.437774665063995e-05,
+      "loss": 0.0034,
+      "step": 15527
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.436011869261783e-05,
+      "loss": 0.1516,
+      "step": 15528
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.434249162558491e-05,
+      "loss": 0.0018,
+      "step": 15529
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.432486544982343e-05,
+      "loss": 0.0017,
+      "step": 15530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.430724016561571e-05,
+      "loss": 0.3442,
+      "step": 15531
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.428961577324395e-05,
+      "loss": 0.3229,
+      "step": 15532
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.427199227299034e-05,
+      "loss": 0.9258,
+      "step": 15533
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.425436966513716e-05,
+      "loss": 0.6367,
+      "step": 15534
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.423674794996663e-05,
+      "loss": 0.0006,
+      "step": 15535
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.421912712776083e-05,
+      "loss": 0.2267,
+      "step": 15536
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.420150719880204e-05,
+      "loss": 0.3122,
+      "step": 15537
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.418388816337237e-05,
+      "loss": 1.2944,
+      "step": 15538
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.41662700217539e-05,
+      "loss": 0.0012,
+      "step": 15539
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.414865277422894e-05,
+      "loss": 0.3506,
+      "step": 15540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.413103642107939e-05,
+      "loss": 0.6428,
+      "step": 15541
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.411342096258746e-05,
+      "loss": 0.0037,
+      "step": 15542
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.409580639903526e-05,
+      "loss": 1.5904,
+      "step": 15543
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.407819273070476e-05,
+      "loss": 0.3951,
+      "step": 15544
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.40605799578781e-05,
+      "loss": 0.8326,
+      "step": 15545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.404296808083733e-05,
+      "loss": 0.2793,
+      "step": 15546
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.402535709986438e-05,
+      "loss": 0.8494,
+      "step": 15547
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.400774701524141e-05,
+      "loss": 0.3992,
+      "step": 15548
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.399013782725028e-05,
+      "loss": 0.5394,
+      "step": 15549
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.397252953617302e-05,
+      "loss": 0.7135,
+      "step": 15550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.395492214229167e-05,
+      "loss": 0.002,
+      "step": 15551
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.393731564588811e-05,
+      "loss": 0.2049,
+      "step": 15552
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.391971004724427e-05,
+      "loss": 0.4859,
+      "step": 15553
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.390210534664216e-05,
+      "loss": 0.1561,
+      "step": 15554
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.388450154436361e-05,
+      "loss": 0.3063,
+      "step": 15555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.386689864069053e-05,
+      "loss": 0.3406,
+      "step": 15556
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.384929663590492e-05,
+      "loss": 0.458,
+      "step": 15557
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.383169553028845e-05,
+      "loss": 0.3927,
+      "step": 15558
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.381409532412314e-05,
+      "loss": 0.5851,
+      "step": 15559
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.379649601769074e-05,
+      "loss": 0.6687,
+      "step": 15560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.377889761127308e-05,
+      "loss": 0.0082,
+      "step": 15561
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.376130010515206e-05,
+      "loss": 0.2444,
+      "step": 15562
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.374370349960938e-05,
+      "loss": 0.6697,
+      "step": 15563
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.372610779492684e-05,
+      "loss": 0.2811,
+      "step": 15564
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.370851299138629e-05,
+      "loss": 1.9981,
+      "step": 15565
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.369091908926938e-05,
+      "loss": 0.2071,
+      "step": 15566
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.367332608885787e-05,
+      "loss": 0.3365,
+      "step": 15567
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.365573399043357e-05,
+      "loss": 2.043,
+      "step": 15568
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.36381427942781e-05,
+      "loss": 0.2541,
+      "step": 15569
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.362055250067318e-05,
+      "loss": 0.444,
+      "step": 15570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.360296310990048e-05,
+      "loss": 0.2897,
+      "step": 15571
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.358537462224171e-05,
+      "loss": 0.3343,
+      "step": 15572
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.356778703797851e-05,
+      "loss": 0.0018,
+      "step": 15573
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.355020035739251e-05,
+      "loss": 0.6702,
+      "step": 15574
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.353261458076528e-05,
+      "loss": 3.311,
+      "step": 15575
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.351502970837854e-05,
+      "loss": 0.7852,
+      "step": 15576
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.34974457405138e-05,
+      "loss": 0.0166,
+      "step": 15577
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.347986267745272e-05,
+      "loss": 0.6502,
+      "step": 15578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.346228051947671e-05,
+      "loss": 0.3564,
+      "step": 15579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.34446992668675e-05,
+      "loss": 0.3903,
+      "step": 15580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.342711891990657e-05,
+      "loss": 0.8586,
+      "step": 15581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.340953947887536e-05,
+      "loss": 0.5036,
+      "step": 15582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.339196094405546e-05,
+      "loss": 0.3703,
+      "step": 15583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.33743833157284e-05,
+      "loss": 0.3599,
+      "step": 15584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.335680659417554e-05,
+      "loss": 0.1839,
+      "step": 15585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.333923077967844e-05,
+      "loss": 0.0099,
+      "step": 15586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.332165587251858e-05,
+      "loss": 0.3619,
+      "step": 15587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.330408187297724e-05,
+      "loss": 0.0292,
+      "step": 15588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.328650878133603e-05,
+      "loss": 4.6344,
+      "step": 15589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.326893659787622e-05,
+      "loss": 0.7481,
+      "step": 15590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.325136532287927e-05,
+      "loss": 0.2684,
+      "step": 15591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.323379495662653e-05,
+      "loss": 0.6134,
+      "step": 15592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.321622549939934e-05,
+      "loss": 0.4338,
+      "step": 15593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.31986569514791e-05,
+      "loss": 0.2117,
+      "step": 15594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.318108931314714e-05,
+      "loss": 0.4612,
+      "step": 15595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.316352258468472e-05,
+      "loss": 0.0009,
+      "step": 15596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.314595676637323e-05,
+      "loss": 0.4642,
+      "step": 15597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.31283918584939e-05,
+      "loss": 0.374,
+      "step": 15598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.311082786132799e-05,
+      "loss": 0.7811,
+      "step": 15599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.309326477515685e-05,
+      "loss": 0.4856,
+      "step": 15600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.307570260026163e-05,
+      "loss": 0.256,
+      "step": 15601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.305814133692359e-05,
+      "loss": 0.1736,
+      "step": 15602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.304058098542401e-05,
+      "loss": 0.4136,
+      "step": 15603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.302302154604401e-05,
+      "loss": 0.0074,
+      "step": 15604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.300546301906476e-05,
+      "loss": 0.4563,
+      "step": 15605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.298790540476758e-05,
+      "loss": 0.0021,
+      "step": 15606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.297034870343348e-05,
+      "loss": 0.0031,
+      "step": 15607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.29527929153437e-05,
+      "loss": 0.4284,
+      "step": 15608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.293523804077927e-05,
+      "loss": 0.0014,
+      "step": 15609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.291768408002135e-05,
+      "loss": 0.177,
+      "step": 15610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.290013103335112e-05,
+      "loss": 0.9066,
+      "step": 15611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.288257890104954e-05,
+      "loss": 0.7207,
+      "step": 15612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.286502768339773e-05,
+      "loss": 2.5586,
+      "step": 15613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.28474773806768e-05,
+      "loss": 0.7174,
+      "step": 15614
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.282992799316773e-05,
+      "loss": 0.3896,
+      "step": 15615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.281237952115152e-05,
+      "loss": 0.0141,
+      "step": 15616
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.279483196490928e-05,
+      "loss": 0.0768,
+      "step": 15617
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.277728532472196e-05,
+      "loss": 0.0155,
+      "step": 15618
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.275973960087055e-05,
+      "loss": 0.4396,
+      "step": 15619
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.274219479363591e-05,
+      "loss": 0.01,
+      "step": 15620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.272465090329916e-05,
+      "loss": 0.6488,
+      "step": 15621
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.270710793014118e-05,
+      "loss": 0.2423,
+      "step": 15622
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.268956587444281e-05,
+      "loss": 0.7372,
+      "step": 15623
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.267202473648509e-05,
+      "loss": 0.2406,
+      "step": 15624
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.265448451654885e-05,
+      "loss": 0.8213,
+      "step": 15625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.263694521491496e-05,
+      "loss": 0.6646,
+      "step": 15626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.261940683186434e-05,
+      "loss": 0.4587,
+      "step": 15627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.260186936767776e-05,
+      "loss": 1.2059,
+      "step": 15628
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.25843328226361e-05,
+      "loss": 1.1289,
+      "step": 15629
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.256679719702022e-05,
+      "loss": 0.3076,
+      "step": 15630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.254926249111084e-05,
+      "loss": 0.2671,
+      "step": 15631
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.253172870518883e-05,
+      "loss": 0.0352,
+      "step": 15632
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.251419583953494e-05,
+      "loss": 0.0037,
+      "step": 15633
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.24966638944299e-05,
+      "loss": 0.4015,
+      "step": 15634
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.24791328701545e-05,
+      "loss": 0.1499,
+      "step": 15635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.24616027669895e-05,
+      "loss": 0.5165,
+      "step": 15636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.244407358521551e-05,
+      "loss": 0.3005,
+      "step": 15637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.242654532511338e-05,
+      "loss": 0.1651,
+      "step": 15638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.240901798696367e-05,
+      "loss": 0.2637,
+      "step": 15639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.239149157104708e-05,
+      "loss": 0.7488,
+      "step": 15640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.237396607764437e-05,
+      "loss": 0.792,
+      "step": 15641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.235644150703604e-05,
+      "loss": 0.2779,
+      "step": 15642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.233891785950284e-05,
+      "loss": 0.028,
+      "step": 15643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.232139513532531e-05,
+      "loss": 0.1527,
+      "step": 15644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.230387333478406e-05,
+      "loss": 0.5313,
+      "step": 15645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.228635245815972e-05,
+      "loss": 3.406,
+      "step": 15646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.226883250573281e-05,
+      "loss": 0.18,
+      "step": 15647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.225131347778386e-05,
+      "loss": 0.2644,
+      "step": 15648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.223379537459354e-05,
+      "loss": 0.2008,
+      "step": 15649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.221627819644225e-05,
+      "loss": 1.8947,
+      "step": 15650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.219876194361053e-05,
+      "loss": 0.2197,
+      "step": 15651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.218124661637892e-05,
+      "loss": 2.1582,
+      "step": 15652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.216373221502786e-05,
+      "loss": 0.706,
+      "step": 15653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.214621873983779e-05,
+      "loss": 0.0529,
+      "step": 15654
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.212870619108926e-05,
+      "loss": 0.0279,
+      "step": 15655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.211119456906264e-05,
+      "loss": 0.5206,
+      "step": 15656
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.209368387403836e-05,
+      "loss": 0.5999,
+      "step": 15657
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.207617410629683e-05,
+      "loss": 0.7958,
+      "step": 15658
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.205866526611839e-05,
+      "loss": 0.3093,
+      "step": 15659
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.204115735378352e-05,
+      "loss": 0.5417,
+      "step": 15660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.20236503695725e-05,
+      "loss": 0.3648,
+      "step": 15661
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.200614431376569e-05,
+      "loss": 0.3812,
+      "step": 15662
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.198863918664352e-05,
+      "loss": 0.2219,
+      "step": 15663
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.197113498848619e-05,
+      "loss": 0.2653,
+      "step": 15664
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.195363171957406e-05,
+      "loss": 0.3597,
+      "step": 15665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.193612938018733e-05,
+      "loss": 0.0102,
+      "step": 15666
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.191862797060639e-05,
+      "loss": 1.1962,
+      "step": 15667
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.19011274911115e-05,
+      "loss": 0.2185,
+      "step": 15668
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.18836279419828e-05,
+      "loss": 0.0094,
+      "step": 15669
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.186612932350058e-05,
+      "loss": 0.195,
+      "step": 15670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.18486316359451e-05,
+      "loss": 0.3474,
+      "step": 15671
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.183113487959647e-05,
+      "loss": 0.7472,
+      "step": 15672
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.181363905473492e-05,
+      "loss": 0.1007,
+      "step": 15673
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.179614416164063e-05,
+      "loss": 0.0077,
+      "step": 15674
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.177865020059371e-05,
+      "loss": 0.4092,
+      "step": 15675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.176115717187441e-05,
+      "loss": 0.6584,
+      "step": 15676
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.17436650757627e-05,
+      "loss": 0.6104,
+      "step": 15677
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.172617391253876e-05,
+      "loss": 0.012,
+      "step": 15678
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.170868368248273e-05,
+      "loss": 0.262,
+      "step": 15679
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.169119438587461e-05,
+      "loss": 0.5194,
+      "step": 15680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.16737060229945e-05,
+      "loss": 0.3622,
+      "step": 15681
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.16562185941225e-05,
+      "loss": 0.0055,
+      "step": 15682
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.163873209953854e-05,
+      "loss": 1.8155,
+      "step": 15683
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.162124653952273e-05,
+      "loss": 0.4621,
+      "step": 15684
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.160376191435505e-05,
+      "loss": 0.3766,
+      "step": 15685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.158627822431545e-05,
+      "loss": 0.1395,
+      "step": 15686
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.156879546968398e-05,
+      "loss": 0.2851,
+      "step": 15687
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.155131365074054e-05,
+      "loss": 0.7921,
+      "step": 15688
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.153383276776506e-05,
+      "loss": 0.2492,
+      "step": 15689
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.151635282103755e-05,
+      "loss": 0.0118,
+      "step": 15690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.149887381083785e-05,
+      "loss": 0.0173,
+      "step": 15691
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.148139573744587e-05,
+      "loss": 0.4929,
+      "step": 15692
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.146391860114158e-05,
+      "loss": 0.0946,
+      "step": 15693
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.144644240220472e-05,
+      "loss": 0.7152,
+      "step": 15694
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.142896714091525e-05,
+      "loss": 1.3918,
+      "step": 15695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.141149281755296e-05,
+      "loss": 0.2071,
+      "step": 15696
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.139401943239764e-05,
+      "loss": 0.0707,
+      "step": 15697
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.13765469857292e-05,
+      "loss": 0.369,
+      "step": 15698
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.135907547782737e-05,
+      "loss": 0.5833,
+      "step": 15699
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.134160490897188e-05,
+      "loss": 0.6248,
+      "step": 15700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.132413527944263e-05,
+      "loss": 0.2651,
+      "step": 15701
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.130666658951926e-05,
+      "loss": 0.4623,
+      "step": 15702
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.12891988394815e-05,
+      "loss": 0.0052,
+      "step": 15703
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.127173202960917e-05,
+      "loss": 0.1829,
+      "step": 15704
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.125426616018188e-05,
+      "loss": 0.6959,
+      "step": 15705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.12368012314794e-05,
+      "loss": 0.4633,
+      "step": 15706
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.12193372437813e-05,
+      "loss": 0.5181,
+      "step": 15707
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.120187419736731e-05,
+      "loss": 1.7785,
+      "step": 15708
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.118441209251708e-05,
+      "loss": 0.0153,
+      "step": 15709
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.116695092951021e-05,
+      "loss": 0.0063,
+      "step": 15710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.114949070862628e-05,
+      "loss": 0.3209,
+      "step": 15711
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.113203143014502e-05,
+      "loss": 0.9059,
+      "step": 15712
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.111457309434587e-05,
+      "loss": 0.1522,
+      "step": 15713
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.10971157015085e-05,
+      "loss": 0.005,
+      "step": 15714
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.107965925191238e-05,
+      "loss": 0.9615,
+      "step": 15715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.106220374583709e-05,
+      "loss": 0.4066,
+      "step": 15716
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.104474918356218e-05,
+      "loss": 0.2067,
+      "step": 15717
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.102729556536708e-05,
+      "loss": 0.2113,
+      "step": 15718
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.100984289153139e-05,
+      "loss": 0.0027,
+      "step": 15719
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.099239116233454e-05,
+      "loss": 0.4258,
+      "step": 15720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.097494037805596e-05,
+      "loss": 0.9131,
+      "step": 15721
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.095749053897513e-05,
+      "loss": 0.8179,
+      "step": 15722
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.094004164537148e-05,
+      "loss": 1.5698,
+      "step": 15723
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.092259369752439e-05,
+      "loss": 0.5279,
+      "step": 15724
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.090514669571342e-05,
+      "loss": 0.6104,
+      "step": 15725
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.08877006402177e-05,
+      "loss": 0.572,
+      "step": 15726
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.08702555313168e-05,
+      "loss": 0.3654,
+      "step": 15727
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.085281136929002e-05,
+      "loss": 1.1201,
+      "step": 15728
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.083536815441667e-05,
+      "loss": 0.6832,
+      "step": 15729
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.081792588697612e-05,
+      "loss": 0.0082,
+      "step": 15730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.08004845672477e-05,
+      "loss": 0.6847,
+      "step": 15731
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.07830441955106e-05,
+      "loss": 0.5967,
+      "step": 15732
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.076560477204423e-05,
+      "loss": 0.4667,
+      "step": 15733
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.074816629712781e-05,
+      "loss": 0.2412,
+      "step": 15734
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.073072877104052e-05,
+      "loss": 0.0149,
+      "step": 15735
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.071329219406172e-05,
+      "loss": 0.5367,
+      "step": 15736
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.069585656647054e-05,
+      "loss": 0.8008,
+      "step": 15737
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.067842188854622e-05,
+      "loss": 0.0036,
+      "step": 15738
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.066098816056797e-05,
+      "loss": 0.4152,
+      "step": 15739
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.064355538281491e-05,
+      "loss": 0.4788,
+      "step": 15740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.062612355556619e-05,
+      "loss": 0.3737,
+      "step": 15741
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.060869267910107e-05,
+      "loss": 0.2355,
+      "step": 15742
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.059126275369857e-05,
+      "loss": 0.9132,
+      "step": 15743
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.057383377963786e-05,
+      "loss": 0.3513,
+      "step": 15744
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.055640575719799e-05,
+      "loss": 0.3024,
+      "step": 15745
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.053897868665805e-05,
+      "loss": 0.6343,
+      "step": 15746
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.052155256829717e-05,
+      "loss": 0.0581,
+      "step": 15747
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.050412740239434e-05,
+      "loss": 0.5673,
+      "step": 15748
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.048670318922861e-05,
+      "loss": 0.2213,
+      "step": 15749
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.046927992907903e-05,
+      "loss": 1.0193,
+      "step": 15750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.045185762222459e-05,
+      "loss": 0.0008,
+      "step": 15751
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.043443626894422e-05,
+      "loss": 0.0066,
+      "step": 15752
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.041701586951704e-05,
+      "loss": 0.2441,
+      "step": 15753
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.03995964242219e-05,
+      "loss": 0.3199,
+      "step": 15754
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.038217793333781e-05,
+      "loss": 0.3816,
+      "step": 15755
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.036476039714361e-05,
+      "loss": 0.4829,
+      "step": 15756
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.034734381591829e-05,
+      "loss": 0.4683,
+      "step": 15757
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.032992818994076e-05,
+      "loss": 0.2139,
+      "step": 15758
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.031251351948981e-05,
+      "loss": 0.4502,
+      "step": 15759
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.029509980484443e-05,
+      "loss": 0.3364,
+      "step": 15760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.027768704628343e-05,
+      "loss": 0.4517,
+      "step": 15761
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.026027524408563e-05,
+      "loss": 0.3602,
+      "step": 15762
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.024286439852988e-05,
+      "loss": 0.5652,
+      "step": 15763
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.022545450989493e-05,
+      "loss": 0.2097,
+      "step": 15764
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.020804557845964e-05,
+      "loss": 0.5024,
+      "step": 15765
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.019063760450278e-05,
+      "loss": 0.008,
+      "step": 15766
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.017323058830306e-05,
+      "loss": 0.444,
+      "step": 15767
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.015582453013928e-05,
+      "loss": 0.3753,
+      "step": 15768
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.01384194302902e-05,
+      "loss": 0.5595,
+      "step": 15769
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.012101528903444e-05,
+      "loss": 2.918,
+      "step": 15770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.010361210665074e-05,
+      "loss": 0.2206,
+      "step": 15771
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.008620988341789e-05,
+      "loss": 0.5521,
+      "step": 15772
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.006880861961437e-05,
+      "loss": 0.2338,
+      "step": 15773
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.0051408315519e-05,
+      "loss": 0.6646,
+      "step": 15774
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.003400897141034e-05,
+      "loss": 0.3206,
+      "step": 15775
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.001661058756698e-05,
+      "loss": 0.0052,
+      "step": 15776
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.999921316426771e-05,
+      "loss": 0.4119,
+      "step": 15777
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.998181670179085e-05,
+      "loss": 0.2864,
+      "step": 15778
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.996442120041518e-05,
+      "loss": 0.4864,
+      "step": 15779
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.994702666041923e-05,
+      "loss": 0.2439,
+      "step": 15780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.992963308208147e-05,
+      "loss": 0.0445,
+      "step": 15781
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.991224046568049e-05,
+      "loss": 4.127,
+      "step": 15782
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.989484881149484e-05,
+      "loss": 0.2088,
+      "step": 15783
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.987745811980292e-05,
+      "loss": 0.3549,
+      "step": 15784
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.986006839088334e-05,
+      "loss": 0.0007,
+      "step": 15785
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.984267962501448e-05,
+      "loss": 0.7439,
+      "step": 15786
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.98252918224748e-05,
+      "loss": 0.4745,
+      "step": 15787
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.98079049835428e-05,
+      "loss": 0.6451,
+      "step": 15788
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.979051910849685e-05,
+      "loss": 0.0044,
+      "step": 15789
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.977313419761534e-05,
+      "loss": 0.2001,
+      "step": 15790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.975575025117676e-05,
+      "loss": 0.1141,
+      "step": 15791
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.973836726945939e-05,
+      "loss": 0.2268,
+      "step": 15792
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.972098525274168e-05,
+      "loss": 0.5735,
+      "step": 15793
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.970360420130188e-05,
+      "loss": 0.5675,
+      "step": 15794
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.968622411541834e-05,
+      "loss": 0.3043,
+      "step": 15795
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.966884499536945e-05,
+      "loss": 0.2954,
+      "step": 15796
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.965146684143344e-05,
+      "loss": 0.2336,
+      "step": 15797
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.963408965388859e-05,
+      "loss": 0.381,
+      "step": 15798
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.961671343301326e-05,
+      "loss": 0.1615,
+      "step": 15799
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.959933817908561e-05,
+      "loss": 0.5792,
+      "step": 15800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.958196389238389e-05,
+      "loss": 0.004,
+      "step": 15801
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.956459057318638e-05,
+      "loss": 0.5352,
+      "step": 15802
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.954721822177123e-05,
+      "loss": 0.2709,
+      "step": 15803
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.952984683841671e-05,
+      "loss": 0.0008,
+      "step": 15804
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.951247642340083e-05,
+      "loss": 0.2726,
+      "step": 15805
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.949510697700193e-05,
+      "loss": 0.2451,
+      "step": 15806
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.94777384994981e-05,
+      "loss": 0.0025,
+      "step": 15807
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.946037099116742e-05,
+      "loss": 0.0217,
+      "step": 15808
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.944300445228804e-05,
+      "loss": 0.1735,
+      "step": 15809
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.942563888313809e-05,
+      "loss": 0.0807,
+      "step": 15810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.940827428399562e-05,
+      "loss": 0.8772,
+      "step": 15811
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.939091065513872e-05,
+      "loss": 0.4138,
+      "step": 15812
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.937354799684536e-05,
+      "loss": 0.0026,
+      "step": 15813
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.935618630939368e-05,
+      "loss": 0.5287,
+      "step": 15814
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.933882559306168e-05,
+      "loss": 0.408,
+      "step": 15815
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.932146584812729e-05,
+      "loss": 0.3653,
+      "step": 15816
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.930410707486861e-05,
+      "loss": 0.5567,
+      "step": 15817
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.928674927356359e-05,
+      "loss": 0.2004,
+      "step": 15818
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.926939244449009e-05,
+      "loss": 2.1123,
+      "step": 15819
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.925203658792613e-05,
+      "loss": 0.6065,
+      "step": 15820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.92346817041497e-05,
+      "loss": 0.0117,
+      "step": 15821
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.921732779343857e-05,
+      "loss": 1.2844,
+      "step": 15822
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.919997485607078e-05,
+      "loss": 0.1588,
+      "step": 15823
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.918262289232413e-05,
+      "loss": 0.3125,
+      "step": 15824
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.916527190247645e-05,
+      "loss": 0.2851,
+      "step": 15825
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.914792188680571e-05,
+      "loss": 0.0027,
+      "step": 15826
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.913057284558963e-05,
+      "loss": 0.3093,
+      "step": 15827
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 8.911322477910605e-05,
+      "loss": 0.2074,
+      "step": 15828
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.909587768763291e-05,
+      "loss": 0.2974,
+      "step": 15829
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.907853157144776e-05,
+      "loss": 0.8391,
+      "step": 15830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.906118643082856e-05,
+      "loss": 0.0005,
+      "step": 15831
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.9043842266053e-05,
+      "loss": 0.3031,
+      "step": 15832
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.902649907739879e-05,
+      "loss": 0.9937,
+      "step": 15833
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.900915686514373e-05,
+      "loss": 0.4899,
+      "step": 15834
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.899181562956548e-05,
+      "loss": 0.2451,
+      "step": 15835
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.897447537094169e-05,
+      "loss": 0.1589,
+      "step": 15836
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.895713608955016e-05,
+      "loss": 0.7669,
+      "step": 15837
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.893979778566844e-05,
+      "loss": 3.3485,
+      "step": 15838
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.892246045957421e-05,
+      "loss": 0.2601,
+      "step": 15839
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.890512411154517e-05,
+      "loss": 0.0009,
+      "step": 15840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.888778874185883e-05,
+      "loss": 0.0958,
+      "step": 15841
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.887045435079289e-05,
+      "loss": 0.4613,
+      "step": 15842
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.88531209386248e-05,
+      "loss": 0.0145,
+      "step": 15843
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.883578850563226e-05,
+      "loss": 0.0038,
+      "step": 15844
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.881845705209278e-05,
+      "loss": 0.3583,
+      "step": 15845
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.880112657828387e-05,
+      "loss": 0.6496,
+      "step": 15846
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.878379708448305e-05,
+      "loss": 0.1978,
+      "step": 15847
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.876646857096787e-05,
+      "loss": 0.29,
+      "step": 15848
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.874914103801581e-05,
+      "loss": 0.0051,
+      "step": 15849
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.873181448590427e-05,
+      "loss": 0.5398,
+      "step": 15850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.871448891491087e-05,
+      "loss": 0.4594,
+      "step": 15851
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.869716432531289e-05,
+      "loss": 0.5375,
+      "step": 15852
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.867984071738785e-05,
+      "loss": 0.1575,
+      "step": 15853
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.86625180914131e-05,
+      "loss": 0.5453,
+      "step": 15854
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.864519644766607e-05,
+      "loss": 0.1619,
+      "step": 15855
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.862787578642418e-05,
+      "loss": 0.3338,
+      "step": 15856
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.86105561079647e-05,
+      "loss": 0.2807,
+      "step": 15857
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.859323741256505e-05,
+      "loss": 0.2235,
+      "step": 15858
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.857591970050259e-05,
+      "loss": 1.4838,
+      "step": 15859
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.85586029720545e-05,
+      "loss": 0.8439,
+      "step": 15860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.854128722749829e-05,
+      "loss": 0.2642,
+      "step": 15861
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.852397246711102e-05,
+      "loss": 0.2086,
+      "step": 15862
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.850665869117011e-05,
+      "loss": 0.2566,
+      "step": 15863
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.848934589995282e-05,
+      "loss": 0.0074,
+      "step": 15864
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.847203409373626e-05,
+      "loss": 0.6058,
+      "step": 15865
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.845472327279775e-05,
+      "loss": 0.0006,
+      "step": 15866
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.843741343741453e-05,
+      "loss": 1.067,
+      "step": 15867
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.842010458786368e-05,
+      "loss": 0.0024,
+      "step": 15868
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.840279672442246e-05,
+      "loss": 0.2225,
+      "step": 15869
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.838548984736803e-05,
+      "loss": 2.3036,
+      "step": 15870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.836818395697745e-05,
+      "loss": 1.3744,
+      "step": 15871
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.835087905352799e-05,
+      "loss": 0.6053,
+      "step": 15872
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.833357513729663e-05,
+      "loss": 2.2461,
+      "step": 15873
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.831627220856049e-05,
+      "loss": 0.0029,
+      "step": 15874
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.829897026759675e-05,
+      "loss": 0.2953,
+      "step": 15875
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.828166931468235e-05,
+      "loss": 0.2349,
+      "step": 15876
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.826436935009437e-05,
+      "loss": 0.2934,
+      "step": 15877
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.824707037410995e-05,
+      "loss": 0.3887,
+      "step": 15878
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.822977238700593e-05,
+      "loss": 0.2981,
+      "step": 15879
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.821247538905943e-05,
+      "loss": 0.4286,
+      "step": 15880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.819517938054745e-05,
+      "loss": 0.189,
+      "step": 15881
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.817788436174683e-05,
+      "loss": 0.3127,
+      "step": 15882
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.81605903329347e-05,
+      "loss": 0.0617,
+      "step": 15883
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.814329729438785e-05,
+      "loss": 0.5056,
+      "step": 15884
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.812600524638325e-05,
+      "loss": 0.8967,
+      "step": 15885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.810871418919787e-05,
+      "loss": 0.3405,
+      "step": 15886
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.809142412310852e-05,
+      "loss": 0.3259,
+      "step": 15887
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.807413504839205e-05,
+      "loss": 0.4243,
+      "step": 15888
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.805684696532545e-05,
+      "loss": 0.4488,
+      "step": 15889
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.803955987418547e-05,
+      "loss": 0.5643,
+      "step": 15890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.802227377524894e-05,
+      "loss": 1.918,
+      "step": 15891
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.800498866879263e-05,
+      "loss": 0.4087,
+      "step": 15892
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.798770455509342e-05,
+      "loss": 0.6761,
+      "step": 15893
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.797042143442809e-05,
+      "loss": 0.5168,
+      "step": 15894
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.795313930707333e-05,
+      "loss": 0.3118,
+      "step": 15895
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.79358581733059e-05,
+      "loss": 0.0581,
+      "step": 15896
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.791857803340262e-05,
+      "loss": 0.4191,
+      "step": 15897
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.790129888764011e-05,
+      "loss": 0.3022,
+      "step": 15898
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.788402073629506e-05,
+      "loss": 0.2514,
+      "step": 15899
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.786674357964427e-05,
+      "loss": 0.0019,
+      "step": 15900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.784946741796428e-05,
+      "loss": 0.566,
+      "step": 15901
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.783219225153186e-05,
+      "loss": 0.5718,
+      "step": 15902
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.781491808062348e-05,
+      "loss": 0.9856,
+      "step": 15903
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.779764490551591e-05,
+      "loss": 1.9668,
+      "step": 15904
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.778037272648572e-05,
+      "loss": 0.286,
+      "step": 15905
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.776310154380945e-05,
+      "loss": 0.4042,
+      "step": 15906
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.774583135776371e-05,
+      "loss": 1.1553,
+      "step": 15907
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.77285621686251e-05,
+      "loss": 0.3317,
+      "step": 15908
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.771129397667003e-05,
+      "loss": 0.2282,
+      "step": 15909
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.76940267821752e-05,
+      "loss": 0.2687,
+      "step": 15910
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.767676058541695e-05,
+      "loss": 0.3567,
+      "step": 15911
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.765949538667184e-05,
+      "loss": 0.6672,
+      "step": 15912
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.764223118621644e-05,
+      "loss": 0.1866,
+      "step": 15913
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.762496798432703e-05,
+      "loss": 1.8741,
+      "step": 15914
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.760770578128017e-05,
+      "loss": 0.0023,
+      "step": 15915
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.759044457735229e-05,
+      "loss": 0.2549,
+      "step": 15916
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.757318437281976e-05,
+      "loss": 0.0013,
+      "step": 15917
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.755592516795892e-05,
+      "loss": 0.945,
+      "step": 15918
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.753866696304636e-05,
+      "loss": 0.3066,
+      "step": 15919
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.752140975835823e-05,
+      "loss": 0.9308,
+      "step": 15920
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.750415355417095e-05,
+      "loss": 0.0109,
+      "step": 15921
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.748689835076087e-05,
+      "loss": 0.4338,
+      "step": 15922
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.746964414840429e-05,
+      "loss": 1.6561,
+      "step": 15923
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.745239094737751e-05,
+      "loss": 0.012,
+      "step": 15924
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.743513874795684e-05,
+      "loss": 0.2568,
+      "step": 15925
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.741788755041852e-05,
+      "loss": 0.2035,
+      "step": 15926
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.740063735503879e-05,
+      "loss": 0.3972,
+      "step": 15927
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.738338816209391e-05,
+      "loss": 0.0007,
+      "step": 15928
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.73661399718601e-05,
+      "loss": 0.0031,
+      "step": 15929
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.734889278461355e-05,
+      "loss": 0.107,
+      "step": 15930
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.733164660063048e-05,
+      "loss": 0.3798,
+      "step": 15931
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.731440142018707e-05,
+      "loss": 0.6211,
+      "step": 15932
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.729715724355932e-05,
+      "loss": 0.7583,
+      "step": 15933
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.727991407102357e-05,
+      "loss": 0.6286,
+      "step": 15934
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.726267190285594e-05,
+      "loss": 0.8579,
+      "step": 15935
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.724543073933231e-05,
+      "loss": 1.3299,
+      "step": 15936
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.722819058072901e-05,
+      "loss": 0.0012,
+      "step": 15937
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.721095142732207e-05,
+      "loss": 0.2492,
+      "step": 15938
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.719371327938742e-05,
+      "loss": 0.3922,
+      "step": 15939
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.717647613720129e-05,
+      "loss": 0.2299,
+      "step": 15940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.715924000103956e-05,
+      "loss": 0.0278,
+      "step": 15941
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.714200487117822e-05,
+      "loss": 0.2559,
+      "step": 15942
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.712477074789346e-05,
+      "loss": 0.9739,
+      "step": 15943
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.710753763146108e-05,
+      "loss": 1.082,
+      "step": 15944
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.709030552215703e-05,
+      "loss": 0.479,
+      "step": 15945
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.707307442025744e-05,
+      "loss": 0.1492,
+      "step": 15946
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.705584432603807e-05,
+      "loss": 0.6562,
+      "step": 15947
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.703861523977487e-05,
+      "loss": 0.3301,
+      "step": 15948
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.702138716174378e-05,
+      "loss": 0.7078,
+      "step": 15949
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.700416009222064e-05,
+      "loss": 0.0023,
+      "step": 15950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.698693403148133e-05,
+      "loss": 2.5823,
+      "step": 15951
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.696970897980173e-05,
+      "loss": 0.0078,
+      "step": 15952
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.695248493745763e-05,
+      "loss": 0.3659,
+      "step": 15953
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.693526190472486e-05,
+      "loss": 0.2592,
+      "step": 15954
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.691803988187921e-05,
+      "loss": 1.2122,
+      "step": 15955
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.69008188691965e-05,
+      "loss": 0.3876,
+      "step": 15956
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.688359886695246e-05,
+      "loss": 0.7092,
+      "step": 15957
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.686637987542287e-05,
+      "loss": 0.3109,
+      "step": 15958
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.684916189488351e-05,
+      "loss": 0.3651,
+      "step": 15959
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.683194492560994e-05,
+      "loss": 0.0063,
+      "step": 15960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.681472896787802e-05,
+      "loss": 0.2916,
+      "step": 15961
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.679751402196345e-05,
+      "loss": 0.0065,
+      "step": 15962
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.678030008814177e-05,
+      "loss": 0.0104,
+      "step": 15963
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.676308716668864e-05,
+      "loss": 1.4204,
+      "step": 15964
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.674587525787991e-05,
+      "loss": 0.8409,
+      "step": 15965
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.672866436199099e-05,
+      "loss": 0.0068,
+      "step": 15966
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.671145447929747e-05,
+      "loss": 0.8096,
+      "step": 15967
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.669424561007517e-05,
+      "loss": 0.0483,
+      "step": 15968
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.667703775459945e-05,
+      "loss": 0.0026,
+      "step": 15969
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.665983091314592e-05,
+      "loss": 1.3571,
+      "step": 15970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.664262508599016e-05,
+      "loss": 0.77,
+      "step": 15971
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.662542027340767e-05,
+      "loss": 0.5103,
+      "step": 15972
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.660821647567396e-05,
+      "loss": 0.5235,
+      "step": 15973
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.65910136930645e-05,
+      "loss": 0.2531,
+      "step": 15974
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.657381192585482e-05,
+      "loss": 0.3845,
+      "step": 15975
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.655661117432034e-05,
+      "loss": 0.0031,
+      "step": 15976
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.653941143873653e-05,
+      "loss": 0.0023,
+      "step": 15977
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.652221271937881e-05,
+      "loss": 0.0028,
+      "step": 15978
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.650501501652258e-05,
+      "loss": 0.004,
+      "step": 15979
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.648781833044323e-05,
+      "loss": 0.0151,
+      "step": 15980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.647062266141622e-05,
+      "loss": 0.0034,
+      "step": 15981
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.645342800971672e-05,
+      "loss": 0.0139,
+      "step": 15982
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.643623437562025e-05,
+      "loss": 0.4653,
+      "step": 15983
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.641904175940216e-05,
+      "loss": 2.1871,
+      "step": 15984
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.640185016133758e-05,
+      "loss": 1.1818,
+      "step": 15985
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.638465958170199e-05,
+      "loss": 0.8733,
+      "step": 15986
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.636747002077066e-05,
+      "loss": 1.9927,
+      "step": 15987
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.635028147881866e-05,
+      "loss": 0.0339,
+      "step": 15988
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.633309395612152e-05,
+      "loss": 0.3186,
+      "step": 15989
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.631590745295425e-05,
+      "loss": 0.4134,
+      "step": 15990
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.62987219695921e-05,
+      "loss": 0.3438,
+      "step": 15991
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.628153750631041e-05,
+      "loss": 0.2083,
+      "step": 15992
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.626435406338424e-05,
+      "loss": 0.1699,
+      "step": 15993
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.624717164108871e-05,
+      "loss": 0.6075,
+      "step": 15994
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.622999023969918e-05,
+      "loss": 0.3279,
+      "step": 15995
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.621280985949057e-05,
+      "loss": 2.2872,
+      "step": 15996
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.619563050073804e-05,
+      "loss": 0.494,
+      "step": 15997
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.617845216371683e-05,
+      "loss": 0.2899,
+      "step": 15998
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.616127484870188e-05,
+      "loss": 1.4119,
+      "step": 15999
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.614409855596831e-05,
+      "loss": 0.2413,
+      "step": 16000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.612692328579116e-05,
+      "loss": 0.0011,
+      "step": 16001
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.610974903844546e-05,
+      "loss": 0.0028,
+      "step": 16002
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.609257581420627e-05,
+      "loss": 0.4805,
+      "step": 16003
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.607540361334855e-05,
+      "loss": 0.7524,
+      "step": 16004
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.605823243614728e-05,
+      "loss": 0.52,
+      "step": 16005
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.60410622828775e-05,
+      "loss": 0.0032,
+      "step": 16006
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.60238931538141e-05,
+      "loss": 0.0714,
+      "step": 16007
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.600672504923211e-05,
+      "loss": 0.0025,
+      "step": 16008
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.598955796940623e-05,
+      "loss": 0.1447,
+      "step": 16009
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.59723919146116e-05,
+      "loss": 0.0005,
+      "step": 16010
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.595522688512306e-05,
+      "loss": 0.5593,
+      "step": 16011
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.593806288121532e-05,
+      "loss": 0.2731,
+      "step": 16012
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.592089990316344e-05,
+      "loss": 0.1461,
+      "step": 16013
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.590373795124222e-05,
+      "loss": 0.044,
+      "step": 16014
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.58865770257264e-05,
+      "loss": 0.4477,
+      "step": 16015
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.586941712689075e-05,
+      "loss": 0.7077,
+      "step": 16016
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.585225825501029e-05,
+      "loss": 0.0083,
+      "step": 16017
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.583510041035955e-05,
+      "loss": 0.3538,
+      "step": 16018
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.581794359321338e-05,
+      "loss": 1.4785,
+      "step": 16019
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.580078780384653e-05,
+      "loss": 0.0125,
+      "step": 16020
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.578363304253372e-05,
+      "loss": 0.2477,
+      "step": 16021
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.576647930954963e-05,
+      "loss": 0.0846,
+      "step": 16022
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.5749326605169e-05,
+      "loss": 0.3808,
+      "step": 16023
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.573217492966645e-05,
+      "loss": 0.2859,
+      "step": 16024
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.571502428331669e-05,
+      "loss": 0.0024,
+      "step": 16025
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.569787466639433e-05,
+      "loss": 0.0034,
+      "step": 16026
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.5680726079174e-05,
+      "loss": 0.4274,
+      "step": 16027
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.56635785219303e-05,
+      "loss": 0.0008,
+      "step": 16028
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.564643199493784e-05,
+      "loss": 0.7588,
+      "step": 16029
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.562928649847127e-05,
+      "loss": 0.7768,
+      "step": 16030
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.561214203280496e-05,
+      "loss": 0.7792,
+      "step": 16031
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.55949985982136e-05,
+      "loss": 0.1254,
+      "step": 16032
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.557785619497173e-05,
+      "loss": 0.0145,
+      "step": 16033
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.556071482335371e-05,
+      "loss": 0.124,
+      "step": 16034
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.554357448363417e-05,
+      "loss": 0.0077,
+      "step": 16035
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.552643517608761e-05,
+      "loss": 0.534,
+      "step": 16036
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.550929690098832e-05,
+      "loss": 0.4217,
+      "step": 16037
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.549215965861099e-05,
+      "loss": 0.3103,
+      "step": 16038
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.547502344922983e-05,
+      "loss": 0.6181,
+      "step": 16039
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.545788827311926e-05,
+      "loss": 0.9738,
+      "step": 16040
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.544075413055388e-05,
+      "loss": 0.2698,
+      "step": 16041
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.542362102180787e-05,
+      "loss": 0.0507,
+      "step": 16042
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.540648894715559e-05,
+      "loss": 0.4438,
+      "step": 16043
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.538935790687158e-05,
+      "loss": 0.3776,
+      "step": 16044
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.537222790122999e-05,
+      "loss": 0.0018,
+      "step": 16045
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.535509893050508e-05,
+      "loss": 0.5681,
+      "step": 16046
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.533797099497137e-05,
+      "loss": 0.7514,
+      "step": 16047
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.532084409490296e-05,
+      "loss": 0.3691,
+      "step": 16048
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.530371823057417e-05,
+      "loss": 0.4904,
+      "step": 16049
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.528659340225922e-05,
+      "loss": 0.2623,
+      "step": 16050
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.526946961023239e-05,
+      "loss": 0.0036,
+      "step": 16051
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.525234685476781e-05,
+      "loss": 0.5972,
+      "step": 16052
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.523522513613976e-05,
+      "loss": 0.7984,
+      "step": 16053
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.521810445462236e-05,
+      "loss": 0.603,
+      "step": 16054
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.520098481048979e-05,
+      "loss": 1.1696,
+      "step": 16055
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.51838662040162e-05,
+      "loss": 0.0014,
+      "step": 16056
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.516674863547579e-05,
+      "loss": 0.7769,
+      "step": 16057
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.514963210514247e-05,
+      "loss": 0.0025,
+      "step": 16058
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.513251661329054e-05,
+      "loss": 0.4363,
+      "step": 16059
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.511540216019406e-05,
+      "loss": 0.7009,
+      "step": 16060
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.509828874612691e-05,
+      "loss": 1.0438,
+      "step": 16061
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.508117637136332e-05,
+      "loss": 1.5527,
+      "step": 16062
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.506406503617731e-05,
+      "loss": 0.0381,
+      "step": 16063
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.504695474084274e-05,
+      "loss": 0.604,
+      "step": 16064
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.502984548563376e-05,
+      "loss": 0.4623,
+      "step": 16065
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.501273727082437e-05,
+      "loss": 0.2979,
+      "step": 16066
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.499563009668839e-05,
+      "loss": 0.5255,
+      "step": 16067
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.497852396349984e-05,
+      "loss": 0.3802,
+      "step": 16068
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.496141887153265e-05,
+      "loss": 0.0046,
+      "step": 16069
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.49443148210607e-05,
+      "loss": 0.2671,
+      "step": 16070
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.492721181235794e-05,
+      "loss": 0.8054,
+      "step": 16071
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.491010984569824e-05,
+      "loss": 0.2399,
+      "step": 16072
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.489300892135541e-05,
+      "loss": 0.2701,
+      "step": 16073
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.487590903960333e-05,
+      "loss": 0.0008,
+      "step": 16074
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.485881020071585e-05,
+      "loss": 0.8655,
+      "step": 16075
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.484171240496676e-05,
+      "loss": 0.2265,
+      "step": 16076
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 8.482461565262984e-05,
+      "loss": 0.001,
+      "step": 16077
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.480751994397889e-05,
+      "loss": 0.6077,
+      "step": 16078
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.479042527928768e-05,
+      "loss": 0.5984,
+      "step": 16079
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.477333165882994e-05,
+      "loss": 0.0008,
+      "step": 16080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.475623908287939e-05,
+      "loss": 1.343,
+      "step": 16081
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.473914755170983e-05,
+      "loss": 0.9304,
+      "step": 16082
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.472205706559476e-05,
+      "loss": 0.0023,
+      "step": 16083
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.470496762480803e-05,
+      "loss": 0.4583,
+      "step": 16084
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.46878792296233e-05,
+      "loss": 0.0009,
+      "step": 16085
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.467079188031405e-05,
+      "loss": 0.9588,
+      "step": 16086
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.465370557715415e-05,
+      "loss": 0.6414,
+      "step": 16087
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.4636620320417e-05,
+      "loss": 1.866,
+      "step": 16088
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.461953611037625e-05,
+      "loss": 0.1974,
+      "step": 16089
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.460245294730561e-05,
+      "loss": 0.2825,
+      "step": 16090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.45853708314785e-05,
+      "loss": 0.4366,
+      "step": 16091
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.456828976316842e-05,
+      "loss": 0.3939,
+      "step": 16092
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.45512097426491e-05,
+      "loss": 0.6901,
+      "step": 16093
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.453413077019387e-05,
+      "loss": 0.0385,
+      "step": 16094
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.451705284607623e-05,
+      "loss": 0.1341,
+      "step": 16095
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.449997597056983e-05,
+      "loss": 0.6574,
+      "step": 16096
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.448290014394797e-05,
+      "loss": 0.3023,
+      "step": 16097
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.446582536648413e-05,
+      "loss": 0.2741,
+      "step": 16098
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.444875163845173e-05,
+      "loss": 0.4247,
+      "step": 16099
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.443167896012421e-05,
+      "loss": 0.0043,
+      "step": 16100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.441460733177494e-05,
+      "loss": 0.7763,
+      "step": 16101
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.439753675367731e-05,
+      "loss": 0.3207,
+      "step": 16102
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.438046722610466e-05,
+      "loss": 0.356,
+      "step": 16103
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.436339874933035e-05,
+      "loss": 0.2757,
+      "step": 16104
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.434633132362771e-05,
+      "loss": 0.2349,
+      "step": 16105
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.432926494927008e-05,
+      "loss": 0.3339,
+      "step": 16106
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.431219962653061e-05,
+      "loss": 0.3332,
+      "step": 16107
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.429513535568275e-05,
+      "loss": 0.3516,
+      "step": 16108
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.427807213699972e-05,
+      "loss": 0.1788,
+      "step": 16109
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.426100997075463e-05,
+      "loss": 0.4998,
+      "step": 16110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.424394885722089e-05,
+      "loss": 0.3296,
+      "step": 16111
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.422688879667164e-05,
+      "loss": 0.4492,
+      "step": 16112
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.420982978937994e-05,
+      "loss": 0.7821,
+      "step": 16113
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.419277183561916e-05,
+      "loss": 0.5054,
+      "step": 16114
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.417571493566242e-05,
+      "loss": 0.2891,
+      "step": 16115
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.415865908978277e-05,
+      "loss": 0.2442,
+      "step": 16116
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.414160429825336e-05,
+      "loss": 0.0027,
+      "step": 16117
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.412455056134734e-05,
+      "loss": 1.9336,
+      "step": 16118
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.410749787933776e-05,
+      "loss": 0.1799,
+      "step": 16119
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.409044625249773e-05,
+      "loss": 0.564,
+      "step": 16120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.407339568110026e-05,
+      "loss": 0.5773,
+      "step": 16121
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.405634616541843e-05,
+      "loss": 0.582,
+      "step": 16122
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.403929770572527e-05,
+      "loss": 0.4211,
+      "step": 16123
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.402225030229373e-05,
+      "loss": 0.5038,
+      "step": 16124
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.400520395539685e-05,
+      "loss": 0.4781,
+      "step": 16125
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.398815866530759e-05,
+      "loss": 0.3506,
+      "step": 16126
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.397111443229888e-05,
+      "loss": 0.5593,
+      "step": 16127
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.395407125664368e-05,
+      "loss": 0.3317,
+      "step": 16128
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.393702913861492e-05,
+      "loss": 0.1175,
+      "step": 16129
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.391998807848551e-05,
+      "loss": 0.7043,
+      "step": 16130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.390294807652828e-05,
+      "loss": 0.0407,
+      "step": 16131
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.388590913301614e-05,
+      "loss": 0.2304,
+      "step": 16132
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.386887124822197e-05,
+      "loss": 0.6499,
+      "step": 16133
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.385183442241862e-05,
+      "loss": 0.4708,
+      "step": 16134
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.383479865587875e-05,
+      "loss": 0.1056,
+      "step": 16135
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.381776394887539e-05,
+      "loss": 0.0096,
+      "step": 16136
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.380073030168118e-05,
+      "loss": 0.0084,
+      "step": 16137
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.378369771456885e-05,
+      "loss": 0.0004,
+      "step": 16138
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.376666618781135e-05,
+      "loss": 0.6559,
+      "step": 16139
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.374963572168122e-05,
+      "loss": 0.0021,
+      "step": 16140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.373260631645118e-05,
+      "loss": 0.4396,
+      "step": 16141
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.371557797239411e-05,
+      "loss": 0.126,
+      "step": 16142
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.369855068978255e-05,
+      "loss": 0.1831,
+      "step": 16143
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.368152446888911e-05,
+      "loss": 0.3179,
+      "step": 16144
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.366449930998666e-05,
+      "loss": 0.0018,
+      "step": 16145
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.364747521334763e-05,
+      "loss": 0.3954,
+      "step": 16146
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.36304521792447e-05,
+      "loss": 0.0073,
+      "step": 16147
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.36134302079505e-05,
+      "loss": 0.2421,
+      "step": 16148
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.359640929973755e-05,
+      "loss": 0.8341,
+      "step": 16149
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.357938945487846e-05,
+      "loss": 3.1794,
+      "step": 16150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.356237067364576e-05,
+      "loss": 0.6674,
+      "step": 16151
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.354535295631199e-05,
+      "loss": 0.7457,
+      "step": 16152
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.352833630314965e-05,
+      "loss": 0.0006,
+      "step": 16153
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.351132071443124e-05,
+      "loss": 0.0009,
+      "step": 16154
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.349430619042933e-05,
+      "loss": 0.5857,
+      "step": 16155
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.347729273141615e-05,
+      "loss": 0.0011,
+      "step": 16156
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.346028033766435e-05,
+      "loss": 0.9695,
+      "step": 16157
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.344326900944637e-05,
+      "loss": 0.3901,
+      "step": 16158
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.342625874703443e-05,
+      "loss": 0.5506,
+      "step": 16159
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.34092495507011e-05,
+      "loss": 0.2633,
+      "step": 16160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.339224142071876e-05,
+      "loss": 0.1852,
+      "step": 16161
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.337523435735959e-05,
+      "loss": 0.0896,
+      "step": 16162
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.33582283608961e-05,
+      "loss": 0.2375,
+      "step": 16163
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.334122343160063e-05,
+      "loss": 0.6948,
+      "step": 16164
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.332421956974532e-05,
+      "loss": 0.0008,
+      "step": 16165
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.33072167756027e-05,
+      "loss": 0.6718,
+      "step": 16166
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.329021504944482e-05,
+      "loss": 0.6656,
+      "step": 16167
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.327321439154404e-05,
+      "loss": 0.6382,
+      "step": 16168
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.32562148021726e-05,
+      "loss": 0.306,
+      "step": 16169
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.323921628160269e-05,
+      "loss": 0.0011,
+      "step": 16170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.322221883010654e-05,
+      "loss": 0.8186,
+      "step": 16171
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.320522244795635e-05,
+      "loss": 0.1506,
+      "step": 16172
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.318822713542428e-05,
+      "loss": 0.2487,
+      "step": 16173
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.317123289278246e-05,
+      "loss": 0.2803,
+      "step": 16174
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.315423972030307e-05,
+      "loss": 0.7483,
+      "step": 16175
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.31372476182582e-05,
+      "loss": 0.6535,
+      "step": 16176
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.312025658691998e-05,
+      "loss": 0.8216,
+      "step": 16177
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.310326662656045e-05,
+      "loss": 0.6186,
+      "step": 16178
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.308627773745171e-05,
+      "loss": 0.5416,
+      "step": 16179
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.306928991986582e-05,
+      "loss": 1.6035,
+      "step": 16180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.30523031740748e-05,
+      "loss": 0.2282,
+      "step": 16181
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.303531750035068e-05,
+      "loss": 0.0922,
+      "step": 16182
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.301833289896548e-05,
+      "loss": 0.7594,
+      "step": 16183
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.300134937019108e-05,
+      "loss": 0.4739,
+      "step": 16184
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.298436691429962e-05,
+      "loss": 0.0008,
+      "step": 16185
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.296738553156287e-05,
+      "loss": 0.3519,
+      "step": 16186
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.29504052222528e-05,
+      "loss": 0.284,
+      "step": 16187
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.293342598664147e-05,
+      "loss": 0.1783,
+      "step": 16188
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.291644782500061e-05,
+      "loss": 0.2638,
+      "step": 16189
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.289947073760212e-05,
+      "loss": 0.5424,
+      "step": 16190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.288249472471799e-05,
+      "loss": 0.6493,
+      "step": 16191
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.286551978661994e-05,
+      "loss": 0.0064,
+      "step": 16192
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.284854592357976e-05,
+      "loss": 0.005,
+      "step": 16193
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.283157313586947e-05,
+      "loss": 2.457,
+      "step": 16194
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.281460142376067e-05,
+      "loss": 2.4024,
+      "step": 16195
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.279763078752519e-05,
+      "loss": 0.3497,
+      "step": 16196
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.27806612274348e-05,
+      "loss": 0.3541,
+      "step": 16197
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.276369274376125e-05,
+      "loss": 0.9505,
+      "step": 16198
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.274672533677622e-05,
+      "loss": 0.4644,
+      "step": 16199
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.272975900675148e-05,
+      "loss": 0.4722,
+      "step": 16200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.271279375395868e-05,
+      "loss": 0.6269,
+      "step": 16201
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.26958295786695e-05,
+      "loss": 0.2699,
+      "step": 16202
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.267886648115562e-05,
+      "loss": 0.342,
+      "step": 16203
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.26619044616887e-05,
+      "loss": 0.3257,
+      "step": 16204
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.264494352054021e-05,
+      "loss": 0.0005,
+      "step": 16205
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.262798365798194e-05,
+      "loss": 0.0035,
+      "step": 16206
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.261102487428542e-05,
+      "loss": 0.6795,
+      "step": 16207
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.25940671697221e-05,
+      "loss": 0.2168,
+      "step": 16208
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.257711054456369e-05,
+      "loss": 0.0009,
+      "step": 16209
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.256015499908173e-05,
+      "loss": 0.3204,
+      "step": 16210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.254320053354754e-05,
+      "loss": 0.1848,
+      "step": 16211
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.252624714823284e-05,
+      "loss": 0.002,
+      "step": 16212
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.250929484340908e-05,
+      "loss": 0.0015,
+      "step": 16213
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.249234361934756e-05,
+      "loss": 0.0044,
+      "step": 16214
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.247539347631994e-05,
+      "loss": 0.78,
+      "step": 16215
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.245844441459752e-05,
+      "loss": 0.0061,
+      "step": 16216
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.244149643445166e-05,
+      "loss": 0.9425,
+      "step": 16217
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.242454953615398e-05,
+      "loss": 0.4105,
+      "step": 16218
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.240760371997566e-05,
+      "loss": 0.8078,
+      "step": 16219
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.239065898618813e-05,
+      "loss": 0.5262,
+      "step": 16220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.237371533506272e-05,
+      "loss": 0.4266,
+      "step": 16221
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.235677276687075e-05,
+      "loss": 1.2897,
+      "step": 16222
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.233983128188358e-05,
+      "loss": 0.7068,
+      "step": 16223
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.232289088037243e-05,
+      "loss": 0.4073,
+      "step": 16224
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.230595156260866e-05,
+      "loss": 0.0041,
+      "step": 16225
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.228901332886344e-05,
+      "loss": 0.3823,
+      "step": 16226
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.227207617940809e-05,
+      "loss": 0.6746,
+      "step": 16227
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.225514011451375e-05,
+      "loss": 0.1844,
+      "step": 16228
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.223820513445171e-05,
+      "loss": 0.4413,
+      "step": 16229
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.22212712394931e-05,
+      "loss": 0.2403,
+      "step": 16230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.220433842990911e-05,
+      "loss": 0.1952,
+      "step": 16231
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.218740670597088e-05,
+      "loss": 0.3796,
+      "step": 16232
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.217047606794958e-05,
+      "loss": 0.1841,
+      "step": 16233
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.215354651611636e-05,
+      "loss": 0.8358,
+      "step": 16234
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.213661805074222e-05,
+      "loss": 0.3603,
+      "step": 16235
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.211969067209821e-05,
+      "loss": 0.1423,
+      "step": 16236
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.210276438045563e-05,
+      "loss": 0.7764,
+      "step": 16237
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.20858391760853e-05,
+      "loss": 0.3549,
+      "step": 16238
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.206891505925827e-05,
+      "loss": 0.2758,
+      "step": 16239
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.205199203024574e-05,
+      "loss": 0.3525,
+      "step": 16240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.203507008931851e-05,
+      "loss": 0.2856,
+      "step": 16241
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.20181492367476e-05,
+      "loss": 0.6795,
+      "step": 16242
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.200122947280408e-05,
+      "loss": 0.3912,
+      "step": 16243
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.19843107977588e-05,
+      "loss": 0.667,
+      "step": 16244
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.196739321188269e-05,
+      "loss": 0.3458,
+      "step": 16245
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.195047671544667e-05,
+      "loss": 1.3885,
+      "step": 16246
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.193356130872165e-05,
+      "loss": 0.2837,
+      "step": 16247
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.191664699197849e-05,
+      "loss": 1.2233,
+      "step": 16248
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.189973376548802e-05,
+      "loss": 0.1919,
+      "step": 16249
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.188282162952113e-05,
+      "loss": 0.6065,
+      "step": 16250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.186591058434865e-05,
+      "loss": 0.5211,
+      "step": 16251
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.184900063024132e-05,
+      "loss": 0.0056,
+      "step": 16252
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.183209176747005e-05,
+      "loss": 0.4832,
+      "step": 16253
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.181518399630536e-05,
+      "loss": 1.1795,
+      "step": 16254
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.179827731701829e-05,
+      "loss": 0.2604,
+      "step": 16255
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.178137172987946e-05,
+      "loss": 0.1325,
+      "step": 16256
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.176446723515948e-05,
+      "loss": 0.4397,
+      "step": 16257
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.174756383312921e-05,
+      "loss": 0.0007,
+      "step": 16258
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.173066152405935e-05,
+      "loss": 1.9222,
+      "step": 16259
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.171376030822034e-05,
+      "loss": 0.2178,
+      "step": 16260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.169686018588305e-05,
+      "loss": 0.7174,
+      "step": 16261
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.167996115731808e-05,
+      "loss": 0.2318,
+      "step": 16262
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.166306322279592e-05,
+      "loss": 1.7373,
+      "step": 16263
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.164616638258731e-05,
+      "loss": 0.0037,
+      "step": 16264
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.162927063696272e-05,
+      "loss": 0.4856,
+      "step": 16265
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.161237598619269e-05,
+      "loss": 4.0334,
+      "step": 16266
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.159548243054796e-05,
+      "loss": 5.9228,
+      "step": 16267
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.157858997029883e-05,
+      "loss": 0.3337,
+      "step": 16268
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.156169860571592e-05,
+      "loss": 1.0686,
+      "step": 16269
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.154480833706967e-05,
+      "loss": 0.1636,
+      "step": 16270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.15279191646306e-05,
+      "loss": 0.0054,
+      "step": 16271
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.151103108866915e-05,
+      "loss": 0.3654,
+      "step": 16272
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.149414410945574e-05,
+      "loss": 0.3322,
+      "step": 16273
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.14772582272608e-05,
+      "loss": 1.5625,
+      "step": 16274
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.146037344235473e-05,
+      "loss": 0.1646,
+      "step": 16275
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.14434897550079e-05,
+      "loss": 0.0991,
+      "step": 16276
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.142660716549071e-05,
+      "loss": 0.0051,
+      "step": 16277
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.140972567407348e-05,
+      "loss": 0.4982,
+      "step": 16278
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.139284528102655e-05,
+      "loss": 0.6649,
+      "step": 16279
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.137596598662026e-05,
+      "loss": 0.6101,
+      "step": 16280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.135908779112487e-05,
+      "loss": 1.0027,
+      "step": 16281
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.134221069481067e-05,
+      "loss": 0.0005,
+      "step": 16282
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.132533469794798e-05,
+      "loss": 0.4573,
+      "step": 16283
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.13084598008069e-05,
+      "loss": 0.1218,
+      "step": 16284
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.129158600365777e-05,
+      "loss": 0.7083,
+      "step": 16285
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.127471330677086e-05,
+      "loss": 2.3711,
+      "step": 16286
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.12578417104162e-05,
+      "loss": 0.1573,
+      "step": 16287
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.124097121486397e-05,
+      "loss": 0.6218,
+      "step": 16288
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.122410182038453e-05,
+      "loss": 0.644,
+      "step": 16289
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.120723352724782e-05,
+      "loss": 0.456,
+      "step": 16290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.119036633572395e-05,
+      "loss": 0.2603,
+      "step": 16291
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.117350024608322e-05,
+      "loss": 2.4492,
+      "step": 16292
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.115663525859553e-05,
+      "loss": 0.0007,
+      "step": 16293
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.113977137353099e-05,
+      "loss": 0.6835,
+      "step": 16294
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.112290859115971e-05,
+      "loss": 0.7,
+      "step": 16295
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.110604691175164e-05,
+      "loss": 0.3153,
+      "step": 16296
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.108918633557685e-05,
+      "loss": 0.0356,
+      "step": 16297
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.107232686290531e-05,
+      "loss": 0.2516,
+      "step": 16298
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.105546849400702e-05,
+      "loss": 0.4796,
+      "step": 16299
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.10386112291519e-05,
+      "loss": 0.0076,
+      "step": 16300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.102175506860997e-05,
+      "loss": 0.8886,
+      "step": 16301
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.100490001265114e-05,
+      "loss": 0.7835,
+      "step": 16302
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.09880460615452e-05,
+      "loss": 1.3165,
+      "step": 16303
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.097119321556216e-05,
+      "loss": 0.2912,
+      "step": 16304
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.095434147497193e-05,
+      "loss": 0.3301,
+      "step": 16305
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.093749084004417e-05,
+      "loss": 0.6506,
+      "step": 16306
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.092064131104892e-05,
+      "loss": 0.1596,
+      "step": 16307
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.090379288825598e-05,
+      "loss": 0.6729,
+      "step": 16308
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.088694557193499e-05,
+      "loss": 0.3083,
+      "step": 16309
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.087009936235588e-05,
+      "loss": 0.3496,
+      "step": 16310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.085325425978846e-05,
+      "loss": 2.4746,
+      "step": 16311
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.083641026450226e-05,
+      "loss": 0.0095,
+      "step": 16312
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.081956737676729e-05,
+      "loss": 0.3666,
+      "step": 16313
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.080272559685306e-05,
+      "loss": 0.3306,
+      "step": 16314
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.078588492502926e-05,
+      "loss": 0.8594,
+      "step": 16315
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.076904536156575e-05,
+      "loss": 0.4728,
+      "step": 16316
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.075220690673204e-05,
+      "loss": 0.8454,
+      "step": 16317
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.073536956079773e-05,
+      "loss": 0.0057,
+      "step": 16318
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.071853332403265e-05,
+      "loss": 0.0033,
+      "step": 16319
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.070169819670622e-05,
+      "loss": 0.005,
+      "step": 16320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.06848641790881e-05,
+      "loss": 0.7564,
+      "step": 16321
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.066803127144785e-05,
+      "loss": 0.435,
+      "step": 16322
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.065119947405501e-05,
+      "loss": 0.006,
+      "step": 16323
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.063436878717915e-05,
+      "loss": 0.4973,
+      "step": 16324
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.061753921108976e-05,
+      "loss": 1.78,
+      "step": 16325
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 8.060071074605636e-05,
+      "loss": 0.6066,
+      "step": 16326
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.058388339234843e-05,
+      "loss": 0.0071,
+      "step": 16327
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.056705715023542e-05,
+      "loss": 0.0303,
+      "step": 16328
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.05502320199868e-05,
+      "loss": 0.2878,
+      "step": 16329
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.053340800187196e-05,
+      "loss": 0.4762,
+      "step": 16330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.051658509616035e-05,
+      "loss": 0.5096,
+      "step": 16331
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.04997633031214e-05,
+      "loss": 0.7505,
+      "step": 16332
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.048294262302432e-05,
+      "loss": 0.2647,
+      "step": 16333
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.046612305613864e-05,
+      "loss": 0.8047,
+      "step": 16334
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.044930460273372e-05,
+      "loss": 0.0039,
+      "step": 16335
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.043248726307873e-05,
+      "loss": 0.695,
+      "step": 16336
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.041567103744299e-05,
+      "loss": 0.7035,
+      "step": 16337
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.039885592609599e-05,
+      "loss": 0.5911,
+      "step": 16338
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.038204192930675e-05,
+      "loss": 0.1847,
+      "step": 16339
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.036522904734459e-05,
+      "loss": 0.2676,
+      "step": 16340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.034841728047893e-05,
+      "loss": 2.9548,
+      "step": 16341
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.033160662897873e-05,
+      "loss": 0.6745,
+      "step": 16342
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.031479709311332e-05,
+      "loss": 0.2063,
+      "step": 16343
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.029798867315183e-05,
+      "loss": 2.5842,
+      "step": 16344
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.028118136936346e-05,
+      "loss": 0.0085,
+      "step": 16345
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.026437518201733e-05,
+      "loss": 0.7286,
+      "step": 16346
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.024757011138257e-05,
+      "loss": 0.4307,
+      "step": 16347
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.023076615772831e-05,
+      "loss": 0.538,
+      "step": 16348
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.021396332132362e-05,
+      "loss": 0.2862,
+      "step": 16349
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.019716160243753e-05,
+      "loss": 0.1472,
+      "step": 16350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.018036100133918e-05,
+      "loss": 0.0022,
+      "step": 16351
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.016356151829756e-05,
+      "loss": 1.1011,
+      "step": 16352
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.014676315358167e-05,
+      "loss": 0.7155,
+      "step": 16353
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.01299659074606e-05,
+      "loss": 0.2415,
+      "step": 16354
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.011316978020319e-05,
+      "loss": 0.2924,
+      "step": 16355
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.009637477207847e-05,
+      "loss": 0.4814,
+      "step": 16356
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.007958088335551e-05,
+      "loss": 0.1848,
+      "step": 16357
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.006278811430298e-05,
+      "loss": 0.0196,
+      "step": 16358
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.004599646519e-05,
+      "loss": 0.3703,
+      "step": 16359
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.002920593628547e-05,
+      "loss": 0.0011,
+      "step": 16360
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 8.001241652785808e-05,
+      "loss": 0.807,
+      "step": 16361
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.99956282401769e-05,
+      "loss": 0.1338,
+      "step": 16362
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.997884107351064e-05,
+      "loss": 0.7851,
+      "step": 16363
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.996205502812808e-05,
+      "loss": 0.002,
+      "step": 16364
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.994527010429822e-05,
+      "loss": 0.1929,
+      "step": 16365
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.992848630228964e-05,
+      "loss": 0.0109,
+      "step": 16366
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.991170362237114e-05,
+      "loss": 0.4216,
+      "step": 16367
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.989492206481165e-05,
+      "loss": 0.2165,
+      "step": 16368
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.98781416298797e-05,
+      "loss": 0.3601,
+      "step": 16369
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.986136231784402e-05,
+      "loss": 0.2807,
+      "step": 16370
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.984458412897348e-05,
+      "loss": 0.0254,
+      "step": 16371
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.982780706353657e-05,
+      "loss": 0.9893,
+      "step": 16372
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.981103112180202e-05,
+      "loss": 2.0871,
+      "step": 16373
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.979425630403847e-05,
+      "loss": 0.4177,
+      "step": 16374
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.977748261051453e-05,
+      "loss": 0.0809,
+      "step": 16375
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.976071004149883e-05,
+      "loss": 0.962,
+      "step": 16376
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.974393859725994e-05,
+      "loss": 0.0299,
+      "step": 16377
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.972716827806644e-05,
+      "loss": 0.4402,
+      "step": 16378
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.971039908418687e-05,
+      "loss": 0.174,
+      "step": 16379
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.969363101588977e-05,
+      "loss": 0.1369,
+      "step": 16380
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.967686407344371e-05,
+      "loss": 0.2308,
+      "step": 16381
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.966009825711702e-05,
+      "loss": 0.2033,
+      "step": 16382
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.964333356717835e-05,
+      "loss": 0.6859,
+      "step": 16383
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.962657000389613e-05,
+      "loss": 0.0012,
+      "step": 16384
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.960980756753871e-05,
+      "loss": 0.4786,
+      "step": 16385
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.959304625837461e-05,
+      "loss": 0.3319,
+      "step": 16386
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.957628607667226e-05,
+      "loss": 0.002,
+      "step": 16387
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.955952702269995e-05,
+      "loss": 0.3562,
+      "step": 16388
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.954276909672605e-05,
+      "loss": 0.597,
+      "step": 16389
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.952601229901907e-05,
+      "loss": 0.5286,
+      "step": 16390
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.950925662984716e-05,
+      "loss": 0.0037,
+      "step": 16391
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.949250208947873e-05,
+      "loss": 0.3311,
+      "step": 16392
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.947574867818207e-05,
+      "loss": 0.2305,
+      "step": 16393
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.945899639622543e-05,
+      "loss": 0.4033,
+      "step": 16394
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.944224524387713e-05,
+      "loss": 0.4518,
+      "step": 16395
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.942549522140535e-05,
+      "loss": 0.4109,
+      "step": 16396
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.940874632907835e-05,
+      "loss": 0.3706,
+      "step": 16397
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.939199856716434e-05,
+      "loss": 0.4839,
+      "step": 16398
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.937525193593151e-05,
+      "loss": 0.1894,
+      "step": 16399
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.935850643564801e-05,
+      "loss": 0.532,
+      "step": 16400
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.934176206658202e-05,
+      "loss": 0.6118,
+      "step": 16401
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.932501882900168e-05,
+      "loss": 0.2334,
+      "step": 16402
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.930827672317514e-05,
+      "loss": 0.6071,
+      "step": 16403
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.929153574937036e-05,
+      "loss": 0.4617,
+      "step": 16404
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.927479590785556e-05,
+      "loss": 0.4326,
+      "step": 16405
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.925805719889881e-05,
+      "loss": 0.7616,
+      "step": 16406
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.924131962276799e-05,
+      "loss": 0.2973,
+      "step": 16407
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.922458317973131e-05,
+      "loss": 0.3138,
+      "step": 16408
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.920784787005677e-05,
+      "loss": 0.7715,
+      "step": 16409
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.919111369401218e-05,
+      "loss": 0.2287,
+      "step": 16410
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.917438065186575e-05,
+      "loss": 0.2113,
+      "step": 16411
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.915764874388526e-05,
+      "loss": 0.1771,
+      "step": 16412
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.914091797033865e-05,
+      "loss": 0.001,
+      "step": 16413
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.912418833149402e-05,
+      "loss": 0.3268,
+      "step": 16414
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.910745982761907e-05,
+      "loss": 0.9341,
+      "step": 16415
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.909073245898171e-05,
+      "loss": 0.0492,
+      "step": 16416
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.907400622584998e-05,
+      "loss": 0.3307,
+      "step": 16417
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.905728112849151e-05,
+      "loss": 0.2102,
+      "step": 16418
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.904055716717415e-05,
+      "loss": 0.0043,
+      "step": 16419
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.902383434216589e-05,
+      "loss": 0.2387,
+      "step": 16420
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.900711265373437e-05,
+      "loss": 0.3483,
+      "step": 16421
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.899039210214737e-05,
+      "loss": 0.3054,
+      "step": 16422
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.897367268767269e-05,
+      "loss": 0.0018,
+      "step": 16423
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.895695441057803e-05,
+      "loss": 0.3072,
+      "step": 16424
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.894023727113114e-05,
+      "loss": 0.2102,
+      "step": 16425
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.89235212695997e-05,
+      "loss": 0.0015,
+      "step": 16426
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.890680640625139e-05,
+      "loss": 0.6611,
+      "step": 16427
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.88900926813539e-05,
+      "loss": 0.0022,
+      "step": 16428
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.887338009517482e-05,
+      "loss": 0.7057,
+      "step": 16429
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.885666864798189e-05,
+      "loss": 0.0135,
+      "step": 16430
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.883995834004252e-05,
+      "loss": 0.5586,
+      "step": 16431
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.882324917162447e-05,
+      "loss": 0.5224,
+      "step": 16432
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.880654114299533e-05,
+      "loss": 0.0037,
+      "step": 16433
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.878983425442248e-05,
+      "loss": 0.5966,
+      "step": 16434
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.877312850617361e-05,
+      "loss": 0.8489,
+      "step": 16435
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.875642389851625e-05,
+      "loss": 0.0053,
+      "step": 16436
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.873972043171772e-05,
+      "loss": 0.3555,
+      "step": 16437
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.87230181060457e-05,
+      "loss": 0.9353,
+      "step": 16438
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.87063169217676e-05,
+      "loss": 1.4282,
+      "step": 16439
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.868961687915081e-05,
+      "loss": 0.5126,
+      "step": 16440
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.867291797846276e-05,
+      "loss": 0.3603,
+      "step": 16441
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.86562202199709e-05,
+      "loss": 1.6593,
+      "step": 16442
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.863952360394258e-05,
+      "loss": 0.22,
+      "step": 16443
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.86228281306452e-05,
+      "loss": 0.1343,
+      "step": 16444
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.86061338003461e-05,
+      "loss": 0.959,
+      "step": 16445
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.858944061331264e-05,
+      "loss": 0.8204,
+      "step": 16446
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.857274856981211e-05,
+      "loss": 0.4536,
+      "step": 16447
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.855605767011181e-05,
+      "loss": 0.2274,
+      "step": 16448
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.853936791447901e-05,
+      "loss": 0.5781,
+      "step": 16449
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.852267930318101e-05,
+      "loss": 1.9945,
+      "step": 16450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.850599183648502e-05,
+      "loss": 0.4726,
+      "step": 16451
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.848930551465828e-05,
+      "loss": 0.0037,
+      "step": 16452
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.847262033796797e-05,
+      "loss": 0.9288,
+      "step": 16453
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.84559363066813e-05,
+      "loss": 0.5953,
+      "step": 16454
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.84392534210655e-05,
+      "loss": 0.045,
+      "step": 16455
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.842257168138756e-05,
+      "loss": 0.0112,
+      "step": 16456
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.840589108791474e-05,
+      "loss": 0.22,
+      "step": 16457
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.838921164091421e-05,
+      "loss": 0.3509,
+      "step": 16458
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.837253334065287e-05,
+      "loss": 0.0013,
+      "step": 16459
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.8355856187398e-05,
+      "loss": 1.5903,
+      "step": 16460
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.833918018141654e-05,
+      "loss": 0.0025,
+      "step": 16461
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.832250532297547e-05,
+      "loss": 0.3037,
+      "step": 16462
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.830583161234205e-05,
+      "loss": 0.2702,
+      "step": 16463
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.828915904978306e-05,
+      "loss": 0.3121,
+      "step": 16464
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.827248763556554e-05,
+      "loss": 0.6396,
+      "step": 16465
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.825581736995656e-05,
+      "loss": 1.539,
+      "step": 16466
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.823914825322297e-05,
+      "loss": 0.5759,
+      "step": 16467
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.822248028563163e-05,
+      "loss": 0.0822,
+      "step": 16468
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.820581346744968e-05,
+      "loss": 0.5289,
+      "step": 16469
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.818914779894381e-05,
+      "loss": 0.5234,
+      "step": 16470
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.817248328038099e-05,
+      "loss": 0.867,
+      "step": 16471
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.815581991202803e-05,
+      "loss": 0.5907,
+      "step": 16472
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.81391576941518e-05,
+      "loss": 0.2642,
+      "step": 16473
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.81224966270191e-05,
+      "loss": 0.4347,
+      "step": 16474
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.810583671089675e-05,
+      "loss": 0.9064,
+      "step": 16475
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.808917794605153e-05,
+      "loss": 0.0616,
+      "step": 16476
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.80725203327502e-05,
+      "loss": 0.2159,
+      "step": 16477
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.805586387125952e-05,
+      "loss": 2.5145,
+      "step": 16478
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.803920856184625e-05,
+      "loss": 0.3263,
+      "step": 16479
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.802255440477695e-05,
+      "loss": 0.0049,
+      "step": 16480
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.800590140031846e-05,
+      "loss": 0.7058,
+      "step": 16481
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.79892495487375e-05,
+      "loss": 0.2799,
+      "step": 16482
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.797259885030049e-05,
+      "loss": 0.6057,
+      "step": 16483
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.795594930527429e-05,
+      "loss": 1.8644,
+      "step": 16484
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.793930091392548e-05,
+      "loss": 0.4909,
+      "step": 16485
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.79226536765205e-05,
+      "loss": 0.0021,
+      "step": 16486
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.790600759332611e-05,
+      "loss": 0.1306,
+      "step": 16487
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.788936266460887e-05,
+      "loss": 0.2421,
+      "step": 16488
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.787271889063521e-05,
+      "loss": 0.2824,
+      "step": 16489
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.785607627167173e-05,
+      "loss": 0.0834,
+      "step": 16490
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.783943480798489e-05,
+      "loss": 0.2336,
+      "step": 16491
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.782279449984122e-05,
+      "loss": 0.2093,
+      "step": 16492
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.780615534750715e-05,
+      "loss": 0.2894,
+      "step": 16493
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.778951735124916e-05,
+      "loss": 0.7593,
+      "step": 16494
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.777288051133369e-05,
+      "loss": 0.4173,
+      "step": 16495
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.775624482802712e-05,
+      "loss": 0.2242,
+      "step": 16496
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.773961030159588e-05,
+      "loss": 1.1212,
+      "step": 16497
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.772297693230633e-05,
+      "loss": 0.4707,
+      "step": 16498
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.770634472042484e-05,
+      "loss": 0.6155,
+      "step": 16499
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.768971366621775e-05,
+      "loss": 0.7428,
+      "step": 16500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.767308376995134e-05,
+      "loss": 0.8872,
+      "step": 16501
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.765645503189197e-05,
+      "loss": 0.1756,
+      "step": 16502
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.76398274523059e-05,
+      "loss": 0.3567,
+      "step": 16503
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.762320103145944e-05,
+      "loss": 0.9851,
+      "step": 16504
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.76065757696187e-05,
+      "loss": 0.2708,
+      "step": 16505
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.758995166705003e-05,
+      "loss": 0.6567,
+      "step": 16506
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.75733287240197e-05,
+      "loss": 0.8073,
+      "step": 16507
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.755670694079369e-05,
+      "loss": 0.5912,
+      "step": 16508
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.75400863176384e-05,
+      "loss": 1.1257,
+      "step": 16509
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.752346685481983e-05,
+      "loss": 0.7689,
+      "step": 16510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.75068485526041e-05,
+      "loss": 0.6427,
+      "step": 16511
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.749023141125751e-05,
+      "loss": 0.3999,
+      "step": 16512
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.7473615431046e-05,
+      "loss": 0.1507,
+      "step": 16513
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.745700061223563e-05,
+      "loss": 0.5443,
+      "step": 16514
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.744038695509262e-05,
+      "loss": 0.3357,
+      "step": 16515
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.742377445988287e-05,
+      "loss": 0.0005,
+      "step": 16516
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.74071631268724e-05,
+      "loss": 0.2324,
+      "step": 16517
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.739055295632738e-05,
+      "loss": 0.0695,
+      "step": 16518
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.737394394851361e-05,
+      "loss": 0.1366,
+      "step": 16519
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.735733610369716e-05,
+      "loss": 0.0106,
+      "step": 16520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.734072942214392e-05,
+      "loss": 0.7951,
+      "step": 16521
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.732412390411987e-05,
+      "loss": 0.7567,
+      "step": 16522
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.73075195498909e-05,
+      "loss": 0.0623,
+      "step": 16523
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.729091635972291e-05,
+      "loss": 1.102,
+      "step": 16524
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.727431433388176e-05,
+      "loss": 0.34,
+      "step": 16525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.725771347263333e-05,
+      "loss": 0.0034,
+      "step": 16526
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.724111377624345e-05,
+      "loss": 0.9419,
+      "step": 16527
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.7224515244978e-05,
+      "loss": 0.0155,
+      "step": 16528
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.72079178791026e-05,
+      "loss": 0.3663,
+      "step": 16529
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.719132167888318e-05,
+      "loss": 2.0229,
+      "step": 16530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.717472664458558e-05,
+      "loss": 2.0684,
+      "step": 16531
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.715813277647529e-05,
+      "loss": 0.0024,
+      "step": 16532
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.714154007481823e-05,
+      "loss": 0.4729,
+      "step": 16533
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.712494853988015e-05,
+      "loss": 0.7643,
+      "step": 16534
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.710835817192652e-05,
+      "loss": 0.3142,
+      "step": 16535
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.70917689712232e-05,
+      "loss": 0.0016,
+      "step": 16536
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.707518093803582e-05,
+      "loss": 0.6005,
+      "step": 16537
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.705859407262988e-05,
+      "loss": 0.0015,
+      "step": 16538
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.704200837527119e-05,
+      "loss": 0.3215,
+      "step": 16539
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.702542384622518e-05,
+      "loss": 0.127,
+      "step": 16540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.700884048575748e-05,
+      "loss": 0.3598,
+      "step": 16541
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.699225829413366e-05,
+      "loss": 0.401,
+      "step": 16542
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.697567727161928e-05,
+      "loss": 0.3397,
+      "step": 16543
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.695909741847982e-05,
+      "loss": 0.8195,
+      "step": 16544
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.69425187349808e-05,
+      "loss": 0.962,
+      "step": 16545
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.692594122138768e-05,
+      "loss": 0.4391,
+      "step": 16546
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.690936487796595e-05,
+      "loss": 0.3176,
+      "step": 16547
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.689278970498104e-05,
+      "loss": 0.1944,
+      "step": 16548
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.687621570269839e-05,
+      "loss": 0.1366,
+      "step": 16549
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.685964287138342e-05,
+      "loss": 0.4647,
+      "step": 16550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.68430712113015e-05,
+      "loss": 0.0026,
+      "step": 16551
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.682650072271798e-05,
+      "loss": 0.3109,
+      "step": 16552
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.680993140589824e-05,
+      "loss": 0.0289,
+      "step": 16553
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.67933632611076e-05,
+      "loss": 0.6229,
+      "step": 16554
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.677679628861141e-05,
+      "loss": 0.2162,
+      "step": 16555
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.676023048867496e-05,
+      "loss": 0.3175,
+      "step": 16556
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.67436658615634e-05,
+      "loss": 0.4309,
+      "step": 16557
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.672710240754219e-05,
+      "loss": 0.2051,
+      "step": 16558
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.671054012687643e-05,
+      "loss": 0.3086,
+      "step": 16559
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.66939790198313e-05,
+      "loss": 0.315,
+      "step": 16560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.667741908667222e-05,
+      "loss": 0.6955,
+      "step": 16561
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.666086032766416e-05,
+      "loss": 0.0005,
+      "step": 16562
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.664430274307228e-05,
+      "loss": 0.1521,
+      "step": 16563
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.662774633316194e-05,
+      "loss": 0.6042,
+      "step": 16564
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.661119109819806e-05,
+      "loss": 3.0216,
+      "step": 16565
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.659463703844575e-05,
+      "loss": 0.9288,
+      "step": 16566
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.657808415417028e-05,
+      "loss": 0.4779,
+      "step": 16567
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.656153244563656e-05,
+      "loss": 0.5724,
+      "step": 16568
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.654498191310968e-05,
+      "loss": 0.0014,
+      "step": 16569
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.652843255685464e-05,
+      "loss": 0.2968,
+      "step": 16570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.651188437713652e-05,
+      "loss": 0.2642,
+      "step": 16571
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.649533737422027e-05,
+      "loss": 0.3238,
+      "step": 16572
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.647879154837087e-05,
+      "loss": 0.3601,
+      "step": 16573
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.64622468998533e-05,
+      "loss": 1.3811,
+      "step": 16574
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 7.644570342893247e-05,
+      "loss": 0.8593,
+      "step": 16575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.64291611358733e-05,
+      "loss": 0.5268,
+      "step": 16576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.641262002094078e-05,
+      "loss": 0.1428,
+      "step": 16577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.63960800843996e-05,
+      "loss": 0.0008,
+      "step": 16578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.637954132651478e-05,
+      "loss": 0.0014,
+      "step": 16579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.636300374755117e-05,
+      "loss": 0.7652,
+      "step": 16580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.634646734777343e-05,
+      "loss": 0.3899,
+      "step": 16581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.632993212744654e-05,
+      "loss": 0.0622,
+      "step": 16582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.631339808683526e-05,
+      "loss": 1.7891,
+      "step": 16583
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.629686522620423e-05,
+      "loss": 0.289,
+      "step": 16584
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.628033354581832e-05,
+      "loss": 0.5517,
+      "step": 16585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.626380304594232e-05,
+      "loss": 0.3464,
+      "step": 16586
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.624727372684073e-05,
+      "loss": 0.0108,
+      "step": 16587
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.623074558877847e-05,
+      "loss": 0.003,
+      "step": 16588
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.621421863202004e-05,
+      "loss": 0.2218,
+      "step": 16589
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.619769285683011e-05,
+      "loss": 1.1797,
+      "step": 16590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.618116826347349e-05,
+      "loss": 0.0225,
+      "step": 16591
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.616464485221462e-05,
+      "loss": 2.1261,
+      "step": 16592
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.614812262331815e-05,
+      "loss": 0.3139,
+      "step": 16593
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.613160157704865e-05,
+      "loss": 0.761,
+      "step": 16594
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.61150817136707e-05,
+      "loss": 0.661,
+      "step": 16595
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.609856303344882e-05,
+      "loss": 0.0032,
+      "step": 16596
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.608204553664755e-05,
+      "loss": 0.3285,
+      "step": 16597
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.60655292235314e-05,
+      "loss": 0.4676,
+      "step": 16598
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.604901409436481e-05,
+      "loss": 0.0431,
+      "step": 16599
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.603250014941229e-05,
+      "loss": 0.2423,
+      "step": 16600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.601598738893828e-05,
+      "loss": 0.0008,
+      "step": 16601
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.599947581320718e-05,
+      "loss": 0.3115,
+      "step": 16602
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.598296542248342e-05,
+      "loss": 1.922,
+      "step": 16603
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.596645621703138e-05,
+      "loss": 0.5959,
+      "step": 16604
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.594994819711544e-05,
+      "loss": 1.4707,
+      "step": 16605
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.593344136299996e-05,
+      "loss": 0.0072,
+      "step": 16606
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.591693571494929e-05,
+      "loss": 0.3854,
+      "step": 16607
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.590043125322767e-05,
+      "loss": 0.0573,
+      "step": 16608
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.588392797809938e-05,
+      "loss": 1.5335,
+      "step": 16609
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.586742588982887e-05,
+      "loss": 0.4429,
+      "step": 16610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.58509249886802e-05,
+      "loss": 0.7112,
+      "step": 16611
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.583442527491764e-05,
+      "loss": 0.001,
+      "step": 16612
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.581792674880557e-05,
+      "loss": 0.2115,
+      "step": 16613
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.5801429410608e-05,
+      "loss": 0.3275,
+      "step": 16614
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.578493326058915e-05,
+      "loss": 0.2937,
+      "step": 16615
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.576843829901329e-05,
+      "loss": 0.2614,
+      "step": 16616
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.575194452614445e-05,
+      "loss": 0.6749,
+      "step": 16617
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.573545194224679e-05,
+      "loss": 0.2843,
+      "step": 16618
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.57189605475844e-05,
+      "loss": 0.0014,
+      "step": 16619
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.570247034242138e-05,
+      "loss": 0.6207,
+      "step": 16620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.568598132702179e-05,
+      "loss": 0.2743,
+      "step": 16621
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.56694935016497e-05,
+      "loss": 0.2978,
+      "step": 16622
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.565300686656908e-05,
+      "loss": 1.0879,
+      "step": 16623
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.5636521422044e-05,
+      "loss": 0.223,
+      "step": 16624
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.562003716833843e-05,
+      "loss": 0.5063,
+      "step": 16625
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.560355410571639e-05,
+      "loss": 0.701,
+      "step": 16626
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.558707223444167e-05,
+      "loss": 0.0027,
+      "step": 16627
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.557059155477836e-05,
+      "loss": 0.0009,
+      "step": 16628
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.555411206699037e-05,
+      "loss": 0.5686,
+      "step": 16629
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.553763377134144e-05,
+      "loss": 0.4526,
+      "step": 16630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.552115666809561e-05,
+      "loss": 0.2528,
+      "step": 16631
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.550468075751674e-05,
+      "loss": 0.3618,
+      "step": 16632
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.54882060398685e-05,
+      "loss": 0.4099,
+      "step": 16633
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.547173251541484e-05,
+      "loss": 0.1458,
+      "step": 16634
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.545526018441962e-05,
+      "loss": 0.8274,
+      "step": 16635
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.54387890471464e-05,
+      "loss": 0.0058,
+      "step": 16636
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.542231910385919e-05,
+      "loss": 0.0222,
+      "step": 16637
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.540585035482158e-05,
+      "loss": 1.1284,
+      "step": 16638
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.538938280029723e-05,
+      "loss": 1.2656,
+      "step": 16639
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.537291644055008e-05,
+      "loss": 0.2712,
+      "step": 16640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.53564512758436e-05,
+      "loss": 0.0619,
+      "step": 16641
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.533998730644155e-05,
+      "loss": 0.6821,
+      "step": 16642
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.532352453260754e-05,
+      "loss": 0.3045,
+      "step": 16643
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.53070629546052e-05,
+      "loss": 0.7981,
+      "step": 16644
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.529060257269817e-05,
+      "loss": 0.1316,
+      "step": 16645
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.527414338715001e-05,
+      "loss": 0.2773,
+      "step": 16646
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.525768539822429e-05,
+      "loss": 0.0009,
+      "step": 16647
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.52412286061846e-05,
+      "loss": 0.4049,
+      "step": 16648
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.522477301129439e-05,
+      "loss": 0.1301,
+      "step": 16649
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.520831861381723e-05,
+      "loss": 0.5537,
+      "step": 16650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.51918654140166e-05,
+      "loss": 0.0099,
+      "step": 16651
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.517541341215599e-05,
+      "loss": 0.2764,
+      "step": 16652
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.515896260849884e-05,
+      "loss": 0.3003,
+      "step": 16653
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.514251300330855e-05,
+      "loss": 0.389,
+      "step": 16654
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.51260645968486e-05,
+      "loss": 0.2291,
+      "step": 16655
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.51096173893824e-05,
+      "loss": 0.0017,
+      "step": 16656
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.509317138117322e-05,
+      "loss": 0.0048,
+      "step": 16657
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.507672657248444e-05,
+      "loss": 0.0005,
+      "step": 16658
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.506028296357956e-05,
+      "loss": 0.0544,
+      "step": 16659
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.50438405547217e-05,
+      "loss": 0.9659,
+      "step": 16660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.50273993461742e-05,
+      "loss": 0.8132,
+      "step": 16661
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.501095933820049e-05,
+      "loss": 0.7734,
+      "step": 16662
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.499452053106365e-05,
+      "loss": 0.1552,
+      "step": 16663
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.497808292502697e-05,
+      "loss": 0.3282,
+      "step": 16664
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.496164652035379e-05,
+      "loss": 0.0007,
+      "step": 16665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.494521131730717e-05,
+      "loss": 1.0031,
+      "step": 16666
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.492877731615036e-05,
+      "loss": 0.2864,
+      "step": 16667
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.49123445171465e-05,
+      "loss": 0.2698,
+      "step": 16668
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.489591292055878e-05,
+      "loss": 0.4672,
+      "step": 16669
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.487948252665027e-05,
+      "loss": 0.4381,
+      "step": 16670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.486305333568411e-05,
+      "loss": 0.7098,
+      "step": 16671
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.48466253479234e-05,
+      "loss": 0.2321,
+      "step": 16672
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.483019856363118e-05,
+      "loss": 0.2423,
+      "step": 16673
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.481377298307054e-05,
+      "loss": 0.0015,
+      "step": 16674
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.479734860650453e-05,
+      "loss": 0.6057,
+      "step": 16675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.478092543419599e-05,
+      "loss": 0.1177,
+      "step": 16676
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.476450346640811e-05,
+      "loss": 0.0018,
+      "step": 16677
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.474808270340385e-05,
+      "loss": 0.4419,
+      "step": 16678
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.4731663145446e-05,
+      "loss": 1.8281,
+      "step": 16679
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.471524479279765e-05,
+      "loss": 0.5886,
+      "step": 16680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.469882764572174e-05,
+      "loss": 0.2189,
+      "step": 16681
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.468241170448096e-05,
+      "loss": 0.558,
+      "step": 16682
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.46659969693384e-05,
+      "loss": 0.4085,
+      "step": 16683
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.464958344055689e-05,
+      "loss": 0.1489,
+      "step": 16684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.463317111839909e-05,
+      "loss": 0.3275,
+      "step": 16685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.461676000312808e-05,
+      "loss": 0.3289,
+      "step": 16686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.460035009500644e-05,
+      "loss": 0.5694,
+      "step": 16687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.458394139429699e-05,
+      "loss": 0.5829,
+      "step": 16688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.456753390126266e-05,
+      "loss": 1.122,
+      "step": 16689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.455112761616598e-05,
+      "loss": 0.6485,
+      "step": 16690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.453472253926972e-05,
+      "loss": 5.0471,
+      "step": 16691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.451831867083674e-05,
+      "loss": 0.3801,
+      "step": 16692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.450191601112953e-05,
+      "loss": 0.842,
+      "step": 16693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.448551456041084e-05,
+      "loss": 0.3512,
+      "step": 16694
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.446911431894332e-05,
+      "loss": 0.1695,
+      "step": 16695
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.445271528698956e-05,
+      "loss": 0.2656,
+      "step": 16696
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.443631746481219e-05,
+      "loss": 0.5049,
+      "step": 16697
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.441992085267378e-05,
+      "loss": 0.2375,
+      "step": 16698
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.44035254508369e-05,
+      "loss": 0.0027,
+      "step": 16699
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.438713125956412e-05,
+      "loss": 0.0257,
+      "step": 16700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.437073827911797e-05,
+      "loss": 1.1036,
+      "step": 16701
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.435434650976092e-05,
+      "loss": 0.6794,
+      "step": 16702
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.433795595175549e-05,
+      "loss": 2.463,
+      "step": 16703
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.432156660536413e-05,
+      "loss": 0.4159,
+      "step": 16704
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.430517847084939e-05,
+      "loss": 0.0016,
+      "step": 16705
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.428879154847347e-05,
+      "loss": 0.4172,
+      "step": 16706
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.427240583849899e-05,
+      "loss": 0.0079,
+      "step": 16707
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.425602134118835e-05,
+      "loss": 0.1638,
+      "step": 16708
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.423963805680378e-05,
+      "loss": 0.0655,
+      "step": 16709
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.422325598560765e-05,
+      "loss": 0.0006,
+      "step": 16710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.420687512786248e-05,
+      "loss": 0.2988,
+      "step": 16711
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.41904954838304e-05,
+      "loss": 0.1939,
+      "step": 16712
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.417411705377368e-05,
+      "loss": 0.0007,
+      "step": 16713
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.415773983795482e-05,
+      "loss": 0.7568,
+      "step": 16714
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.414136383663585e-05,
+      "loss": 0.2327,
+      "step": 16715
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.41249890500791e-05,
+      "loss": 0.2843,
+      "step": 16716
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.410861547854678e-05,
+      "loss": 1.39,
+      "step": 16717
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.409224312230108e-05,
+      "loss": 0.8581,
+      "step": 16718
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.407587198160419e-05,
+      "loss": 0.3569,
+      "step": 16719
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.405950205671825e-05,
+      "loss": 1.627,
+      "step": 16720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.404313334790544e-05,
+      "loss": 1.1718,
+      "step": 16721
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.402676585542783e-05,
+      "loss": 0.0052,
+      "step": 16722
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.401039957954757e-05,
+      "loss": 0.0016,
+      "step": 16723
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.399403452052675e-05,
+      "loss": 0.003,
+      "step": 16724
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.39776706786273e-05,
+      "loss": 0.2847,
+      "step": 16725
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.39613080541114e-05,
+      "loss": 0.4637,
+      "step": 16726
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.394494664724111e-05,
+      "loss": 0.4866,
+      "step": 16727
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.392858645827824e-05,
+      "loss": 0.7371,
+      "step": 16728
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.391222748748492e-05,
+      "loss": 0.1964,
+      "step": 16729
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.389586973512319e-05,
+      "loss": 1.1237,
+      "step": 16730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.387951320145474e-05,
+      "loss": 0.5774,
+      "step": 16731
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.38631578867417e-05,
+      "loss": 0.3632,
+      "step": 16732
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.3846803791246e-05,
+      "loss": 0.7105,
+      "step": 16733
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.383045091522933e-05,
+      "loss": 0.0056,
+      "step": 16734
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.381409925895378e-05,
+      "loss": 0.0023,
+      "step": 16735
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.379774882268104e-05,
+      "loss": 0.2808,
+      "step": 16736
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.378139960667292e-05,
+      "loss": 0.0058,
+      "step": 16737
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.376505161119142e-05,
+      "loss": 0.5887,
+      "step": 16738
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.374870483649815e-05,
+      "loss": 0.4454,
+      "step": 16739
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.373235928285487e-05,
+      "loss": 0.0128,
+      "step": 16740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.371601495052351e-05,
+      "loss": 0.4163,
+      "step": 16741
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.369967183976564e-05,
+      "loss": 0.2469,
+      "step": 16742
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.368332995084301e-05,
+      "loss": 0.5507,
+      "step": 16743
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.36669892840173e-05,
+      "loss": 0.7256,
+      "step": 16744
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.365064983955021e-05,
+      "loss": 0.4782,
+      "step": 16745
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.363431161770338e-05,
+      "loss": 0.3705,
+      "step": 16746
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.361797461873844e-05,
+      "loss": 0.6715,
+      "step": 16747
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.360163884291699e-05,
+      "loss": 0.5029,
+      "step": 16748
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.358530429050064e-05,
+      "loss": 0.3668,
+      "step": 16749
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.356897096175097e-05,
+      "loss": 0.9064,
+      "step": 16750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.355263885692952e-05,
+      "loss": 0.1408,
+      "step": 16751
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.353630797629782e-05,
+      "loss": 0.2574,
+      "step": 16752
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.351997832011742e-05,
+      "loss": 0.0502,
+      "step": 16753
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.35036498886498e-05,
+      "loss": 0.0585,
+      "step": 16754
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.348732268215635e-05,
+      "loss": 0.5518,
+      "step": 16755
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.347099670089863e-05,
+      "loss": 0.2118,
+      "step": 16756
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.34546719451381e-05,
+      "loss": 1.1548,
+      "step": 16757
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.343834841513601e-05,
+      "loss": 0.2315,
+      "step": 16758
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.342202611115393e-05,
+      "loss": 0.0043,
+      "step": 16759
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.340570503345324e-05,
+      "loss": 0.4215,
+      "step": 16760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.338938518229516e-05,
+      "loss": 0.3146,
+      "step": 16761
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.337306655794103e-05,
+      "loss": 0.2613,
+      "step": 16762
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.335674916065238e-05,
+      "loss": 0.1791,
+      "step": 16763
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.33404329906903e-05,
+      "loss": 0.0015,
+      "step": 16764
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.332411804831612e-05,
+      "loss": 0.7647,
+      "step": 16765
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.330780433379113e-05,
+      "loss": 0.6087,
+      "step": 16766
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.329149184737656e-05,
+      "loss": 0.1687,
+      "step": 16767
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.327518058933363e-05,
+      "loss": 0.4723,
+      "step": 16768
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.32588705599235e-05,
+      "loss": 0.1469,
+      "step": 16769
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.324256175940743e-05,
+      "loss": 0.2508,
+      "step": 16770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.322625418804651e-05,
+      "loss": 0.4566,
+      "step": 16771
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.320994784610191e-05,
+      "loss": 0.0225,
+      "step": 16772
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.319364273383474e-05,
+      "loss": 0.9757,
+      "step": 16773
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.317733885150613e-05,
+      "loss": 0.7242,
+      "step": 16774
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.316103619937713e-05,
+      "loss": 0.3623,
+      "step": 16775
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.314473477770888e-05,
+      "loss": 0.8761,
+      "step": 16776
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.312843458676227e-05,
+      "loss": 0.0911,
+      "step": 16777
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.311213562679842e-05,
+      "loss": 0.3193,
+      "step": 16778
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.309583789807843e-05,
+      "loss": 0.5097,
+      "step": 16779
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.307954140086302e-05,
+      "loss": 0.2386,
+      "step": 16780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.306324613541339e-05,
+      "loss": 0.4086,
+      "step": 16781
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.304695210199047e-05,
+      "loss": 0.3414,
+      "step": 16782
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.303065930085501e-05,
+      "loss": 0.6978,
+      "step": 16783
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.301436773226814e-05,
+      "loss": 0.6561,
+      "step": 16784
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.299807739649056e-05,
+      "loss": 0.2465,
+      "step": 16785
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.298178829378317e-05,
+      "loss": 0.385,
+      "step": 16786
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.296550042440695e-05,
+      "loss": 0.2373,
+      "step": 16787
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.294921378862257e-05,
+      "loss": 0.4248,
+      "step": 16788
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.293292838669085e-05,
+      "loss": 0.0134,
+      "step": 16789
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.291664421887272e-05,
+      "loss": 0.295,
+      "step": 16790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.29003612854288e-05,
+      "loss": 1.2529,
+      "step": 16791
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.288407958661988e-05,
+      "loss": 0.4695,
+      "step": 16792
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.286779912270669e-05,
+      "loss": 0.001,
+      "step": 16793
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.285151989394995e-05,
+      "loss": 0.0021,
+      "step": 16794
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.283524190061033e-05,
+      "loss": 0.1768,
+      "step": 16795
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.28189651429485e-05,
+      "loss": 0.2128,
+      "step": 16796
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.280268962122511e-05,
+      "loss": 0.9696,
+      "step": 16797
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.278641533570077e-05,
+      "loss": 0.3616,
+      "step": 16798
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.277014228663612e-05,
+      "loss": 0.4233,
+      "step": 16799
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.275387047429175e-05,
+      "loss": 1.0066,
+      "step": 16800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.27375998989282e-05,
+      "loss": 0.0015,
+      "step": 16801
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.272133056080605e-05,
+      "loss": 0.4415,
+      "step": 16802
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.270506246018586e-05,
+      "loss": 0.3013,
+      "step": 16803
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.268879559732797e-05,
+      "loss": 0.2857,
+      "step": 16804
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.267252997249304e-05,
+      "loss": 0.4549,
+      "step": 16805
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.265626558594156e-05,
+      "loss": 0.1427,
+      "step": 16806
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.264000243793379e-05,
+      "loss": 0.1752,
+      "step": 16807
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.262374052873035e-05,
+      "loss": 0.2806,
+      "step": 16808
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.260747985859163e-05,
+      "loss": 0.1818,
+      "step": 16809
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.259122042777792e-05,
+      "loss": 0.2634,
+      "step": 16810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.257496223654955e-05,
+      "loss": 0.3786,
+      "step": 16811
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.255870528516712e-05,
+      "loss": 0.6522,
+      "step": 16812
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.254244957389071e-05,
+      "loss": 0.2237,
+      "step": 16813
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.252619510298074e-05,
+      "loss": 0.2372,
+      "step": 16814
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.250994187269748e-05,
+      "loss": 0.2312,
+      "step": 16815
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.249368988330121e-05,
+      "loss": 0.7704,
+      "step": 16816
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.247743913505218e-05,
+      "loss": 0.4446,
+      "step": 16817
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.246118962821059e-05,
+      "loss": 0.6713,
+      "step": 16818
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.24449413630367e-05,
+      "loss": 1.5778,
+      "step": 16819
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.242869433979068e-05,
+      "loss": 0.6808,
+      "step": 16820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.24124485587327e-05,
+      "loss": 0.4639,
+      "step": 16821
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.239620402012292e-05,
+      "loss": 1.2364,
+      "step": 16822
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.237996072422146e-05,
+      "loss": 0.9741,
+      "step": 16823
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.236371867128844e-05,
+      "loss": 0.1453,
+      "step": 16824
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 7.234747786158394e-05,
+      "loss": 0.3179,
+      "step": 16825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.233123829536805e-05,
+      "loss": 0.0032,
+      "step": 16826
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.23149999729008e-05,
+      "loss": 0.564,
+      "step": 16827
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.229876289444231e-05,
+      "loss": 0.1865,
+      "step": 16828
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.22825270602524e-05,
+      "loss": 0.0018,
+      "step": 16829
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.226629247059125e-05,
+      "loss": 0.0587,
+      "step": 16830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.22500591257188e-05,
+      "loss": 0.8682,
+      "step": 16831
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.223382702589487e-05,
+      "loss": 1.3514,
+      "step": 16832
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.221759617137959e-05,
+      "loss": 0.4952,
+      "step": 16833
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.220136656243272e-05,
+      "loss": 0.8294,
+      "step": 16834
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.218513819931415e-05,
+      "loss": 0.2006,
+      "step": 16835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.216891108228392e-05,
+      "loss": 0.3532,
+      "step": 16836
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.215268521160172e-05,
+      "loss": 0.0099,
+      "step": 16837
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.213646058752736e-05,
+      "loss": 0.3435,
+      "step": 16838
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.212023721032086e-05,
+      "loss": 0.1406,
+      "step": 16839
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.210401508024182e-05,
+      "loss": 0.2535,
+      "step": 16840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.208779419755007e-05,
+      "loss": 0.5434,
+      "step": 16841
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.207157456250537e-05,
+      "loss": 0.5809,
+      "step": 16842
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.205535617536744e-05,
+      "loss": 0.0065,
+      "step": 16843
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.203913903639601e-05,
+      "loss": 0.0829,
+      "step": 16844
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.202292314585076e-05,
+      "loss": 0.2272,
+      "step": 16845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.200670850399137e-05,
+      "loss": 1.6238,
+      "step": 16846
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.19904951110775e-05,
+      "loss": 0.7534,
+      "step": 16847
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.197428296736876e-05,
+      "loss": 0.2651,
+      "step": 16848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.195807207312479e-05,
+      "loss": 0.624,
+      "step": 16849
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.194186242860517e-05,
+      "loss": 0.0038,
+      "step": 16850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.192565403406947e-05,
+      "loss": 0.4847,
+      "step": 16851
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.190944688977729e-05,
+      "loss": 0.096,
+      "step": 16852
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.189324099598804e-05,
+      "loss": 1.9009,
+      "step": 16853
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.187703635296135e-05,
+      "loss": 0.1669,
+      "step": 16854
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.186083296095673e-05,
+      "loss": 0.7121,
+      "step": 16855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.18446308202335e-05,
+      "loss": 0.2243,
+      "step": 16856
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.182842993105125e-05,
+      "loss": 0.5743,
+      "step": 16857
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.181223029366947e-05,
+      "loss": 0.4636,
+      "step": 16858
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.179603190834732e-05,
+      "loss": 0.8017,
+      "step": 16859
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.177983477534442e-05,
+      "loss": 0.6101,
+      "step": 16860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.176363889492013e-05,
+      "loss": 0.2545,
+      "step": 16861
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.174744426733373e-05,
+      "loss": 0.2496,
+      "step": 16862
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.173125089284453e-05,
+      "loss": 0.1503,
+      "step": 16863
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.171505877171189e-05,
+      "loss": 0.3369,
+      "step": 16864
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.169886790419509e-05,
+      "loss": 0.2709,
+      "step": 16865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.168267829055341e-05,
+      "loss": 0.2803,
+      "step": 16866
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.16664899310461e-05,
+      "loss": 0.2981,
+      "step": 16867
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.165030282593239e-05,
+      "loss": 0.563,
+      "step": 16868
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.16341169754715e-05,
+      "loss": 0.3472,
+      "step": 16869
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.161793237992259e-05,
+      "loss": 0.2753,
+      "step": 16870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.16017490395449e-05,
+      "loss": 0.0011,
+      "step": 16871
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.158556695459753e-05,
+      "loss": 0.5791,
+      "step": 16872
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.156938612533961e-05,
+      "loss": 0.741,
+      "step": 16873
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.155320655203025e-05,
+      "loss": 0.4849,
+      "step": 16874
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.153702823492857e-05,
+      "loss": 0.2812,
+      "step": 16875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.152085117429362e-05,
+      "loss": 0.3739,
+      "step": 16876
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.150467537038452e-05,
+      "loss": 0.2167,
+      "step": 16877
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.148850082346014e-05,
+      "loss": 0.2916,
+      "step": 16878
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.147232753377962e-05,
+      "loss": 0.6207,
+      "step": 16879
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.145615550160201e-05,
+      "loss": 0.2816,
+      "step": 16880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.143998472718604e-05,
+      "loss": 2.3145,
+      "step": 16881
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.142381521079095e-05,
+      "loss": 1.9815,
+      "step": 16882
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.140764695267546e-05,
+      "loss": 0.2853,
+      "step": 16883
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.139147995309851e-05,
+      "loss": 0.4745,
+      "step": 16884
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.137531421231915e-05,
+      "loss": 0.315,
+      "step": 16885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.135914973059608e-05,
+      "loss": 0.1738,
+      "step": 16886
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.134298650818814e-05,
+      "loss": 0.2924,
+      "step": 16887
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.132682454535433e-05,
+      "loss": 0.3441,
+      "step": 16888
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.13106638423533e-05,
+      "loss": 1.7632,
+      "step": 16889
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.12945043994439e-05,
+      "loss": 0.7846,
+      "step": 16890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.127834621688489e-05,
+      "loss": 1.1202,
+      "step": 16891
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.126218929493502e-05,
+      "loss": 0.3776,
+      "step": 16892
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.1246033633853e-05,
+      "loss": 0.0028,
+      "step": 16893
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.122987923389758e-05,
+      "loss": 0.5889,
+      "step": 16894
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.121372609532742e-05,
+      "loss": 0.2944,
+      "step": 16895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.119757421840118e-05,
+      "loss": 0.7379,
+      "step": 16896
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.118142360337755e-05,
+      "loss": 0.5955,
+      "step": 16897
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.116527425051512e-05,
+      "loss": 0.0076,
+      "step": 16898
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.114912616007252e-05,
+      "loss": 0.0331,
+      "step": 16899
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.11329793323083e-05,
+      "loss": 0.5012,
+      "step": 16900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.111683376748114e-05,
+      "loss": 0.5553,
+      "step": 16901
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.110068946584936e-05,
+      "loss": 0.0867,
+      "step": 16902
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.108454642767169e-05,
+      "loss": 0.8552,
+      "step": 16903
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.106840465320664e-05,
+      "loss": 0.6014,
+      "step": 16904
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.105226414271251e-05,
+      "loss": 0.2592,
+      "step": 16905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.103612489644794e-05,
+      "loss": 0.7817,
+      "step": 16906
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.101998691467138e-05,
+      "loss": 0.7508,
+      "step": 16907
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.100385019764109e-05,
+      "loss": 0.0007,
+      "step": 16908
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.098771474561564e-05,
+      "loss": 0.2901,
+      "step": 16909
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.09715805588534e-05,
+      "loss": 0.3013,
+      "step": 16910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.095544763761258e-05,
+      "loss": 0.8384,
+      "step": 16911
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.093931598215176e-05,
+      "loss": 0.5421,
+      "step": 16912
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.09231855927291e-05,
+      "loss": 0.0867,
+      "step": 16913
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.090705646960292e-05,
+      "loss": 0.832,
+      "step": 16914
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.089092861303153e-05,
+      "loss": 0.0013,
+      "step": 16915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.08748020232732e-05,
+      "loss": 0.8023,
+      "step": 16916
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.085867670058617e-05,
+      "loss": 0.1044,
+      "step": 16917
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.084255264522865e-05,
+      "loss": 0.7181,
+      "step": 16918
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.082642985745886e-05,
+      "loss": 1.0421,
+      "step": 16919
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.081030833753496e-05,
+      "loss": 0.0603,
+      "step": 16920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.079418808571512e-05,
+      "loss": 1.9419,
+      "step": 16921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.077806910225752e-05,
+      "loss": 0.4277,
+      "step": 16922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.076195138742022e-05,
+      "loss": 0.0034,
+      "step": 16923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.074583494146135e-05,
+      "loss": 2.9688,
+      "step": 16924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.072971976463898e-05,
+      "loss": 0.0015,
+      "step": 16925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.07136058572112e-05,
+      "loss": 0.8081,
+      "step": 16926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.069749321943603e-05,
+      "loss": 0.2903,
+      "step": 16927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.068138185157148e-05,
+      "loss": 0.3478,
+      "step": 16928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.06652717538756e-05,
+      "loss": 0.2948,
+      "step": 16929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.064916292660624e-05,
+      "loss": 0.324,
+      "step": 16930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.063305537002157e-05,
+      "loss": 0.5122,
+      "step": 16931
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.061694908437933e-05,
+      "loss": 0.0022,
+      "step": 16932
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.060084406993746e-05,
+      "loss": 0.2045,
+      "step": 16933
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.058474032695401e-05,
+      "loss": 0.0422,
+      "step": 16934
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.056863785568672e-05,
+      "loss": 0.2579,
+      "step": 16935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.055253665639342e-05,
+      "loss": 0.156,
+      "step": 16936
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.053643672933212e-05,
+      "loss": 0.3858,
+      "step": 16937
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.05203380747605e-05,
+      "loss": 0.3989,
+      "step": 16938
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.050424069293635e-05,
+      "loss": 0.6441,
+      "step": 16939
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.048814458411749e-05,
+      "loss": 0.2898,
+      "step": 16940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.047204974856166e-05,
+      "loss": 0.2498,
+      "step": 16941
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.04559561865266e-05,
+      "loss": 0.3586,
+      "step": 16942
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.043986389827003e-05,
+      "loss": 0.3211,
+      "step": 16943
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.042377288404963e-05,
+      "loss": 0.3816,
+      "step": 16944
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.040768314412309e-05,
+      "loss": 0.5287,
+      "step": 16945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.039159467874804e-05,
+      "loss": 1.128,
+      "step": 16946
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.037550748818214e-05,
+      "loss": 0.5481,
+      "step": 16947
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.035942157268298e-05,
+      "loss": 0.0654,
+      "step": 16948
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.034333693250815e-05,
+      "loss": 0.2746,
+      "step": 16949
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.032725356791529e-05,
+      "loss": 0.4324,
+      "step": 16950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.031117147916179e-05,
+      "loss": 0.0155,
+      "step": 16951
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.029509066650533e-05,
+      "loss": 0.7261,
+      "step": 16952
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.027901113020343e-05,
+      "loss": 1.0607,
+      "step": 16953
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.026293287051342e-05,
+      "loss": 0.1798,
+      "step": 16954
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.024685588769293e-05,
+      "loss": 0.0487,
+      "step": 16955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.023078018199939e-05,
+      "loss": 0.7491,
+      "step": 16956
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.021470575369008e-05,
+      "loss": 0.0013,
+      "step": 16957
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.019863260302257e-05,
+      "loss": 0.0027,
+      "step": 16958
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.018256073025426e-05,
+      "loss": 0.9642,
+      "step": 16959
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.016649013564234e-05,
+      "loss": 0.6637,
+      "step": 16960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.015042081944437e-05,
+      "loss": 1.7961,
+      "step": 16961
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.013435278191752e-05,
+      "loss": 0.1166,
+      "step": 16962
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.011828602331915e-05,
+      "loss": 0.4254,
+      "step": 16963
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.010222054390656e-05,
+      "loss": 0.2361,
+      "step": 16964
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.008615634393698e-05,
+      "loss": 0.3328,
+      "step": 16965
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.007009342366767e-05,
+      "loss": 0.6234,
+      "step": 16966
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.005403178335584e-05,
+      "loss": 0.002,
+      "step": 16967
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.003797142325873e-05,
+      "loss": 2.9235,
+      "step": 16968
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.00219123436335e-05,
+      "loss": 0.7294,
+      "step": 16969
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 7.000585454473733e-05,
+      "loss": 0.2968,
+      "step": 16970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.998979802682733e-05,
+      "loss": 0.3242,
+      "step": 16971
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.997374279016063e-05,
+      "loss": 0.4458,
+      "step": 16972
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.995768883499434e-05,
+      "loss": 0.1959,
+      "step": 16973
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.994163616158554e-05,
+      "loss": 0.0092,
+      "step": 16974
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.992558477019129e-05,
+      "loss": 0.4313,
+      "step": 16975
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.990953466106862e-05,
+      "loss": 0.2391,
+      "step": 16976
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.989348583447454e-05,
+      "loss": 0.2439,
+      "step": 16977
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.987743829066608e-05,
+      "loss": 0.288,
+      "step": 16978
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.986139202990019e-05,
+      "loss": 0.0023,
+      "step": 16979
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.984534705243389e-05,
+      "loss": 0.1439,
+      "step": 16980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.982930335852401e-05,
+      "loss": 0.1779,
+      "step": 16981
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.981326094842748e-05,
+      "loss": 0.2359,
+      "step": 16982
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.979721982240131e-05,
+      "loss": 0.4137,
+      "step": 16983
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.978117998070225e-05,
+      "loss": 0.3144,
+      "step": 16984
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.976514142358716e-05,
+      "loss": 0.3253,
+      "step": 16985
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.974910415131302e-05,
+      "loss": 0.6925,
+      "step": 16986
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.973306816413648e-05,
+      "loss": 1.0186,
+      "step": 16987
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.971703346231438e-05,
+      "loss": 0.7713,
+      "step": 16988
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.97010000461035e-05,
+      "loss": 0.0088,
+      "step": 16989
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.968496791576061e-05,
+      "loss": 0.3864,
+      "step": 16990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.966893707154242e-05,
+      "loss": 0.1386,
+      "step": 16991
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.965290751370565e-05,
+      "loss": 0.2046,
+      "step": 16992
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.963687924250696e-05,
+      "loss": 0.9097,
+      "step": 16993
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.962085225820304e-05,
+      "loss": 0.7421,
+      "step": 16994
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.960482656105055e-05,
+      "loss": 0.2093,
+      "step": 16995
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.958880215130612e-05,
+      "loss": 0.0015,
+      "step": 16996
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.957277902922632e-05,
+      "loss": 0.2068,
+      "step": 16997
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.955675719506778e-05,
+      "loss": 1.3782,
+      "step": 16998
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.954073664908708e-05,
+      "loss": 0.3491,
+      "step": 16999
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.952471739154063e-05,
+      "loss": 1.2847,
+      "step": 17000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.95086994226851e-05,
+      "loss": 0.0971,
+      "step": 17001
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.949268274277703e-05,
+      "loss": 1.0214,
+      "step": 17002
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.947666735207267e-05,
+      "loss": 0.5395,
+      "step": 17003
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.946065325082872e-05,
+      "loss": 0.0014,
+      "step": 17004
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.944464043930159e-05,
+      "loss": 0.496,
+      "step": 17005
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.942862891774752e-05,
+      "loss": 0.0019,
+      "step": 17006
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.94126186864231e-05,
+      "loss": 0.4384,
+      "step": 17007
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.939660974558471e-05,
+      "loss": 1.5692,
+      "step": 17008
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.938060209548853e-05,
+      "loss": 0.2512,
+      "step": 17009
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.936459573639112e-05,
+      "loss": 0.7046,
+      "step": 17010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.934859066854865e-05,
+      "loss": 0.264,
+      "step": 17011
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.93325868922174e-05,
+      "loss": 1.6216,
+      "step": 17012
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.931658440765379e-05,
+      "loss": 0.3404,
+      "step": 17013
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.930058321511397e-05,
+      "loss": 0.5628,
+      "step": 17014
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.928458331485419e-05,
+      "loss": 0.3636,
+      "step": 17015
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.926858470713066e-05,
+      "loss": 0.2101,
+      "step": 17016
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.92525873921996e-05,
+      "loss": 0.3417,
+      "step": 17017
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.923659137031715e-05,
+      "loss": 0.0007,
+      "step": 17018
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.922059664173949e-05,
+      "loss": 0.8375,
+      "step": 17019
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.920460320672274e-05,
+      "loss": 0.1699,
+      "step": 17020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.918861106552301e-05,
+      "loss": 2.3058,
+      "step": 17021
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.917262021839642e-05,
+      "loss": 0.0344,
+      "step": 17022
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.9156630665599e-05,
+      "loss": 0.0019,
+      "step": 17023
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.914064240738679e-05,
+      "loss": 0.6095,
+      "step": 17024
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.912465544401585e-05,
+      "loss": 0.5431,
+      "step": 17025
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.910866977574218e-05,
+      "loss": 0.2893,
+      "step": 17026
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.909268540282176e-05,
+      "loss": 0.3387,
+      "step": 17027
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.907670232551055e-05,
+      "loss": 0.4535,
+      "step": 17028
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.906072054406457e-05,
+      "loss": 0.3612,
+      "step": 17029
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.904474005873962e-05,
+      "loss": 1.2695,
+      "step": 17030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.90287608697916e-05,
+      "loss": 0.244,
+      "step": 17031
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.901278297747657e-05,
+      "loss": 0.0009,
+      "step": 17032
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.899680638205022e-05,
+      "loss": 0.1753,
+      "step": 17033
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.89808310837684e-05,
+      "loss": 0.4202,
+      "step": 17034
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.896485708288707e-05,
+      "loss": 0.0071,
+      "step": 17035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.894888437966189e-05,
+      "loss": 0.3454,
+      "step": 17036
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.893291297434868e-05,
+      "loss": 0.7262,
+      "step": 17037
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.891694286720322e-05,
+      "loss": 0.2981,
+      "step": 17038
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.890097405848122e-05,
+      "loss": 0.3674,
+      "step": 17039
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.888500654843842e-05,
+      "loss": 0.7326,
+      "step": 17040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.886904033733049e-05,
+      "loss": 0.0007,
+      "step": 17041
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.885307542541312e-05,
+      "loss": 0.2576,
+      "step": 17042
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.883711181294195e-05,
+      "loss": 0.6697,
+      "step": 17043
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.882114950017266e-05,
+      "loss": 0.0941,
+      "step": 17044
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.88051884873608e-05,
+      "loss": 0.4652,
+      "step": 17045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.8789228774762e-05,
+      "loss": 0.5814,
+      "step": 17046
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.877327036263182e-05,
+      "loss": 0.3562,
+      "step": 17047
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.875731325122587e-05,
+      "loss": 0.3082,
+      "step": 17048
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.87413574407995e-05,
+      "loss": 0.3943,
+      "step": 17049
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.87254029316084e-05,
+      "loss": 0.5064,
+      "step": 17050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.870944972390807e-05,
+      "loss": 0.4706,
+      "step": 17051
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.869349781795376e-05,
+      "loss": 2.6371,
+      "step": 17052
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.867754721400112e-05,
+      "loss": 0.5044,
+      "step": 17053
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.866159791230559e-05,
+      "loss": 0.1665,
+      "step": 17054
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.864564991312236e-05,
+      "loss": 0.2234,
+      "step": 17055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.862970321670702e-05,
+      "loss": 0.3295,
+      "step": 17056
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.861375782331491e-05,
+      "loss": 1.1638,
+      "step": 17057
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.859781373320123e-05,
+      "loss": 0.0332,
+      "step": 17058
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.85818709466215e-05,
+      "loss": 0.4525,
+      "step": 17059
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.856592946383086e-05,
+      "loss": 0.4332,
+      "step": 17060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.854998928508458e-05,
+      "loss": 0.2398,
+      "step": 17061
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.85340504106381e-05,
+      "loss": 0.7056,
+      "step": 17062
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.851811284074647e-05,
+      "loss": 0.0013,
+      "step": 17063
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.850217657566494e-05,
+      "loss": 0.3536,
+      "step": 17064
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.848624161564884e-05,
+      "loss": 0.4295,
+      "step": 17065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.84703079609532e-05,
+      "loss": 0.898,
+      "step": 17066
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.845437561183321e-05,
+      "loss": 0.0009,
+      "step": 17067
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.843844456854403e-05,
+      "loss": 0.2916,
+      "step": 17068
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.842251483134074e-05,
+      "loss": 4.4304,
+      "step": 17069
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.840658640047844e-05,
+      "loss": 0.001,
+      "step": 17070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.839065927621221e-05,
+      "loss": 0.366,
+      "step": 17071
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.837473345879708e-05,
+      "loss": 0.6543,
+      "step": 17072
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.835880894848811e-05,
+      "loss": 0.5059,
+      "step": 17073
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.834288574554029e-05,
+      "loss": 0.0474,
+      "step": 17074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.832696385020859e-05,
+      "loss": 2.0586,
+      "step": 17075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.831104326274799e-05,
+      "loss": 0.0441,
+      "step": 17076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.829512398341343e-05,
+      "loss": 0.8394,
+      "step": 17077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.827920601245991e-05,
+      "loss": 1.1394,
+      "step": 17078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.826328935014215e-05,
+      "loss": 0.8242,
+      "step": 17079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.824737399671519e-05,
+      "loss": 0.2065,
+      "step": 17080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.823145995243389e-05,
+      "loss": 0.2083,
+      "step": 17081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.821554721755299e-05,
+      "loss": 0.4332,
+      "step": 17082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.81996357923273e-05,
+      "loss": 0.3352,
+      "step": 17083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.818372567701182e-05,
+      "loss": 0.0003,
+      "step": 17084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.81678168718611e-05,
+      "loss": 1.2891,
+      "step": 17085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.815190937712997e-05,
+      "loss": 1.2182,
+      "step": 17086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.813600319307319e-05,
+      "loss": 0.5437,
+      "step": 17087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.812009831994545e-05,
+      "loss": 0.6677,
+      "step": 17088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.810419475800147e-05,
+      "loss": 0.6279,
+      "step": 17089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.808829250749589e-05,
+      "loss": 0.0158,
+      "step": 17090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.807239156868337e-05,
+      "loss": 0.3603,
+      "step": 17091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.805649194181855e-05,
+      "loss": 0.0554,
+      "step": 17092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.804059362715602e-05,
+      "loss": 0.6375,
+      "step": 17093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.80246966249504e-05,
+      "loss": 0.2841,
+      "step": 17094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.800880093545621e-05,
+      "loss": 0.7427,
+      "step": 17095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.799290655892803e-05,
+      "loss": 0.1843,
+      "step": 17096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.797701349562044e-05,
+      "loss": 0.0078,
+      "step": 17097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.796112174578776e-05,
+      "loss": 0.9667,
+      "step": 17098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.794523130968466e-05,
+      "loss": 0.6581,
+      "step": 17099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.792934218756559e-05,
+      "loss": 0.5097,
+      "step": 17100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.791345437968481e-05,
+      "loss": 0.426,
+      "step": 17101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.789756788629691e-05,
+      "loss": 0.2244,
+      "step": 17102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.788168270765633e-05,
+      "loss": 0.8614,
+      "step": 17103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.786579884401721e-05,
+      "loss": 0.2094,
+      "step": 17104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.784991629563414e-05,
+      "loss": 0.4944,
+      "step": 17105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.783403506276141e-05,
+      "loss": 1.0976,
+      "step": 17106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.78181551456532e-05,
+      "loss": 0.446,
+      "step": 17107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.780227654456398e-05,
+      "loss": 0.0024,
+      "step": 17108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.778639925974791e-05,
+      "loss": 0.748,
+      "step": 17109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.77705232914592e-05,
+      "loss": 0.2022,
+      "step": 17110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.775464863995227e-05,
+      "loss": 0.6789,
+      "step": 17111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.773877530548116e-05,
+      "loss": 0.5713,
+      "step": 17112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.772290328830003e-05,
+      "loss": 0.5702,
+      "step": 17113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.770703258866326e-05,
+      "loss": 0.165,
+      "step": 17114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.769116320682479e-05,
+      "loss": 0.4448,
+      "step": 17115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.767529514303881e-05,
+      "loss": 0.2109,
+      "step": 17116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.765942839755943e-05,
+      "loss": 0.5512,
+      "step": 17117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.764356297064073e-05,
+      "loss": 0.8918,
+      "step": 17118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.762769886253676e-05,
+      "loss": 0.6836,
+      "step": 17119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.761183607350158e-05,
+      "loss": 0.2165,
+      "step": 17120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.75959746037892e-05,
+      "loss": 0.3934,
+      "step": 17121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.75801144536536e-05,
+      "loss": 0.3903,
+      "step": 17122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.75642556233488e-05,
+      "loss": 0.3805,
+      "step": 17123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.75483981131287e-05,
+      "loss": 0.2746,
+      "step": 17124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.753254192324727e-05,
+      "loss": 0.1622,
+      "step": 17125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.751668705395841e-05,
+      "loss": 0.0006,
+      "step": 17126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.750083350551607e-05,
+      "loss": 0.3643,
+      "step": 17127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.748498127817399e-05,
+      "loss": 0.5607,
+      "step": 17128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.746913037218611e-05,
+      "loss": 1.1955,
+      "step": 17129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.745328078780632e-05,
+      "loss": 0.4383,
+      "step": 17130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.743743252528825e-05,
+      "loss": 0.4534,
+      "step": 17131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.742158558488583e-05,
+      "loss": 0.2566,
+      "step": 17132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.740573996685284e-05,
+      "loss": 0.143,
+      "step": 17133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.738989567144293e-05,
+      "loss": 0.2274,
+      "step": 17134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.737405269890983e-05,
+      "loss": 0.8791,
+      "step": 17135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.73582110495073e-05,
+      "loss": 0.3258,
+      "step": 17136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.734237072348897e-05,
+      "loss": 0.3984,
+      "step": 17137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.732653172110853e-05,
+      "loss": 0.4737,
+      "step": 17138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.731069404261962e-05,
+      "loss": 1.5155,
+      "step": 17139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.729485768827584e-05,
+      "loss": 0.2497,
+      "step": 17140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.727902265833078e-05,
+      "loss": 0.8398,
+      "step": 17141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.726318895303803e-05,
+      "loss": 0.3257,
+      "step": 17142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.724735657265115e-05,
+      "loss": 0.2116,
+      "step": 17143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.723152551742365e-05,
+      "loss": 0.4111,
+      "step": 17144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.721569578760905e-05,
+      "loss": 1.0613,
+      "step": 17145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.719986738346084e-05,
+      "loss": 0.4083,
+      "step": 17146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.71840403052325e-05,
+      "loss": 0.001,
+      "step": 17147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.716821455317745e-05,
+      "loss": 0.2164,
+      "step": 17148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.715239012754919e-05,
+      "loss": 0.6543,
+      "step": 17149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.713656702860096e-05,
+      "loss": 0.3756,
+      "step": 17150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.712074525658631e-05,
+      "loss": 3.3333,
+      "step": 17151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.710492481175859e-05,
+      "loss": 0.3068,
+      "step": 17152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.708910569437096e-05,
+      "loss": 0.2933,
+      "step": 17153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.707328790467696e-05,
+      "loss": 0.2621,
+      "step": 17154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.705747144292983e-05,
+      "loss": 0.4566,
+      "step": 17155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.70416563093827e-05,
+      "loss": 1.0368,
+      "step": 17156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.702584250428907e-05,
+      "loss": 0.0087,
+      "step": 17157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.701003002790197e-05,
+      "loss": 0.0021,
+      "step": 17158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.699421888047465e-05,
+      "loss": 0.0011,
+      "step": 17159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.697840906226045e-05,
+      "loss": 2.6836,
+      "step": 17160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.696260057351238e-05,
+      "loss": 1.1861,
+      "step": 17161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.694679341448357e-05,
+      "loss": 1.2049,
+      "step": 17162
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.693098758542733e-05,
+      "loss": 0.2856,
+      "step": 17163
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.691518308659662e-05,
+      "loss": 0.4106,
+      "step": 17164
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.689937991824454e-05,
+      "loss": 0.6867,
+      "step": 17165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.688357808062418e-05,
+      "loss": 0.4336,
+      "step": 17166
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.686777757398859e-05,
+      "loss": 0.3912,
+      "step": 17167
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.685197839859076e-05,
+      "loss": 0.176,
+      "step": 17168
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.683618055468371e-05,
+      "loss": 0.318,
+      "step": 17169
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.682038404252041e-05,
+      "loss": 0.1724,
+      "step": 17170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.680458886235384e-05,
+      "loss": 0.6944,
+      "step": 17171
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.678879501443692e-05,
+      "loss": 0.3197,
+      "step": 17172
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.677300249902256e-05,
+      "loss": 0.0163,
+      "step": 17173
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.675721131636366e-05,
+      "loss": 0.7047,
+      "step": 17174
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.67414214667131e-05,
+      "loss": 0.773,
+      "step": 17175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.672563295032375e-05,
+      "loss": 0.5251,
+      "step": 17176
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.670984576744833e-05,
+      "loss": 0.0237,
+      "step": 17177
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.669405991833976e-05,
+      "loss": 0.6808,
+      "step": 17178
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.667827540325086e-05,
+      "loss": 0.2303,
+      "step": 17179
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.666249222243422e-05,
+      "loss": 0.3408,
+      "step": 17180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.664671037614277e-05,
+      "loss": 0.0024,
+      "step": 17181
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.66309298646292e-05,
+      "loss": 0.7353,
+      "step": 17182
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.66151506881461e-05,
+      "loss": 0.0023,
+      "step": 17183
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.659937284694623e-05,
+      "loss": 0.0153,
+      "step": 17184
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.658359634128223e-05,
+      "loss": 2.5028,
+      "step": 17185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.656782117140675e-05,
+      "loss": 1.1471,
+      "step": 17186
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.65520473375724e-05,
+      "loss": 0.4003,
+      "step": 17187
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.653627484003177e-05,
+      "loss": 0.1244,
+      "step": 17188
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.652050367903744e-05,
+      "loss": 0.401,
+      "step": 17189
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.650473385484196e-05,
+      "loss": 0.2354,
+      "step": 17190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.648896536769787e-05,
+      "loss": 0.01,
+      "step": 17191
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.647319821785765e-05,
+      "loss": 0.7817,
+      "step": 17192
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.64574324055738e-05,
+      "loss": 0.6949,
+      "step": 17193
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.644166793109881e-05,
+      "loss": 0.7205,
+      "step": 17194
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.642590479468509e-05,
+      "loss": 0.3987,
+      "step": 17195
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.641014299658508e-05,
+      "loss": 0.137,
+      "step": 17196
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.639438253705118e-05,
+      "loss": 0.2771,
+      "step": 17197
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.637862341633577e-05,
+      "loss": 0.0023,
+      "step": 17198
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.63628656346912e-05,
+      "loss": 0.4894,
+      "step": 17199
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.634710919236982e-05,
+      "loss": 0.0055,
+      "step": 17200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.6331354089624e-05,
+      "loss": 1.0042,
+      "step": 17201
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.631560032670586e-05,
+      "loss": 0.0032,
+      "step": 17202
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.629984790386784e-05,
+      "loss": 0.5792,
+      "step": 17203
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.628409682136221e-05,
+      "loss": 0.3372,
+      "step": 17204
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.626834707944102e-05,
+      "loss": 0.0036,
+      "step": 17205
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.625259867835667e-05,
+      "loss": 1.6268,
+      "step": 17206
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.623685161836123e-05,
+      "loss": 0.2523,
+      "step": 17207
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.622110589970685e-05,
+      "loss": 0.4306,
+      "step": 17208
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.620536152264584e-05,
+      "loss": 0.6561,
+      "step": 17209
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.618961848743013e-05,
+      "loss": 0.5007,
+      "step": 17210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.617387679431186e-05,
+      "loss": 1.9268,
+      "step": 17211
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.615813644354323e-05,
+      "loss": 0.0039,
+      "step": 17212
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.61423974353762e-05,
+      "loss": 0.3626,
+      "step": 17213
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.61266597700628e-05,
+      "loss": 1.023,
+      "step": 17214
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.611092344785508e-05,
+      "loss": 0.2371,
+      "step": 17215
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.6095188469005e-05,
+      "loss": 0.9191,
+      "step": 17216
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.607945483376457e-05,
+      "loss": 0.8314,
+      "step": 17217
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.606372254238573e-05,
+      "loss": 0.3495,
+      "step": 17218
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.604799159512037e-05,
+      "loss": 0.002,
+      "step": 17219
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.603226199222046e-05,
+      "loss": 0.2846,
+      "step": 17220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.601653373393785e-05,
+      "loss": 0.2232,
+      "step": 17221
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.60008068205244e-05,
+      "loss": 0.3464,
+      "step": 17222
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.598508125223198e-05,
+      "loss": 0.0041,
+      "step": 17223
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.59693570293124e-05,
+      "loss": 0.6642,
+      "step": 17224
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.595363415201749e-05,
+      "loss": 0.7304,
+      "step": 17225
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.59379126205989e-05,
+      "loss": 0.0178,
+      "step": 17226
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.592219243530853e-05,
+      "loss": 0.1427,
+      "step": 17227
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.59064735963981e-05,
+      "loss": 0.133,
+      "step": 17228
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.58907561041192e-05,
+      "loss": 0.2763,
+      "step": 17229
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.587503995872367e-05,
+      "loss": 0.343,
+      "step": 17230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.585932516046316e-05,
+      "loss": 0.1993,
+      "step": 17231
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.584361170958918e-05,
+      "loss": 0.394,
+      "step": 17232
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.582789960635357e-05,
+      "loss": 0.8662,
+      "step": 17233
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.581218885100775e-05,
+      "loss": 0.2689,
+      "step": 17234
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.579647944380339e-05,
+      "loss": 0.002,
+      "step": 17235
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.578077138499204e-05,
+      "loss": 2.3653,
+      "step": 17236
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.576506467482525e-05,
+      "loss": 1.069,
+      "step": 17237
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.574935931355452e-05,
+      "loss": 0.3385,
+      "step": 17238
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.573365530143135e-05,
+      "loss": 0.327,
+      "step": 17239
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.571795263870724e-05,
+      "loss": 0.3835,
+      "step": 17240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.570225132563362e-05,
+      "loss": 0.4082,
+      "step": 17241
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.568655136246193e-05,
+      "loss": 0.796,
+      "step": 17242
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.56708527494436e-05,
+      "loss": 0.0028,
+      "step": 17243
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.565515548682999e-05,
+      "loss": 0.8112,
+      "step": 17244
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.563945957487247e-05,
+      "loss": 0.4027,
+      "step": 17245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.562376501382241e-05,
+      "loss": 0.1454,
+      "step": 17246
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.560807180393111e-05,
+      "loss": 0.592,
+      "step": 17247
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.559237994544992e-05,
+      "loss": 0.5811,
+      "step": 17248
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.557668943863004e-05,
+      "loss": 0.0008,
+      "step": 17249
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.556100028372287e-05,
+      "loss": 0.2253,
+      "step": 17250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.554531248097943e-05,
+      "loss": 0.3257,
+      "step": 17251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.552962603065112e-05,
+      "loss": 0.0016,
+      "step": 17252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.551394093298911e-05,
+      "loss": 0.0564,
+      "step": 17253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.549825718824446e-05,
+      "loss": 0.5061,
+      "step": 17254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.548257479666851e-05,
+      "loss": 0.2782,
+      "step": 17255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.546689375851221e-05,
+      "loss": 0.2665,
+      "step": 17256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.545121407402669e-05,
+      "loss": 0.4478,
+      "step": 17257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.543553574346321e-05,
+      "loss": 0.6298,
+      "step": 17258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.541985876707265e-05,
+      "loss": 0.8975,
+      "step": 17259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.540418314510604e-05,
+      "loss": 0.2483,
+      "step": 17260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.53885088778146e-05,
+      "loss": 0.5802,
+      "step": 17261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.537283596544916e-05,
+      "loss": 0.2212,
+      "step": 17262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.535716440826075e-05,
+      "loss": 0.2273,
+      "step": 17263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.53414942065003e-05,
+      "loss": 0.6507,
+      "step": 17264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.532582536041877e-05,
+      "loss": 0.44,
+      "step": 17265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.531015787026707e-05,
+      "loss": 0.0022,
+      "step": 17266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.529449173629611e-05,
+      "loss": 0.282,
+      "step": 17267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.52788269587567e-05,
+      "loss": 0.0123,
+      "step": 17268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.526316353789976e-05,
+      "loss": 0.1891,
+      "step": 17269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.524750147397608e-05,
+      "loss": 0.1628,
+      "step": 17270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.523184076723646e-05,
+      "loss": 0.3351,
+      "step": 17271
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.52161814179317e-05,
+      "loss": 0.4518,
+      "step": 17272
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.520052342631253e-05,
+      "loss": 0.4298,
+      "step": 17273
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.51848667926298e-05,
+      "loss": 0.4775,
+      "step": 17274
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.5169211517134e-05,
+      "loss": 0.0065,
+      "step": 17275
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.515355760007603e-05,
+      "loss": 0.0036,
+      "step": 17276
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.513790504170656e-05,
+      "loss": 0.1454,
+      "step": 17277
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.512225384227606e-05,
+      "loss": 0.2265,
+      "step": 17278
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.510660400203534e-05,
+      "loss": 0.2111,
+      "step": 17279
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.5090955521235e-05,
+      "loss": 0.0012,
+      "step": 17280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.507530840012547e-05,
+      "loss": 0.3117,
+      "step": 17281
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.505966263895754e-05,
+      "loss": 0.0252,
+      "step": 17282
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.504401823798158e-05,
+      "loss": 0.8384,
+      "step": 17283
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.502837519744809e-05,
+      "loss": 0.5618,
+      "step": 17284
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.501273351760779e-05,
+      "loss": 0.9053,
+      "step": 17285
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.499709319871093e-05,
+      "loss": 0.4139,
+      "step": 17286
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.498145424100807e-05,
+      "loss": 0.0055,
+      "step": 17287
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.496581664474961e-05,
+      "loss": 0.5055,
+      "step": 17288
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.495018041018597e-05,
+      "loss": 0.02,
+      "step": 17289
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.493454553756758e-05,
+      "loss": 0.3348,
+      "step": 17290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.491891202714477e-05,
+      "loss": 0.7649,
+      "step": 17291
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.490327987916787e-05,
+      "loss": 0.0063,
+      "step": 17292
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.488764909388726e-05,
+      "loss": 0.7241,
+      "step": 17293
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.487201967155319e-05,
+      "loss": 0.1478,
+      "step": 17294
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.485639161241597e-05,
+      "loss": 0.4909,
+      "step": 17295
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.484076491672588e-05,
+      "loss": 0.2705,
+      "step": 17296
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.482513958473313e-05,
+      "loss": 0.2407,
+      "step": 17297
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.48095156166879e-05,
+      "loss": 0.6685,
+      "step": 17298
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.479389301284046e-05,
+      "loss": 0.62,
+      "step": 17299
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.477827177344095e-05,
+      "loss": 0.2011,
+      "step": 17300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.47626518987395e-05,
+      "loss": 3.0157,
+      "step": 17301
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.474703338898631e-05,
+      "loss": 1.0913,
+      "step": 17302
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.473141624443134e-05,
+      "loss": 0.2071,
+      "step": 17303
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.471580046532486e-05,
+      "loss": 0.6393,
+      "step": 17304
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.470018605191681e-05,
+      "loss": 0.3254,
+      "step": 17305
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.468457300445719e-05,
+      "loss": 1.0099,
+      "step": 17306
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.46689613231962e-05,
+      "loss": 0.256,
+      "step": 17307
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.465335100838367e-05,
+      "loss": 0.5163,
+      "step": 17308
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.463774206026962e-05,
+      "loss": 1.155,
+      "step": 17309
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.462213447910409e-05,
+      "loss": 0.0059,
+      "step": 17310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.46065282651369e-05,
+      "loss": 0.2258,
+      "step": 17311
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.459092341861801e-05,
+      "loss": 0.234,
+      "step": 17312
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.45753199397973e-05,
+      "loss": 0.3995,
+      "step": 17313
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.455971782892463e-05,
+      "loss": 1.2074,
+      "step": 17314
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.454411708624986e-05,
+      "loss": 0.673,
+      "step": 17315
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.452851771202279e-05,
+      "loss": 0.0196,
+      "step": 17316
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.451291970649322e-05,
+      "loss": 0.014,
+      "step": 17317
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.449732306991097e-05,
+      "loss": 0.4561,
+      "step": 17318
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.448172780252576e-05,
+      "loss": 0.5993,
+      "step": 17319
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.446613390458733e-05,
+      "loss": 2.1856,
+      "step": 17320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.445054137634541e-05,
+      "loss": 0.3052,
+      "step": 17321
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.443495021804968e-05,
+      "loss": 0.176,
+      "step": 17322
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.441936042994984e-05,
+      "loss": 0.1586,
+      "step": 17323
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.440377201229541e-05,
+      "loss": 0.4307,
+      "step": 17324
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.438818496533616e-05,
+      "loss": 0.228,
+      "step": 17325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.43725992893217e-05,
+      "loss": 0.4475,
+      "step": 17326
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.435701498450145e-05,
+      "loss": 0.5209,
+      "step": 17327
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.434143205112511e-05,
+      "loss": 0.0184,
+      "step": 17328
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.432585048944226e-05,
+      "loss": 0.412,
+      "step": 17329
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.431027029970222e-05,
+      "loss": 0.6734,
+      "step": 17330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.429469148215469e-05,
+      "loss": 0.034,
+      "step": 17331
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.427911403704901e-05,
+      "loss": 0.3032,
+      "step": 17332
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.426353796463461e-05,
+      "loss": 0.0015,
+      "step": 17333
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.424796326516107e-05,
+      "loss": 0.083,
+      "step": 17334
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.423238993887766e-05,
+      "loss": 0.7236,
+      "step": 17335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.421681798603382e-05,
+      "loss": 0.3204,
+      "step": 17336
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.42012474068789e-05,
+      "loss": 0.3448,
+      "step": 17337
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.418567820166221e-05,
+      "loss": 0.1767,
+      "step": 17338
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.417011037063311e-05,
+      "loss": 1.8076,
+      "step": 17339
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.415454391404089e-05,
+      "loss": 0.8114,
+      "step": 17340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.413897883213479e-05,
+      "loss": 0.3556,
+      "step": 17341
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.412341512516411e-05,
+      "loss": 0.1319,
+      "step": 17342
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.410785279337804e-05,
+      "loss": 0.8298,
+      "step": 17343
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.409229183702579e-05,
+      "loss": 0.0218,
+      "step": 17344
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.407673225635657e-05,
+      "loss": 0.773,
+      "step": 17345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.406117405161953e-05,
+      "loss": 1.7222,
+      "step": 17346
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.404561722306381e-05,
+      "loss": 0.312,
+      "step": 17347
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.403006177093854e-05,
+      "loss": 0.3671,
+      "step": 17348
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.40145076954928e-05,
+      "loss": 0.0453,
+      "step": 17349
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.399895499697567e-05,
+      "loss": 0.0085,
+      "step": 17350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.398340367563621e-05,
+      "loss": 1.0017,
+      "step": 17351
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.396785373172344e-05,
+      "loss": 0.0011,
+      "step": 17352
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.395230516548644e-05,
+      "loss": 0.3436,
+      "step": 17353
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.393675797717407e-05,
+      "loss": 0.8272,
+      "step": 17354
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.392121216703531e-05,
+      "loss": 0.0119,
+      "step": 17355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.390566773531927e-05,
+      "loss": 0.0591,
+      "step": 17356
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.389012468227469e-05,
+      "loss": 0.7869,
+      "step": 17357
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.387458300815047e-05,
+      "loss": 0.2038,
+      "step": 17358
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.385904271319565e-05,
+      "loss": 0.7756,
+      "step": 17359
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.384350379765894e-05,
+      "loss": 0.4135,
+      "step": 17360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.38279662617892e-05,
+      "loss": 0.0478,
+      "step": 17361
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.381243010583527e-05,
+      "loss": 0.004,
+      "step": 17362
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.379689533004593e-05,
+      "loss": 0.3049,
+      "step": 17363
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.37813619346699e-05,
+      "loss": 0.6001,
+      "step": 17364
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.3765829919956e-05,
+      "loss": 0.6606,
+      "step": 17365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.37502992861529e-05,
+      "loss": 0.9683,
+      "step": 17366
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.373477003350932e-05,
+      "loss": 0.3206,
+      "step": 17367
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.371924216227393e-05,
+      "loss": 0.2941,
+      "step": 17368
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.370371567269538e-05,
+      "loss": 0.0028,
+      "step": 17369
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.36881905650223e-05,
+      "loss": 0.6623,
+      "step": 17370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.367266683950334e-05,
+      "loss": 1.076,
+      "step": 17371
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.365714449638711e-05,
+      "loss": 2.3828,
+      "step": 17372
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.364162353592202e-05,
+      "loss": 1.0843,
+      "step": 17373
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.362610395835677e-05,
+      "loss": 0.1346,
+      "step": 17374
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.361058576393989e-05,
+      "loss": 0.5717,
+      "step": 17375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.359506895291971e-05,
+      "loss": 0.276,
+      "step": 17376
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.35795535255449e-05,
+      "loss": 0.1973,
+      "step": 17377
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.356403948206389e-05,
+      "loss": 0.7656,
+      "step": 17378
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.354852682272494e-05,
+      "loss": 0.1278,
+      "step": 17379
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.35330155477767e-05,
+      "loss": 0.3263,
+      "step": 17380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.35175056574674e-05,
+      "loss": 0.3392,
+      "step": 17381
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.35019971520454e-05,
+      "loss": 0.6353,
+      "step": 17382
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.348649003175922e-05,
+      "loss": 0.7192,
+      "step": 17383
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.3470984296857e-05,
+      "loss": 0.2455,
+      "step": 17384
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.345547994758704e-05,
+      "loss": 0.2519,
+      "step": 17385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.343997698419781e-05,
+      "loss": 1.7512,
+      "step": 17386
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.342447540693738e-05,
+      "loss": 0.9007,
+      "step": 17387
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.340897521605404e-05,
+      "loss": 0.1555,
+      "step": 17388
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.339347641179601e-05,
+      "loss": 0.501,
+      "step": 17389
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.337797899441147e-05,
+      "loss": 0.1756,
+      "step": 17390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.336248296414861e-05,
+      "loss": 0.4868,
+      "step": 17391
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.334698832125555e-05,
+      "loss": 0.399,
+      "step": 17392
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.333149506598043e-05,
+      "loss": 0.0715,
+      "step": 17393
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.331600319857133e-05,
+      "loss": 0.9257,
+      "step": 17394
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.330051271927637e-05,
+      "loss": 1.1034,
+      "step": 17395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.328502362834357e-05,
+      "loss": 0.2623,
+      "step": 17396
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.326953592602096e-05,
+      "loss": 0.2448,
+      "step": 17397
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.325404961255658e-05,
+      "loss": 0.2194,
+      "step": 17398
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.323856468819842e-05,
+      "loss": 0.0895,
+      "step": 17399
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.322308115319441e-05,
+      "loss": 0.2626,
+      "step": 17400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.320759900779254e-05,
+      "loss": 1.1613,
+      "step": 17401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.319211825224076e-05,
+      "loss": 0.7972,
+      "step": 17402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.317663888678688e-05,
+      "loss": 0.188,
+      "step": 17403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.316116091167877e-05,
+      "loss": 1.2905,
+      "step": 17404
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.314568432716444e-05,
+      "loss": 0.0023,
+      "step": 17405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.313020913349157e-05,
+      "loss": 0.0021,
+      "step": 17406
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.3114735330908e-05,
+      "loss": 0.16,
+      "step": 17407
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.309926291966163e-05,
+      "loss": 0.1551,
+      "step": 17408
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.30837919000001e-05,
+      "loss": 0.0911,
+      "step": 17409
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.30683222721712e-05,
+      "loss": 0.7802,
+      "step": 17410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.305285403642264e-05,
+      "loss": 0.7735,
+      "step": 17411
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.303738719300213e-05,
+      "loss": 0.0043,
+      "step": 17412
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.302192174215738e-05,
+      "loss": 0.3045,
+      "step": 17413
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.300645768413599e-05,
+      "loss": 0.5393,
+      "step": 17414
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.299099501918564e-05,
+      "loss": 0.7203,
+      "step": 17415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.297553374755391e-05,
+      "loss": 1.0093,
+      "step": 17416
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.296007386948841e-05,
+      "loss": 0.3304,
+      "step": 17417
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.29446153852367e-05,
+      "loss": 0.0037,
+      "step": 17418
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.292915829504631e-05,
+      "loss": 0.6302,
+      "step": 17419
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.291370259916477e-05,
+      "loss": 0.1454,
+      "step": 17420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.289824829783967e-05,
+      "loss": 0.6007,
+      "step": 17421
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.288279539131826e-05,
+      "loss": 0.1947,
+      "step": 17422
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.286734387984821e-05,
+      "loss": 0.3688,
+      "step": 17423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.285189376367694e-05,
+      "loss": 0.3902,
+      "step": 17424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.283644504305169e-05,
+      "loss": 0.6832,
+      "step": 17425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.282099771822e-05,
+      "loss": 2.2813,
+      "step": 17426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.280555178942926e-05,
+      "loss": 0.0123,
+      "step": 17427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.279010725692662e-05,
+      "loss": 0.9662,
+      "step": 17428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.277466412095965e-05,
+      "loss": 0.2865,
+      "step": 17429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.275922238177548e-05,
+      "loss": 0.2683,
+      "step": 17430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.274378203962135e-05,
+      "loss": 0.7602,
+      "step": 17431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.272834309474475e-05,
+      "loss": 0.4743,
+      "step": 17432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.271290554739266e-05,
+      "loss": 1.5099,
+      "step": 17433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.269746939781237e-05,
+      "loss": 0.5866,
+      "step": 17434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.268203464625118e-05,
+      "loss": 0.5094,
+      "step": 17435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.26666012929561e-05,
+      "loss": 0.2703,
+      "step": 17436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.265116933817428e-05,
+      "loss": 0.1561,
+      "step": 17437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.2635738782153e-05,
+      "loss": 0.5158,
+      "step": 17438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.26203096251392e-05,
+      "loss": 0.0017,
+      "step": 17439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.260488186738e-05,
+      "loss": 0.3409,
+      "step": 17440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.258945550912247e-05,
+      "loss": 0.3024,
+      "step": 17441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.25740305506136e-05,
+      "loss": 0.2166,
+      "step": 17442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.255860699210042e-05,
+      "loss": 0.2845,
+      "step": 17443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.254318483382992e-05,
+      "loss": 0.2623,
+      "step": 17444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.252776407604907e-05,
+      "loss": 0.4013,
+      "step": 17445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.25123447190048e-05,
+      "loss": 0.2649,
+      "step": 17446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.249692676294404e-05,
+      "loss": 0.0011,
+      "step": 17447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.248151020811366e-05,
+      "loss": 0.0027,
+      "step": 17448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.246609505476054e-05,
+      "loss": 0.6156,
+      "step": 17449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.245068130313156e-05,
+      "loss": 0.004,
+      "step": 17450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.243526895347355e-05,
+      "loss": 0.3454,
+      "step": 17451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.241985800603322e-05,
+      "loss": 0.0054,
+      "step": 17452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.240444846105745e-05,
+      "loss": 1.0314,
+      "step": 17453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.238904031879305e-05,
+      "loss": 0.1932,
+      "step": 17454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.237363357948661e-05,
+      "loss": 0.0009,
+      "step": 17455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.235822824338486e-05,
+      "loss": 0.1774,
+      "step": 17456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.23428243107347e-05,
+      "loss": 0.5607,
+      "step": 17457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.232742178178257e-05,
+      "loss": 0.3344,
+      "step": 17458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.23120206567752e-05,
+      "loss": 0.4007,
+      "step": 17459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.229662093595921e-05,
+      "loss": 0.3081,
+      "step": 17460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.228122261958124e-05,
+      "loss": 0.9027,
+      "step": 17461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.226582570788783e-05,
+      "loss": 1.0274,
+      "step": 17462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.225043020112555e-05,
+      "loss": 0.0014,
+      "step": 17463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.223503609954092e-05,
+      "loss": 0.2374,
+      "step": 17464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.221964340338051e-05,
+      "loss": 0.3937,
+      "step": 17465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.220425211289073e-05,
+      "loss": 0.7885,
+      "step": 17466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.21888622283181e-05,
+      "loss": 0.5975,
+      "step": 17467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.217347374990907e-05,
+      "loss": 0.4433,
+      "step": 17468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.215808667791004e-05,
+      "loss": 1.238,
+      "step": 17469
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.214270101256747e-05,
+      "loss": 0.3119,
+      "step": 17470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.212731675412759e-05,
+      "loss": 0.4614,
+      "step": 17471
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.21119339028369e-05,
+      "loss": 0.0104,
+      "step": 17472
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.209655245894175e-05,
+      "loss": 0.0071,
+      "step": 17473
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.208117242268827e-05,
+      "loss": 0.6548,
+      "step": 17474
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.206579379432291e-05,
+      "loss": 0.3954,
+      "step": 17475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.205041657409196e-05,
+      "loss": 0.4205,
+      "step": 17476
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.20350407622415e-05,
+      "loss": 0.0072,
+      "step": 17477
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.201966635901794e-05,
+      "loss": 0.6727,
+      "step": 17478
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.200429336466731e-05,
+      "loss": 0.6973,
+      "step": 17479
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.198892177943582e-05,
+      "loss": 1.1786,
+      "step": 17480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.197355160356977e-05,
+      "loss": 0.2421,
+      "step": 17481
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.195818283731514e-05,
+      "loss": 0.3348,
+      "step": 17482
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.1942815480918e-05,
+      "loss": 0.4889,
+      "step": 17483
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.192744953462466e-05,
+      "loss": 0.186,
+      "step": 17484
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.191208499868094e-05,
+      "loss": 0.3195,
+      "step": 17485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.189672187333295e-05,
+      "loss": 0.4865,
+      "step": 17486
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.188136015882685e-05,
+      "loss": 0.2933,
+      "step": 17487
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.186599985540845e-05,
+      "loss": 0.0028,
+      "step": 17488
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.18506409633238e-05,
+      "loss": 0.2195,
+      "step": 17489
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.183528348281884e-05,
+      "loss": 0.0021,
+      "step": 17490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.181992741413947e-05,
+      "loss": 0.8025,
+      "step": 17491
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.180457275753166e-05,
+      "loss": 2.6817,
+      "step": 17492
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.178921951324123e-05,
+      "loss": 0.1294,
+      "step": 17493
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.177386768151407e-05,
+      "loss": 0.6102,
+      "step": 17494
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.175851726259602e-05,
+      "loss": 0.0205,
+      "step": 17495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.17431682567329e-05,
+      "loss": 0.8546,
+      "step": 17496
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.172782066417044e-05,
+      "loss": 0.7488,
+      "step": 17497
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.171247448515449e-05,
+      "loss": 0.3094,
+      "step": 17498
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.169712971993074e-05,
+      "loss": 0.7536,
+      "step": 17499
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.1681786368745e-05,
+      "loss": 0.6183,
+      "step": 17500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.16664444318428e-05,
+      "loss": 0.959,
+      "step": 17501
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.165110390946997e-05,
+      "loss": 0.31,
+      "step": 17502
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.163576480187217e-05,
+      "loss": 1.0736,
+      "step": 17503
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.162042710929492e-05,
+      "loss": 0.0041,
+      "step": 17504
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.160509083198384e-05,
+      "loss": 0.3863,
+      "step": 17505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.158975597018468e-05,
+      "loss": 0.253,
+      "step": 17506
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.157442252414285e-05,
+      "loss": 0.0171,
+      "step": 17507
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.155909049410391e-05,
+      "loss": 0.6759,
+      "step": 17508
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.154375988031342e-05,
+      "loss": 2.541,
+      "step": 17509
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.152843068301686e-05,
+      "loss": 1.2059,
+      "step": 17510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.15131029024597e-05,
+      "loss": 0.4541,
+      "step": 17511
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.14977765388874e-05,
+      "loss": 0.1667,
+      "step": 17512
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.148245159254539e-05,
+      "loss": 0.6157,
+      "step": 17513
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.146712806367907e-05,
+      "loss": 0.5698,
+      "step": 17514
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.145180595253383e-05,
+      "loss": 0.2949,
+      "step": 17515
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.143648525935501e-05,
+      "loss": 0.0064,
+      "step": 17516
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.142116598438798e-05,
+      "loss": 0.0066,
+      "step": 17517
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.140584812787803e-05,
+      "loss": 0.3192,
+      "step": 17518
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.139053169007046e-05,
+      "loss": 0.5914,
+      "step": 17519
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.137521667121054e-05,
+      "loss": 0.4869,
+      "step": 17520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.135990307154352e-05,
+      "loss": 0.6741,
+      "step": 17521
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.134459089131467e-05,
+      "loss": 0.5862,
+      "step": 17522
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.132928013076904e-05,
+      "loss": 0.0009,
+      "step": 17523
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.131397079015198e-05,
+      "loss": 0.7969,
+      "step": 17524
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.129866286970861e-05,
+      "loss": 0.53,
+      "step": 17525
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.128335636968394e-05,
+      "loss": 0.4464,
+      "step": 17526
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.126805129032329e-05,
+      "loss": 0.3646,
+      "step": 17527
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.125274763187155e-05,
+      "loss": 0.006,
+      "step": 17528
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.123744539457383e-05,
+      "loss": 0.4497,
+      "step": 17529
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.12221445786753e-05,
+      "loss": 0.2349,
+      "step": 17530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.120684518442086e-05,
+      "loss": 0.2654,
+      "step": 17531
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.119154721205546e-05,
+      "loss": 1.4735,
+      "step": 17532
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.117625066182425e-05,
+      "loss": 0.7852,
+      "step": 17533
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.116095553397205e-05,
+      "loss": 0.0029,
+      "step": 17534
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.114566182874374e-05,
+      "loss": 0.1197,
+      "step": 17535
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.113036954638444e-05,
+      "loss": 0.7746,
+      "step": 17536
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.111507868713881e-05,
+      "loss": 1.4356,
+      "step": 17537
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.109978925125183e-05,
+      "loss": 0.0147,
+      "step": 17538
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.108450123896829e-05,
+      "loss": 0.7033,
+      "step": 17539
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.1069214650533e-05,
+      "loss": 0.3358,
+      "step": 17540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.10539294861908e-05,
+      "loss": 2.1811,
+      "step": 17541
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.103864574618643e-05,
+      "loss": 0.854,
+      "step": 17542
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.1023363430764624e-05,
+      "loss": 1.4101,
+      "step": 17543
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.100808254017011e-05,
+      "loss": 0.2407,
+      "step": 17544
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.09928030746476e-05,
+      "loss": 0.9741,
+      "step": 17545
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0977525034441774e-05,
+      "loss": 5.0759,
+      "step": 17546
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.096224841979727e-05,
+      "loss": 2.2088,
+      "step": 17547
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.094697323095874e-05,
+      "loss": 0.5507,
+      "step": 17548
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.093169946817085e-05,
+      "loss": 0.3799,
+      "step": 17549
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0916427131678e-05,
+      "loss": 0.4497,
+      "step": 17550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0901156221724926e-05,
+      "loss": 0.1653,
+      "step": 17551
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.088588673855617e-05,
+      "loss": 0.8552,
+      "step": 17552
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.087061868241608e-05,
+      "loss": 0.398,
+      "step": 17553
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.085535205354933e-05,
+      "loss": 0.5874,
+      "step": 17554
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0840086852200386e-05,
+      "loss": 0.3924,
+      "step": 17555
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0824823078613586e-05,
+      "loss": 0.8986,
+      "step": 17556
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.080956073303341e-05,
+      "loss": 0.2971,
+      "step": 17557
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.079429981570427e-05,
+      "loss": 0.7005,
+      "step": 17558
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0779040326870534e-05,
+      "loss": 0.3936,
+      "step": 17559
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.076378226677656e-05,
+      "loss": 0.121,
+      "step": 17560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0748525635666695e-05,
+      "loss": 0.0138,
+      "step": 17561
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.073327043378524e-05,
+      "loss": 0.278,
+      "step": 17562
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0718016661376486e-05,
+      "loss": 0.044,
+      "step": 17563
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.070276431868471e-05,
+      "loss": 0.3067,
+      "step": 17564
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.068751340595413e-05,
+      "loss": 0.8847,
+      "step": 17565
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0672263923429e-05,
+      "loss": 0.3828,
+      "step": 17566
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0657015871353506e-05,
+      "loss": 0.5657,
+      "step": 17567
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0641769249971804e-05,
+      "loss": 0.5186,
+      "step": 17568
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.062652405952806e-05,
+      "loss": 0.5616,
+      "step": 17569
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0611280300266416e-05,
+      "loss": 0.018,
+      "step": 17570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0596037972431007e-05,
+      "loss": 0.7129,
+      "step": 17571
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.0580797076265775e-05,
+      "loss": 0.4,
+      "step": 17572
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.056555761201491e-05,
+      "loss": 0.0172,
+      "step": 17573
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.05503195799225e-05,
+      "loss": 0.4095,
+      "step": 17574
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.053508298023234e-05,
+      "loss": 0.429,
+      "step": 17575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.051984781318869e-05,
+      "loss": 1.5494,
+      "step": 17576
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.0504614079035325e-05,
+      "loss": 0.4393,
+      "step": 17577
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.0489381778016186e-05,
+      "loss": 0.1782,
+      "step": 17578
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.047415091037539e-05,
+      "loss": 0.7084,
+      "step": 17579
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.0458921476356634e-05,
+      "loss": 0.2734,
+      "step": 17580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.044369347620383e-05,
+      "loss": 0.5732,
+      "step": 17581
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.042846691016097e-05,
+      "loss": 0.3086,
+      "step": 17582
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.0413241778471734e-05,
+      "loss": 0.4126,
+      "step": 17583
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.039801808137992e-05,
+      "loss": 0.0051,
+      "step": 17584
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.038279581912948e-05,
+      "loss": 1.1312,
+      "step": 17585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.036757499196403e-05,
+      "loss": 0.5945,
+      "step": 17586
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.0352355600127334e-05,
+      "loss": 0.7097,
+      "step": 17587
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.033713764386311e-05,
+      "loss": 0.0056,
+      "step": 17588
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.032192112341508e-05,
+      "loss": 0.7528,
+      "step": 17589
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.03067060390269e-05,
+      "loss": 0.3146,
+      "step": 17590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.029149239094219e-05,
+      "loss": 0.8579,
+      "step": 17591
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.02762801794046e-05,
+      "loss": 0.1783,
+      "step": 17592
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.0261069404657746e-05,
+      "loss": 0.2237,
+      "step": 17593
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.024586006694517e-05,
+      "loss": 0.1375,
+      "step": 17594
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.023065216651045e-05,
+      "loss": 0.3223,
+      "step": 17595
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.02154457035971e-05,
+      "loss": 0.003,
+      "step": 17596
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.020024067844865e-05,
+      "loss": 0.4125,
+      "step": 17597
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.018503709130862e-05,
+      "loss": 0.3593,
+      "step": 17598
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.016983494242032e-05,
+      "loss": 0.3029,
+      "step": 17599
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.015463423202737e-05,
+      "loss": 0.2645,
+      "step": 17600
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.013943496037315e-05,
+      "loss": 0.5841,
+      "step": 17601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.012423712770092e-05,
+      "loss": 0.4953,
+      "step": 17602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.010904073425418e-05,
+      "loss": 1.6507,
+      "step": 17603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.009384578027632e-05,
+      "loss": 0.3082,
+      "step": 17604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.007865226601048e-05,
+      "loss": 0.2623,
+      "step": 17605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.0063460191700174e-05,
+      "loss": 1.5442,
+      "step": 17606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.004826955758853e-05,
+      "loss": 0.7525,
+      "step": 17607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.0033080363918834e-05,
+      "loss": 2.6002,
+      "step": 17608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.001789261093434e-05,
+      "loss": 0.3877,
+      "step": 17609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.000270629887826e-05,
+      "loss": 0.3709,
+      "step": 17610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.998752142799378e-05,
+      "loss": 0.766,
+      "step": 17611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9972337998524034e-05,
+      "loss": 2.5313,
+      "step": 17612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.99571560107122e-05,
+      "loss": 0.0035,
+      "step": 17613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.994197546480136e-05,
+      "loss": 0.7148,
+      "step": 17614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.992679636103463e-05,
+      "loss": 0.4008,
+      "step": 17615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.991161869965509e-05,
+      "loss": 0.041,
+      "step": 17616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.989644248090575e-05,
+      "loss": 0.6163,
+      "step": 17617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.988126770502968e-05,
+      "loss": 0.0091,
+      "step": 17618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.986609437226984e-05,
+      "loss": 1.6699,
+      "step": 17619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9850922482869224e-05,
+      "loss": 1.2163,
+      "step": 17620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.98357520370708e-05,
+      "loss": 0.0053,
+      "step": 17621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9820583035117486e-05,
+      "loss": 1.6555,
+      "step": 17622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.980541547725225e-05,
+      "loss": 1.3986,
+      "step": 17623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.979024936371783e-05,
+      "loss": 1.9121,
+      "step": 17624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9775084694757266e-05,
+      "loss": 0.5536,
+      "step": 17625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.975992147061326e-05,
+      "loss": 0.2967,
+      "step": 17626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.974475969152864e-05,
+      "loss": 0.4081,
+      "step": 17627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.972959935774634e-05,
+      "loss": 0.7647,
+      "step": 17628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9714440469508975e-05,
+      "loss": 0.1707,
+      "step": 17629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.969928302705928e-05,
+      "loss": 0.0066,
+      "step": 17630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9684127030640164e-05,
+      "loss": 0.8672,
+      "step": 17631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9668972480494155e-05,
+      "loss": 0.3099,
+      "step": 17632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.965381937686393e-05,
+      "loss": 0.4571,
+      "step": 17633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.963866771999231e-05,
+      "loss": 0.2363,
+      "step": 17634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9623517510121746e-05,
+      "loss": 0.4307,
+      "step": 17635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.960836874749493e-05,
+      "loss": 0.344,
+      "step": 17636
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.959322143235442e-05,
+      "loss": 0.0185,
+      "step": 17637
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.957807556494278e-05,
+      "loss": 0.6808,
+      "step": 17638
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9562931145502564e-05,
+      "loss": 0.2978,
+      "step": 17639
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.954778817427629e-05,
+      "loss": 0.0192,
+      "step": 17640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.953264665150642e-05,
+      "loss": 1.2015,
+      "step": 17641
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.951750657743545e-05,
+      "loss": 0.3178,
+      "step": 17642
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9502367952305805e-05,
+      "loss": 0.2331,
+      "step": 17643
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.948723077635994e-05,
+      "loss": 0.3681,
+      "step": 17644
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.947209504984022e-05,
+      "loss": 0.0057,
+      "step": 17645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9456960772989034e-05,
+      "loss": 0.1885,
+      "step": 17646
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9441827946048786e-05,
+      "loss": 0.3944,
+      "step": 17647
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.942669656926166e-05,
+      "loss": 0.985,
+      "step": 17648
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9411566642870115e-05,
+      "loss": 0.0355,
+      "step": 17649
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.939643816711641e-05,
+      "loss": 0.4256,
+      "step": 17650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.938131114224268e-05,
+      "loss": 0.8239,
+      "step": 17651
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.936618556849128e-05,
+      "loss": 0.016,
+      "step": 17652
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.935106144610448e-05,
+      "loss": 0.4532,
+      "step": 17653
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.933593877532426e-05,
+      "loss": 0.0847,
+      "step": 17654
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.932081755639302e-05,
+      "loss": 1.7246,
+      "step": 17655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9305697789552765e-05,
+      "loss": 0.1072,
+      "step": 17656
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.929057947504562e-05,
+      "loss": 0.0051,
+      "step": 17657
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.927546261311373e-05,
+      "loss": 0.1999,
+      "step": 17658
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.926034720399913e-05,
+      "loss": 0.6117,
+      "step": 17659
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.924523324794389e-05,
+      "loss": 0.1803,
+      "step": 17660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9230120745190045e-05,
+      "loss": 0.5169,
+      "step": 17661
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.921500969597959e-05,
+      "loss": 0.3364,
+      "step": 17662
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.91999001005545e-05,
+      "loss": 0.7309,
+      "step": 17663
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9184791959156736e-05,
+      "loss": 0.3407,
+      "step": 17664
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.916968527202824e-05,
+      "loss": 0.0027,
+      "step": 17665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.915458003941092e-05,
+      "loss": 0.6484,
+      "step": 17666
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.913947626154665e-05,
+      "loss": 0.3374,
+      "step": 17667
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.912437393867731e-05,
+      "loss": 0.4077,
+      "step": 17668
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.910927307104475e-05,
+      "loss": 0.3574,
+      "step": 17669
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.909417365889076e-05,
+      "loss": 0.4195,
+      "step": 17670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.907907570245715e-05,
+      "loss": 0.3436,
+      "step": 17671
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9063979201985684e-05,
+      "loss": 0.6928,
+      "step": 17672
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.904888415771811e-05,
+      "loss": 0.0437,
+      "step": 17673
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.903379056989624e-05,
+      "loss": 0.3396,
+      "step": 17674
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.901869843876162e-05,
+      "loss": 0.057,
+      "step": 17675
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.9003607764555944e-05,
+      "loss": 0.9487,
+      "step": 17676
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8988518547521035e-05,
+      "loss": 0.0089,
+      "step": 17677
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8973430787898354e-05,
+      "loss": 0.7276,
+      "step": 17678
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8958344485929515e-05,
+      "loss": 0.0012,
+      "step": 17679
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.894325964185627e-05,
+      "loss": 0.5933,
+      "step": 17680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.892817625592e-05,
+      "loss": 0.1281,
+      "step": 17681
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8913094328362256e-05,
+      "loss": 0.3241,
+      "step": 17682
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8898013859424714e-05,
+      "loss": 0.3139,
+      "step": 17683
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.888293484934872e-05,
+      "loss": 0.6339,
+      "step": 17684
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.886785729837574e-05,
+      "loss": 0.2855,
+      "step": 17685
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.885278120674728e-05,
+      "loss": 0.2888,
+      "step": 17686
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.883770657470474e-05,
+      "loss": 1.6442,
+      "step": 17687
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.88226334024895e-05,
+      "loss": 0.6034,
+      "step": 17688
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.880756169034295e-05,
+      "loss": 0.271,
+      "step": 17689
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8792491438506435e-05,
+      "loss": 1.1382,
+      "step": 17690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.87774226472213e-05,
+      "loss": 0.0057,
+      "step": 17691
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.876235531672882e-05,
+      "loss": 0.6429,
+      "step": 17692
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.87472894472703e-05,
+      "loss": 0.0971,
+      "step": 17693
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8732225039087e-05,
+      "loss": 0.2182,
+      "step": 17694
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.871716209242012e-05,
+      "loss": 0.4127,
+      "step": 17695
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.870210060751096e-05,
+      "loss": 0.0067,
+      "step": 17696
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.868704058460052e-05,
+      "loss": 0.3965,
+      "step": 17697
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.867198202393016e-05,
+      "loss": 0.2426,
+      "step": 17698
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8656924925740966e-05,
+      "loss": 0.151,
+      "step": 17699
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.864186929027395e-05,
+      "loss": 0.2461,
+      "step": 17700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8626815117770336e-05,
+      "loss": 0.2318,
+      "step": 17701
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.86117624084712e-05,
+      "loss": 0.29,
+      "step": 17702
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.859671116261742e-05,
+      "loss": 0.3898,
+      "step": 17703
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.858166138045025e-05,
+      "loss": 0.8024,
+      "step": 17704
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.856661306221051e-05,
+      "loss": 0.3351,
+      "step": 17705
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.855156620813919e-05,
+      "loss": 2.1445,
+      "step": 17706
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.853652081847739e-05,
+      "loss": 0.3344,
+      "step": 17707
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.852147689346592e-05,
+      "loss": 0.7558,
+      "step": 17708
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.850643443334568e-05,
+      "loss": 0.5163,
+      "step": 17709
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.849139343835758e-05,
+      "loss": 0.325,
+      "step": 17710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.847635390874248e-05,
+      "loss": 0.0934,
+      "step": 17711
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.846131584474124e-05,
+      "loss": 0.4211,
+      "step": 17712
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.844627924659461e-05,
+      "loss": 0.3392,
+      "step": 17713
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.843124411454344e-05,
+      "loss": 1.0668,
+      "step": 17714
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.841621044882846e-05,
+      "loss": 0.4059,
+      "step": 17715
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8401178249690424e-05,
+      "loss": 0.3472,
+      "step": 17716
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.838614751737004e-05,
+      "loss": 0.3708,
+      "step": 17717
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8371118252108024e-05,
+      "loss": 0.5618,
+      "step": 17718
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.835609045414501e-05,
+      "loss": 0.6885,
+      "step": 17719
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.834106412372169e-05,
+      "loss": 0.3669,
+      "step": 17720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.832603926107866e-05,
+      "loss": 2.6367,
+      "step": 17721
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8311015866456537e-05,
+      "loss": 0.5663,
+      "step": 17722
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.829599394009593e-05,
+      "loss": 0.5447,
+      "step": 17723
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.828097348223731e-05,
+      "loss": 0.5739,
+      "step": 17724
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8265954493121184e-05,
+      "loss": 0.7651,
+      "step": 17725
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.825093697298824e-05,
+      "loss": 0.0091,
+      "step": 17726
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.823592092207879e-05,
+      "loss": 0.6622,
+      "step": 17727
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.822090634063329e-05,
+      "loss": 0.317,
+      "step": 17728
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.820589322889233e-05,
+      "loss": 0.193,
+      "step": 17729
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.819088158709619e-05,
+      "loss": 0.4075,
+      "step": 17730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.817587141548524e-05,
+      "loss": 0.4317,
+      "step": 17731
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.81608627143e-05,
+      "loss": 0.8012,
+      "step": 17732
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.814585548378067e-05,
+      "loss": 0.5318,
+      "step": 17733
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.81308497241676e-05,
+      "loss": 0.3997,
+      "step": 17734
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.811584543570108e-05,
+      "loss": 0.214,
+      "step": 17735
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.810084261862143e-05,
+      "loss": 0.0108,
+      "step": 17736
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8085841273168855e-05,
+      "loss": 0.7432,
+      "step": 17737
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.807084139958357e-05,
+      "loss": 0.6166,
+      "step": 17738
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8055842998105816e-05,
+      "loss": 0.9458,
+      "step": 17739
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.804084606897573e-05,
+      "loss": 0.0053,
+      "step": 17740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.802585061243349e-05,
+      "loss": 0.0056,
+      "step": 17741
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.8010856628719275e-05,
+      "loss": 0.2818,
+      "step": 17742
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.799586411807302e-05,
+      "loss": 0.2406,
+      "step": 17743
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.798087308073498e-05,
+      "loss": 0.2943,
+      "step": 17744
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.796588351694521e-05,
+      "loss": 0.7106,
+      "step": 17745
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.795089542694358e-05,
+      "loss": 0.0022,
+      "step": 17746
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7935908810970286e-05,
+      "loss": 0.6555,
+      "step": 17747
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.792092366926528e-05,
+      "loss": 0.0039,
+      "step": 17748
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.790594000206838e-05,
+      "loss": 0.3337,
+      "step": 17749
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.78909578096197e-05,
+      "loss": 0.3666,
+      "step": 17750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7875977092159146e-05,
+      "loss": 1.2578,
+      "step": 17751
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.786099784992646e-05,
+      "loss": 0.0112,
+      "step": 17752
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.784602008316169e-05,
+      "loss": 0.0048,
+      "step": 17753
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.783104379210459e-05,
+      "loss": 0.0043,
+      "step": 17754
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.781606897699491e-05,
+      "loss": 0.003,
+      "step": 17755
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.780109563807266e-05,
+      "loss": 1.9818,
+      "step": 17756
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.778612377557745e-05,
+      "loss": 0.1132,
+      "step": 17757
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7771153389749e-05,
+      "loss": 2.9589,
+      "step": 17758
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7756184480827235e-05,
+      "loss": 0.3649,
+      "step": 17759
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7741217049051666e-05,
+      "loss": 0.7422,
+      "step": 17760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.772625109466207e-05,
+      "loss": 0.8422,
+      "step": 17761
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.771128661789806e-05,
+      "loss": 0.3135,
+      "step": 17762
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.769632361899929e-05,
+      "loss": 0.3822,
+      "step": 17763
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.768136209820538e-05,
+      "loss": 0.4422,
+      "step": 17764
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.766640205575588e-05,
+      "loss": 1.199,
+      "step": 17765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.765144349189041e-05,
+      "loss": 0.2294,
+      "step": 17766
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.763648640684846e-05,
+      "loss": 0.3826,
+      "step": 17767
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7621530800869556e-05,
+      "loss": 0.3162,
+      "step": 17768
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.76065766741932e-05,
+      "loss": 0.1712,
+      "step": 17769
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7591624027058866e-05,
+      "loss": 0.0089,
+      "step": 17770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7576672859705974e-05,
+      "loss": 0.448,
+      "step": 17771
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.756172317237402e-05,
+      "loss": 0.198,
+      "step": 17772
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.754677496530224e-05,
+      "loss": 0.1279,
+      "step": 17773
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7531828238730145e-05,
+      "loss": 0.3333,
+      "step": 17774
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.75168829928971e-05,
+      "loss": 0.7122,
+      "step": 17775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.750193922804232e-05,
+      "loss": 0.3799,
+      "step": 17776
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.748699694440513e-05,
+      "loss": 0.7126,
+      "step": 17777
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7472056142224924e-05,
+      "loss": 1.0536,
+      "step": 17778
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.745711682174084e-05,
+      "loss": 0.002,
+      "step": 17779
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.744217898319208e-05,
+      "loss": 0.7661,
+      "step": 17780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7427242626818014e-05,
+      "loss": 0.3378,
+      "step": 17781
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.741230775285768e-05,
+      "loss": 2.3242,
+      "step": 17782
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.739737436155028e-05,
+      "loss": 0.3024,
+      "step": 17783
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.738244245313495e-05,
+      "loss": 1.3394,
+      "step": 17784
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7367512027850805e-05,
+      "loss": 0.3918,
+      "step": 17785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.735258308593694e-05,
+      "loss": 0.0062,
+      "step": 17786
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.733765562763241e-05,
+      "loss": 0.2634,
+      "step": 17787
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7322729653176256e-05,
+      "loss": 0.3059,
+      "step": 17788
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.73078051628075e-05,
+      "loss": 0.0119,
+      "step": 17789
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.729288215676513e-05,
+      "loss": 1.2038,
+      "step": 17790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7277960635288175e-05,
+      "loss": 0.0768,
+      "step": 17791
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.726304059861541e-05,
+      "loss": 0.2801,
+      "step": 17792
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.724812204698594e-05,
+      "loss": 0.1662,
+      "step": 17793
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7233204980638636e-05,
+      "loss": 0.0541,
+      "step": 17794
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.721828939981222e-05,
+      "loss": 0.325,
+      "step": 17795
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.720337530474571e-05,
+      "loss": 0.6164,
+      "step": 17796
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.718846269567794e-05,
+      "loss": 0.442,
+      "step": 17797
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7173551572847525e-05,
+      "loss": 0.5033,
+      "step": 17798
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.715864193649344e-05,
+      "loss": 0.6247,
+      "step": 17799
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.714373378685441e-05,
+      "loss": 0.9342,
+      "step": 17800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.712882712416902e-05,
+      "loss": 0.0042,
+      "step": 17801
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.711392194867619e-05,
+      "loss": 3.0194,
+      "step": 17802
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7099018260614446e-05,
+      "loss": 0.3309,
+      "step": 17803
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.7084116060222465e-05,
+      "loss": 0.0496,
+      "step": 17804
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.706921534773899e-05,
+      "loss": 0.3452,
+      "step": 17805
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.705431612340254e-05,
+      "loss": 0.4793,
+      "step": 17806
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.703941838745165e-05,
+      "loss": 0.5298,
+      "step": 17807
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.702452214012507e-05,
+      "loss": 0.5364,
+      "step": 17808
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.700962738166119e-05,
+      "loss": 0.0223,
+      "step": 17809
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6994734112298555e-05,
+      "loss": 0.5194,
+      "step": 17810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6979842332275685e-05,
+      "loss": 0.2255,
+      "step": 17811
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.696495204183103e-05,
+      "loss": 0.6915,
+      "step": 17812
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.695006324120304e-05,
+      "loss": 0.1728,
+      "step": 17813
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6935175930630134e-05,
+      "loss": 0.5226,
+      "step": 17814
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.692029011035071e-05,
+      "loss": 0.3965,
+      "step": 17815
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6905405780603175e-05,
+      "loss": 0.897,
+      "step": 17816
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6890522941625826e-05,
+      "loss": 0.7799,
+      "step": 17817
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.687564159365703e-05,
+      "loss": 0.1913,
+      "step": 17818
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.686076173693506e-05,
+      "loss": 0.0117,
+      "step": 17819
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.684588337169822e-05,
+      "loss": 0.4999,
+      "step": 17820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.6831006498184794e-05,
+      "loss": 0.2707,
+      "step": 17821
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.681613111663287e-05,
+      "loss": 0.5641,
+      "step": 17822
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.680125722728082e-05,
+      "loss": 0.4647,
+      "step": 17823
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.67863848303668e-05,
+      "loss": 0.405,
+      "step": 17824
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.677151392612883e-05,
+      "loss": 0.6403,
+      "step": 17825
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6756644514805204e-05,
+      "loss": 2.0056,
+      "step": 17826
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.674177659663403e-05,
+      "loss": 0.4139,
+      "step": 17827
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.67269101718533e-05,
+      "loss": 0.312,
+      "step": 17828
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.671204524070106e-05,
+      "loss": 0.0031,
+      "step": 17829
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.669718180341551e-05,
+      "loss": 0.0029,
+      "step": 17830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.668231986023453e-05,
+      "loss": 0.5652,
+      "step": 17831
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.666745941139614e-05,
+      "loss": 0.7613,
+      "step": 17832
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.665260045713831e-05,
+      "loss": 0.6275,
+      "step": 17833
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.663774299769898e-05,
+      "loss": 0.6045,
+      "step": 17834
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.662288703331608e-05,
+      "loss": 0.0518,
+      "step": 17835
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.660803256422752e-05,
+      "loss": 0.0035,
+      "step": 17836
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.659317959067114e-05,
+      "loss": 0.7442,
+      "step": 17837
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.657832811288482e-05,
+      "loss": 0.4186,
+      "step": 17838
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.656347813110637e-05,
+      "loss": 0.3061,
+      "step": 17839
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6548629645573576e-05,
+      "loss": 0.6333,
+      "step": 17840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.653378265652424e-05,
+      "loss": 0.7215,
+      "step": 17841
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6518937164196097e-05,
+      "loss": 0.2592,
+      "step": 17842
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.650409316882693e-05,
+      "loss": 0.1578,
+      "step": 17843
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.648925067065428e-05,
+      "loss": 0.966,
+      "step": 17844
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.647440966991601e-05,
+      "loss": 0.2521,
+      "step": 17845
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.645957016684975e-05,
+      "loss": 0.0022,
+      "step": 17846
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.644473216169299e-05,
+      "loss": 1.2018,
+      "step": 17847
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.642989565468349e-05,
+      "loss": 0.3222,
+      "step": 17848
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.641506064605882e-05,
+      "loss": 0.6112,
+      "step": 17849
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.640022713605641e-05,
+      "loss": 0.5535,
+      "step": 17850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6385395124913985e-05,
+      "loss": 0.4201,
+      "step": 17851
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.637056461286892e-05,
+      "loss": 2.2832,
+      "step": 17852
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6355735600158685e-05,
+      "loss": 0.4577,
+      "step": 17853
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6340908087020916e-05,
+      "loss": 0.2942,
+      "step": 17854
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.632608207369288e-05,
+      "loss": 0.1565,
+      "step": 17855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.631125756041201e-05,
+      "loss": 0.247,
+      "step": 17856
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.629643454741585e-05,
+      "loss": 0.4581,
+      "step": 17857
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.628161303494159e-05,
+      "loss": 0.3751,
+      "step": 17858
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.626679302322662e-05,
+      "loss": 1.0537,
+      "step": 17859
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6251974512508376e-05,
+      "loss": 1.1865,
+      "step": 17860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6237157503024014e-05,
+      "loss": 0.1838,
+      "step": 17861
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6222341995010854e-05,
+      "loss": 0.0132,
+      "step": 17862
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6207527988706154e-05,
+      "loss": 0.2929,
+      "step": 17863
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6192715484347116e-05,
+      "loss": 0.3293,
+      "step": 17864
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.617790448217096e-05,
+      "loss": 0.0043,
+      "step": 17865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6163094982414855e-05,
+      "loss": 0.7139,
+      "step": 17866
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.614828698531597e-05,
+      "loss": 0.1305,
+      "step": 17867
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6133480491111396e-05,
+      "loss": 0.347,
+      "step": 17868
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.611867550003827e-05,
+      "loss": 0.4027,
+      "step": 17869
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.610387201233371e-05,
+      "loss": 0.3414,
+      "step": 17870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.608907002823462e-05,
+      "loss": 0.6548,
+      "step": 17871
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.607426954797818e-05,
+      "loss": 0.0017,
+      "step": 17872
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.60594705718014e-05,
+      "loss": 0.3508,
+      "step": 17873
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.6044673099941114e-05,
+      "loss": 0.0787,
+      "step": 17874
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.602987713263444e-05,
+      "loss": 4.8011,
+      "step": 17875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.601508267011831e-05,
+      "loss": 0.001,
+      "step": 17876
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.600028971262951e-05,
+      "loss": 0.7587,
+      "step": 17877
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.598549826040494e-05,
+      "loss": 0.2911,
+      "step": 17878
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5970708313681635e-05,
+      "loss": 0.0328,
+      "step": 17879
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.595591987269627e-05,
+      "loss": 0.9138,
+      "step": 17880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.594113293768571e-05,
+      "loss": 0.5212,
+      "step": 17881
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.592634750888674e-05,
+      "loss": 0.3234,
+      "step": 17882
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.591156358653611e-05,
+      "loss": 0.4696,
+      "step": 17883
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.589678117087061e-05,
+      "loss": 0.0997,
+      "step": 17884
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.588200026212692e-05,
+      "loss": 0.8341,
+      "step": 17885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.586722086054174e-05,
+      "loss": 1.0208,
+      "step": 17886
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.585244296635175e-05,
+      "loss": 0.7395,
+      "step": 17887
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.583766657979358e-05,
+      "loss": 0.4421,
+      "step": 17888
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.582289170110388e-05,
+      "loss": 0.2838,
+      "step": 17889
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5808118330519216e-05,
+      "loss": 1.0761,
+      "step": 17890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.579334646827619e-05,
+      "loss": 0.7126,
+      "step": 17891
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.577857611461132e-05,
+      "loss": 0.1763,
+      "step": 17892
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.576380726976114e-05,
+      "loss": 0.0055,
+      "step": 17893
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.574903993396218e-05,
+      "loss": 0.234,
+      "step": 17894
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5734274107450935e-05,
+      "loss": 0.5395,
+      "step": 17895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.571950979046371e-05,
+      "loss": 0.8091,
+      "step": 17896
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.57047469832371e-05,
+      "loss": 0.2416,
+      "step": 17897
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.56899856860075e-05,
+      "loss": 1.876,
+      "step": 17898
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5675225899011124e-05,
+      "loss": 1.5065,
+      "step": 17899
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5660467622484564e-05,
+      "loss": 0.5203,
+      "step": 17900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.564571085666398e-05,
+      "loss": 1.4779,
+      "step": 17901
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5630955601785664e-05,
+      "loss": 1.9395,
+      "step": 17902
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.561620185808607e-05,
+      "loss": 0.4293,
+      "step": 17903
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.560144962580131e-05,
+      "loss": 0.4531,
+      "step": 17904
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.558669890516761e-05,
+      "loss": 0.2984,
+      "step": 17905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.557194969642133e-05,
+      "loss": 0.6739,
+      "step": 17906
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5557201999798524e-05,
+      "loss": 0.8922,
+      "step": 17907
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.554245581553532e-05,
+      "loss": 0.0599,
+      "step": 17908
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.552771114386803e-05,
+      "loss": 0.5709,
+      "step": 17909
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.551296798503262e-05,
+      "loss": 0.0028,
+      "step": 17910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.549822633926521e-05,
+      "loss": 0.5254,
+      "step": 17911
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.548348620680189e-05,
+      "loss": 1.8665,
+      "step": 17912
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5468747587878685e-05,
+      "loss": 0.0035,
+      "step": 17913
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.545401048273161e-05,
+      "loss": 0.8578,
+      "step": 17914
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.543927489159668e-05,
+      "loss": 0.504,
+      "step": 17915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5424540814709825e-05,
+      "loss": 0.0025,
+      "step": 17916
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.540980825230703e-05,
+      "loss": 0.7692,
+      "step": 17917
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5395077204624184e-05,
+      "loss": 0.0791,
+      "step": 17918
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.538034767189725e-05,
+      "loss": 0.7409,
+      "step": 17919
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.536561965436193e-05,
+      "loss": 0.002,
+      "step": 17920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5350893152254247e-05,
+      "loss": 0.722,
+      "step": 17921
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.533616816581001e-05,
+      "loss": 0.8929,
+      "step": 17922
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.532144469526488e-05,
+      "loss": 0.2462,
+      "step": 17923
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5306722740854755e-05,
+      "loss": 0.2707,
+      "step": 17924
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.529200230281542e-05,
+      "loss": 0.2612,
+      "step": 17925
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5277283381382406e-05,
+      "loss": 0.8575,
+      "step": 17926
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.52625659767916e-05,
+      "loss": 0.2189,
+      "step": 17927
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5247850089278675e-05,
+      "loss": 1.0175,
+      "step": 17928
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.523313571907917e-05,
+      "loss": 0.7572,
+      "step": 17929
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.521842286642879e-05,
+      "loss": 0.2257,
+      "step": 17930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.52037115315631e-05,
+      "loss": 0.5109,
+      "step": 17931
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5189001714717706e-05,
+      "loss": 0.5639,
+      "step": 17932
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.517429341612815e-05,
+      "loss": 0.0024,
+      "step": 17933
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.515958663602998e-05,
+      "loss": 0.175,
+      "step": 17934
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.51448813746587e-05,
+      "loss": 0.1365,
+      "step": 17935
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.513017763224978e-05,
+      "loss": 0.8273,
+      "step": 17936
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.511547540903869e-05,
+      "loss": 0.2507,
+      "step": 17937
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.510077470526085e-05,
+      "loss": 0.4407,
+      "step": 17938
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.50860755211517e-05,
+      "loss": 3.3526,
+      "step": 17939
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.507137785694658e-05,
+      "loss": 0.4149,
+      "step": 17940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.505668171288088e-05,
+      "loss": 0.3682,
+      "step": 17941
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.504198708918994e-05,
+      "loss": 0.0098,
+      "step": 17942
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.5027293986109064e-05,
+      "loss": 0.534,
+      "step": 17943
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.50126024038736e-05,
+      "loss": 0.0251,
+      "step": 17944
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4997912342718635e-05,
+      "loss": 0.4053,
+      "step": 17945
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.498322380287958e-05,
+      "loss": 2.4089,
+      "step": 17946
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.496853678459165e-05,
+      "loss": 0.1893,
+      "step": 17947
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.495385128808988e-05,
+      "loss": 0.1994,
+      "step": 17948
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.493916731360963e-05,
+      "loss": 0.5969,
+      "step": 17949
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.492448486138591e-05,
+      "loss": 0.2945,
+      "step": 17950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.490980393165383e-05,
+      "loss": 0.8635,
+      "step": 17951
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.489512452464862e-05,
+      "loss": 0.0011,
+      "step": 17952
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4880446640605214e-05,
+      "loss": 0.4138,
+      "step": 17953
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.486577027975864e-05,
+      "loss": 1.3125,
+      "step": 17954
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4851095442344085e-05,
+      "loss": 0.2001,
+      "step": 17955
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.483642212859639e-05,
+      "loss": 0.6059,
+      "step": 17956
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4821750338750526e-05,
+      "loss": 0.3743,
+      "step": 17957
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.480708007304158e-05,
+      "loss": 0.0016,
+      "step": 17958
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4792411331704334e-05,
+      "loss": 0.4413,
+      "step": 17959
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.477774411497374e-05,
+      "loss": 0.0032,
+      "step": 17960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4763078423084674e-05,
+      "loss": 0.6144,
+      "step": 17961
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.474841425627195e-05,
+      "loss": 0.0018,
+      "step": 17962
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4733751614770435e-05,
+      "loss": 0.0632,
+      "step": 17963
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.471909049881491e-05,
+      "loss": 0.6654,
+      "step": 17964
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.470443090864016e-05,
+      "loss": 0.7482,
+      "step": 17965
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.468977284448092e-05,
+      "loss": 0.9278,
+      "step": 17966
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4675116306571945e-05,
+      "loss": 2.4352,
+      "step": 17967
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.466046129514795e-05,
+      "loss": 1.0347,
+      "step": 17968
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.464580781044349e-05,
+      "loss": 0.7671,
+      "step": 17969
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.463115585269337e-05,
+      "loss": 0.2475,
+      "step": 17970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.461650542213219e-05,
+      "loss": 0.9931,
+      "step": 17971
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4601856518994444e-05,
+      "loss": 0.0035,
+      "step": 17972
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.458720914351485e-05,
+      "loss": 0.02,
+      "step": 17973
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4572563295927927e-05,
+      "loss": 1.1711,
+      "step": 17974
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.455791897646812e-05,
+      "loss": 0.5018,
+      "step": 17975
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.454327618537003e-05,
+      "loss": 0.1925,
+      "step": 17976
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.452863492286817e-05,
+      "loss": 1.4785,
+      "step": 17977
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.451399518919682e-05,
+      "loss": 0.8622,
+      "step": 17978
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.449935698459065e-05,
+      "loss": 1.0353,
+      "step": 17979
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.44847203092839e-05,
+      "loss": 0.7196,
+      "step": 17980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.447008516351102e-05,
+      "loss": 0.8528,
+      "step": 17981
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.445545154750635e-05,
+      "loss": 0.8809,
+      "step": 17982
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4440819461504215e-05,
+      "loss": 0.3633,
+      "step": 17983
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4426188905738935e-05,
+      "loss": 1.1364,
+      "step": 17984
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4411559880444826e-05,
+      "loss": 0.0801,
+      "step": 17985
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.439693238585613e-05,
+      "loss": 0.6973,
+      "step": 17986
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4382306422207075e-05,
+      "loss": 0.375,
+      "step": 17987
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.436768198973187e-05,
+      "loss": 2.1732,
+      "step": 17988
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.435305908866472e-05,
+      "loss": 0.5349,
+      "step": 17989
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.433843771923978e-05,
+      "loss": 0.0031,
+      "step": 17990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.43238178816912e-05,
+      "loss": 0.007,
+      "step": 17991
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.430919957625308e-05,
+      "loss": 0.0072,
+      "step": 17992
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.429458280315952e-05,
+      "loss": 0.7354,
+      "step": 17993
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4279967562644575e-05,
+      "loss": 0.3492,
+      "step": 17994
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.426535385494231e-05,
+      "loss": 0.3815,
+      "step": 17995
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.425074168028677e-05,
+      "loss": 0.2268,
+      "step": 17996
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.42361310389118e-05,
+      "loss": 0.5913,
+      "step": 17997
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.422152193105157e-05,
+      "loss": 0.349,
+      "step": 17998
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4206914356939876e-05,
+      "loss": 0.0092,
+      "step": 17999
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.419230831681062e-05,
+      "loss": 0.3763,
+      "step": 18000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.417770381089787e-05,
+      "loss": 1.1215,
+      "step": 18001
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.416310083943533e-05,
+      "loss": 0.3279,
+      "step": 18002
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.414849940265686e-05,
+      "loss": 0.603,
+      "step": 18003
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.413389950079641e-05,
+      "loss": 0.4161,
+      "step": 18004
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4119301134087624e-05,
+      "loss": 0.3817,
+      "step": 18005
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.41047043027643e-05,
+      "loss": 0.2441,
+      "step": 18006
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4090109007060315e-05,
+      "loss": 0.3576,
+      "step": 18007
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.407551524720923e-05,
+      "loss": 0.0345,
+      "step": 18008
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.406092302344483e-05,
+      "loss": 0.605,
+      "step": 18009
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.404633233600075e-05,
+      "loss": 0.3992,
+      "step": 18010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4031743185110645e-05,
+      "loss": 0.6947,
+      "step": 18011
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.4017155571008165e-05,
+      "loss": 0.2565,
+      "step": 18012
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.400256949392687e-05,
+      "loss": 0.001,
+      "step": 18013
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.398798495410035e-05,
+      "loss": 0.0077,
+      "step": 18014
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.397340195176217e-05,
+      "loss": 0.2683,
+      "step": 18015
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.395882048714584e-05,
+      "loss": 0.2815,
+      "step": 18016
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.394424056048491e-05,
+      "loss": 0.254,
+      "step": 18017
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.39296621720127e-05,
+      "loss": 0.3815,
+      "step": 18018
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.391508532196283e-05,
+      "loss": 0.2833,
+      "step": 18019
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.390051001056871e-05,
+      "loss": 0.8068,
+      "step": 18020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.38859362380636e-05,
+      "loss": 0.4977,
+      "step": 18021
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3871364004681025e-05,
+      "loss": 0.1207,
+      "step": 18022
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.385679331065433e-05,
+      "loss": 0.3401,
+      "step": 18023
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.384222415621671e-05,
+      "loss": 0.0516,
+      "step": 18024
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3827656541601594e-05,
+      "loss": 0.5879,
+      "step": 18025
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.381309046704229e-05,
+      "loss": 0.3058,
+      "step": 18026
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.379852593277187e-05,
+      "loss": 0.1724,
+      "step": 18027
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.378396293902379e-05,
+      "loss": 0.2248,
+      "step": 18028
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.376940148603108e-05,
+      "loss": 0.0027,
+      "step": 18029
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3754841574027e-05,
+      "loss": 0.4556,
+      "step": 18030
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.374028320324468e-05,
+      "loss": 0.4349,
+      "step": 18031
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.372572637391726e-05,
+      "loss": 0.5402,
+      "step": 18032
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.371117108627784e-05,
+      "loss": 1.0581,
+      "step": 18033
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.369661734055952e-05,
+      "loss": 0.0315,
+      "step": 18034
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3682065136995324e-05,
+      "loss": 0.3854,
+      "step": 18035
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3667514475818304e-05,
+      "loss": 0.1152,
+      "step": 18036
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.365296535726149e-05,
+      "loss": 0.1628,
+      "step": 18037
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.363841778155782e-05,
+      "loss": 0.4254,
+      "step": 18038
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.362387174894027e-05,
+      "loss": 0.3088,
+      "step": 18039
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3609327259641764e-05,
+      "loss": 0.3275,
+      "step": 18040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.359478431389523e-05,
+      "loss": 0.2961,
+      "step": 18041
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.358024291193354e-05,
+      "loss": 0.0044,
+      "step": 18042
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3565703053989564e-05,
+      "loss": 0.3337,
+      "step": 18043
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.355116474029612e-05,
+      "loss": 0.005,
+      "step": 18044
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.353662797108602e-05,
+      "loss": 0.1939,
+      "step": 18045
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.352209274659205e-05,
+      "loss": 0.7048,
+      "step": 18046
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.350755906704703e-05,
+      "loss": 0.4374,
+      "step": 18047
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3493026932683586e-05,
+      "loss": 0.8379,
+      "step": 18048
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3478496343734415e-05,
+      "loss": 0.0084,
+      "step": 18049
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.346396730043237e-05,
+      "loss": 0.2501,
+      "step": 18050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.344943980300997e-05,
+      "loss": 0.044,
+      "step": 18051
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.343491385169983e-05,
+      "loss": 0.205,
+      "step": 18052
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3420389446734735e-05,
+      "loss": 0.2037,
+      "step": 18053
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.340586658834709e-05,
+      "loss": 0.5049,
+      "step": 18054
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3391345276769474e-05,
+      "loss": 0.4812,
+      "step": 18055
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.337682551223459e-05,
+      "loss": 0.0037,
+      "step": 18056
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.336230729497477e-05,
+      "loss": 0.3382,
+      "step": 18057
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.334779062522257e-05,
+      "loss": 0.3953,
+      "step": 18058
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.333327550321045e-05,
+      "loss": 0.1265,
+      "step": 18059
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3318761929170845e-05,
+      "loss": 0.292,
+      "step": 18060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3304249903336174e-05,
+      "loss": 0.226,
+      "step": 18061
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.328973942593881e-05,
+      "loss": 0.001,
+      "step": 18062
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.327523049721114e-05,
+      "loss": 1.3076,
+      "step": 18063
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.326072311738548e-05,
+      "loss": 0.5141,
+      "step": 18064
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.324621728669417e-05,
+      "loss": 0.2308,
+      "step": 18065
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3231713005369514e-05,
+      "loss": 0.3487,
+      "step": 18066
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.321721027364366e-05,
+      "loss": 0.7942,
+      "step": 18067
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.3202709091748965e-05,
+      "loss": 0.6562,
+      "step": 18068
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.318820945991768e-05,
+      "loss": 0.1252,
+      "step": 18069
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.317371137838182e-05,
+      "loss": 0.0026,
+      "step": 18070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 5.315921484737371e-05,
+      "loss": 0.2222,
+      "step": 18071
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.3144719867125475e-05,
+      "loss": 1.4594,
+      "step": 18072
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.3130226437869094e-05,
+      "loss": 0.4251,
+      "step": 18073
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.3115734559836814e-05,
+      "loss": 0.628,
+      "step": 18074
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.310124423326068e-05,
+      "loss": 0.308,
+      "step": 18075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.308675545837261e-05,
+      "loss": 0.177,
+      "step": 18076
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.30722682354048e-05,
+      "loss": 0.6531,
+      "step": 18077
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.305778256458909e-05,
+      "loss": 0.6865,
+      "step": 18078
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.3043298446157456e-05,
+      "loss": 0.0279,
+      "step": 18079
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.302881588034199e-05,
+      "loss": 0.0068,
+      "step": 18080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.301433486737444e-05,
+      "loss": 0.1336,
+      "step": 18081
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.299985540748677e-05,
+      "loss": 0.2011,
+      "step": 18082
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.298537750091085e-05,
+      "loss": 0.0025,
+      "step": 18083
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.297090114787851e-05,
+      "loss": 0.0785,
+      "step": 18084
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2956426348621554e-05,
+      "loss": 0.2632,
+      "step": 18085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.294195310337179e-05,
+      "loss": 0.7895,
+      "step": 18086
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.292748141236098e-05,
+      "loss": 0.0045,
+      "step": 18087
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.291301127582088e-05,
+      "loss": 0.4971,
+      "step": 18088
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2898542693983204e-05,
+      "loss": 0.2537,
+      "step": 18089
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.288407566707963e-05,
+      "loss": 0.3547,
+      "step": 18090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.286961019534182e-05,
+      "loss": 0.3798,
+      "step": 18091
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2855146279001446e-05,
+      "loss": 0.4457,
+      "step": 18092
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2840683918290106e-05,
+      "loss": 0.2237,
+      "step": 18093
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2826223113439405e-05,
+      "loss": 0.2362,
+      "step": 18094
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.281176386468089e-05,
+      "loss": 0.4887,
+      "step": 18095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2797306172246176e-05,
+      "loss": 0.687,
+      "step": 18096
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.278285003636667e-05,
+      "loss": 0.1191,
+      "step": 18097
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.276839545727386e-05,
+      "loss": 2.8312,
+      "step": 18098
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.275394243519938e-05,
+      "loss": 0.4286,
+      "step": 18099
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2739490970374515e-05,
+      "loss": 0.6729,
+      "step": 18100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2725041063030696e-05,
+      "loss": 0.2589,
+      "step": 18101
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.271059271339945e-05,
+      "loss": 1.9911,
+      "step": 18102
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.269614592171199e-05,
+      "loss": 0.0038,
+      "step": 18103
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2681700688199674e-05,
+      "loss": 3.7884,
+      "step": 18104
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2667257013093976e-05,
+      "loss": 0.1384,
+      "step": 18105
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.265281489662602e-05,
+      "loss": 0.0059,
+      "step": 18106
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.263837433902714e-05,
+      "loss": 0.3748,
+      "step": 18107
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2623935340528575e-05,
+      "loss": 0.7393,
+      "step": 18108
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.260949790136155e-05,
+      "loss": 0.7082,
+      "step": 18109
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2595062021757244e-05,
+      "loss": 0.4655,
+      "step": 18110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.258062770194684e-05,
+      "loss": 0.3628,
+      "step": 18111
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2566194942161485e-05,
+      "loss": 0.0009,
+      "step": 18112
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.255176374263228e-05,
+      "loss": 0.2715,
+      "step": 18113
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.253733410359033e-05,
+      "loss": 1.2217,
+      "step": 18114
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2522906025266764e-05,
+      "loss": 0.5668,
+      "step": 18115
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.250847950789245e-05,
+      "loss": 0.5024,
+      "step": 18116
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2494054551698576e-05,
+      "loss": 0.2535,
+      "step": 18117
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2479631156916136e-05,
+      "loss": 0.0142,
+      "step": 18118
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.246520932377596e-05,
+      "loss": 0.1739,
+      "step": 18119
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.245078905250911e-05,
+      "loss": 2.4606,
+      "step": 18120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.243637034334653e-05,
+      "loss": 0.0238,
+      "step": 18121
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.242195319651894e-05,
+      "loss": 0.2765,
+      "step": 18122
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2407537612257394e-05,
+      "loss": 0.2624,
+      "step": 18123
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.239312359079271e-05,
+      "loss": 0.2775,
+      "step": 18124
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.237871113235556e-05,
+      "loss": 0.7664,
+      "step": 18125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.236430023717695e-05,
+      "loss": 0.3903,
+      "step": 18126
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.234989090548749e-05,
+      "loss": 0.8785,
+      "step": 18127
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.233548313751792e-05,
+      "loss": 0.1234,
+      "step": 18128
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2321076933499125e-05,
+      "loss": 0.0013,
+      "step": 18129
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.230667229366163e-05,
+      "loss": 0.4493,
+      "step": 18130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.22922692182361e-05,
+      "loss": 0.4423,
+      "step": 18131
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2277867707453355e-05,
+      "loss": 0.2612,
+      "step": 18132
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.226346776154384e-05,
+      "loss": 0.5299,
+      "step": 18133
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2249069380738216e-05,
+      "loss": 0.9578,
+      "step": 18134
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2234672565267024e-05,
+      "loss": 0.3586,
+      "step": 18135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2220277315360845e-05,
+      "loss": 0.2422,
+      "step": 18136
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.220588363125018e-05,
+      "loss": 0.0048,
+      "step": 18137
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2191491513165504e-05,
+      "loss": 0.2447,
+      "step": 18138
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.21771009613373e-05,
+      "loss": 0.674,
+      "step": 18139
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.216271197599602e-05,
+      "loss": 0.5432,
+      "step": 18140
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2148324557372076e-05,
+      "loss": 1.3333,
+      "step": 18141
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.213393870569586e-05,
+      "loss": 0.7471,
+      "step": 18142
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.211955442119775e-05,
+      "loss": 1.5059,
+      "step": 18143
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.210517170410807e-05,
+      "loss": 0.4637,
+      "step": 18144
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2090790554657176e-05,
+      "loss": 0.0008,
+      "step": 18145
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.207641097307525e-05,
+      "loss": 0.3795,
+      "step": 18146
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2062032959592694e-05,
+      "loss": 0.2584,
+      "step": 18147
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.2047656514439734e-05,
+      "loss": 0.5847,
+      "step": 18148
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.203328163784649e-05,
+      "loss": 0.6787,
+      "step": 18149
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.201890833004318e-05,
+      "loss": 0.206,
+      "step": 18150
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.20045365912601e-05,
+      "loss": 0.4124,
+      "step": 18151
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1990166421727245e-05,
+      "loss": 0.7113,
+      "step": 18152
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.197579782167474e-05,
+      "loss": 0.9995,
+      "step": 18153
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1961430791332815e-05,
+      "loss": 0.0274,
+      "step": 18154
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.194706533093138e-05,
+      "loss": 1.0663,
+      "step": 18155
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1932701440700544e-05,
+      "loss": 0.2647,
+      "step": 18156
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1918339120870315e-05,
+      "loss": 0.5183,
+      "step": 18157
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.190397837167069e-05,
+      "loss": 0.0086,
+      "step": 18158
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.188961919333163e-05,
+      "loss": 0.7821,
+      "step": 18159
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.187526158608307e-05,
+      "loss": 0.002,
+      "step": 18160
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.186090555015492e-05,
+      "loss": 0.0029,
+      "step": 18161
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.18465510857771e-05,
+      "loss": 0.2784,
+      "step": 18162
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1832198193179466e-05,
+      "loss": 0.5366,
+      "step": 18163
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1817846872591865e-05,
+      "loss": 0.5801,
+      "step": 18164
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1803497124244024e-05,
+      "loss": 0.7989,
+      "step": 18165
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1789148948365836e-05,
+      "loss": 0.3598,
+      "step": 18166
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1774802345187105e-05,
+      "loss": 0.4857,
+      "step": 18167
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1760457314937376e-05,
+      "loss": 2.0957,
+      "step": 18168
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.174611385784655e-05,
+      "loss": 0.0018,
+      "step": 18169
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.173177197414431e-05,
+      "loss": 0.5464,
+      "step": 18170
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.171743166406014e-05,
+      "loss": 0.3724,
+      "step": 18171
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.170309292782387e-05,
+      "loss": 0.003,
+      "step": 18172
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.168875576566509e-05,
+      "loss": 0.4371,
+      "step": 18173
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.167442017781322e-05,
+      "loss": 0.302,
+      "step": 18174
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.166008616449807e-05,
+      "loss": 0.7815,
+      "step": 18175
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.164575372594899e-05,
+      "loss": 0.3156,
+      "step": 18176
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.163142286239551e-05,
+      "loss": 1.1816,
+      "step": 18177
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.161709357406726e-05,
+      "loss": 0.5926,
+      "step": 18178
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1602765861193554e-05,
+      "loss": 0.414,
+      "step": 18179
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.158843972400384e-05,
+      "loss": 0.2669,
+      "step": 18180
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.157411516272766e-05,
+      "loss": 0.1647,
+      "step": 18181
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.155979217759426e-05,
+      "loss": 0.309,
+      "step": 18182
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1545470768833044e-05,
+      "loss": 0.6735,
+      "step": 18183
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.153115093667337e-05,
+      "loss": 0.1314,
+      "step": 18184
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1516832681344516e-05,
+      "loss": 0.4878,
+      "step": 18185
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.150251600307578e-05,
+      "loss": 1.7793,
+      "step": 18186
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.148820090209643e-05,
+      "loss": 1.9467,
+      "step": 18187
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1473887378635714e-05,
+      "loss": 0.0009,
+      "step": 18188
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.145957543292281e-05,
+      "loss": 0.2025,
+      "step": 18189
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.144526506518692e-05,
+      "loss": 0.388,
+      "step": 18190
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1430956275657205e-05,
+      "loss": 0.9865,
+      "step": 18191
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.141664906456278e-05,
+      "loss": 0.5123,
+      "step": 18192
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.140234343213279e-05,
+      "loss": 0.2909,
+      "step": 18193
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1388039378596334e-05,
+      "loss": 0.247,
+      "step": 18194
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1373736904182326e-05,
+      "loss": 0.7614,
+      "step": 18195
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1359436009119956e-05,
+      "loss": 0.5846,
+      "step": 18196
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.134513669363823e-05,
+      "loss": 0.2873,
+      "step": 18197
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.133083895796598e-05,
+      "loss": 0.5416,
+      "step": 18198
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.131654280233231e-05,
+      "loss": 0.3214,
+      "step": 18199
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.130224822696617e-05,
+      "loss": 0.6775,
+      "step": 18200
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.128795523209633e-05,
+      "loss": 0.392,
+      "step": 18201
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1273663817951706e-05,
+      "loss": 0.1748,
+      "step": 18202
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.125937398476128e-05,
+      "loss": 0.5724,
+      "step": 18203
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.124508573275374e-05,
+      "loss": 0.4053,
+      "step": 18204
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.123079906215794e-05,
+      "loss": 0.3938,
+      "step": 18205
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1216513973202674e-05,
+      "loss": 2.1797,
+      "step": 18206
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.120223046611667e-05,
+      "loss": 0.0808,
+      "step": 18207
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1187948541128675e-05,
+      "loss": 0.002,
+      "step": 18208
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.117366819846738e-05,
+      "loss": 0.3329,
+      "step": 18209
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.115938943836148e-05,
+      "loss": 0.7319,
+      "step": 18210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.114511226103962e-05,
+      "loss": 0.5111,
+      "step": 18211
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1130836666730424e-05,
+      "loss": 1.8683,
+      "step": 18212
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1116562655662496e-05,
+      "loss": 0.0357,
+      "step": 18213
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.110229022806443e-05,
+      "loss": 0.2991,
+      "step": 18214
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.108801938416474e-05,
+      "loss": 0.4696,
+      "step": 18215
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1073750124192034e-05,
+      "loss": 0.3845,
+      "step": 18216
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1059482448374646e-05,
+      "loss": 0.7736,
+      "step": 18217
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1045216356941196e-05,
+      "loss": 0.6582,
+      "step": 18218
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.103095185012016e-05,
+      "loss": 0.594,
+      "step": 18219
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.10166889281398e-05,
+      "loss": 0.3716,
+      "step": 18220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.1002427591228664e-05,
+      "loss": 0.2398,
+      "step": 18221
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.098816783961512e-05,
+      "loss": 0.2511,
+      "step": 18222
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0973909673527384e-05,
+      "loss": 0.6529,
+      "step": 18223
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0959653093193964e-05,
+      "loss": 0.0469,
+      "step": 18224
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0945398098843e-05,
+      "loss": 0.2684,
+      "step": 18225
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0931144690702776e-05,
+      "loss": 0.8428,
+      "step": 18226
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0916892869001694e-05,
+      "loss": 0.4839,
+      "step": 18227
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0902642633967824e-05,
+      "loss": 0.5008,
+      "step": 18228
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.088839398582935e-05,
+      "loss": 0.2861,
+      "step": 18229
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0874146924814594e-05,
+      "loss": 0.4963,
+      "step": 18230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0859901451151574e-05,
+      "loss": 0.0029,
+      "step": 18231
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0845657565068435e-05,
+      "loss": 0.8153,
+      "step": 18232
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0831415266793285e-05,
+      "loss": 1.3633,
+      "step": 18233
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0817174556554186e-05,
+      "loss": 0.3575,
+      "step": 18234
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.080293543457919e-05,
+      "loss": 0.3074,
+      "step": 18235
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.07886979010963e-05,
+      "loss": 0.682,
+      "step": 18236
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.077446195633353e-05,
+      "loss": 0.6314,
+      "step": 18237
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.076022760051882e-05,
+      "loss": 0.2643,
+      "step": 18238
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.074599483388013e-05,
+      "loss": 0.4422,
+      "step": 18239
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.073176365664538e-05,
+      "loss": 0.0124,
+      "step": 18240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0717534069042445e-05,
+      "loss": 0.0013,
+      "step": 18241
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.07033060712992e-05,
+      "loss": 0.5252,
+      "step": 18242
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.068907966364355e-05,
+      "loss": 2.0373,
+      "step": 18243
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.067485484630313e-05,
+      "loss": 0.4123,
+      "step": 18244
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.066063161950589e-05,
+      "loss": 0.0204,
+      "step": 18245
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0646409983479616e-05,
+      "loss": 0.3454,
+      "step": 18246
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.063218993845186e-05,
+      "loss": 0.3765,
+      "step": 18247
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0617971484650504e-05,
+      "loss": 0.3205,
+      "step": 18248
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.060375462230325e-05,
+      "loss": 1.8416,
+      "step": 18249
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.058953935163763e-05,
+      "loss": 0.4532,
+      "step": 18250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0575325672881284e-05,
+      "loss": 0.2675,
+      "step": 18251
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.056111358626199e-05,
+      "loss": 0.1964,
+      "step": 18252
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0546903092007175e-05,
+      "loss": 0.1062,
+      "step": 18253
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0532694190344446e-05,
+      "loss": 0.7357,
+      "step": 18254
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.051848688150135e-05,
+      "loss": 2.0957,
+      "step": 18255
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.050428116570537e-05,
+      "loss": 0.0028,
+      "step": 18256
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.049007704318403e-05,
+      "loss": 0.2257,
+      "step": 18257
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.047587451416475e-05,
+      "loss": 2.688,
+      "step": 18258
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0461673578874965e-05,
+      "loss": 0.007,
+      "step": 18259
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.04474742375421e-05,
+      "loss": 0.1861,
+      "step": 18260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.043327649039352e-05,
+      "loss": 0.8941,
+      "step": 18261
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.041908033765659e-05,
+      "loss": 1.6035,
+      "step": 18262
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.040488577955865e-05,
+      "loss": 2.5958,
+      "step": 18263
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.039069281632698e-05,
+      "loss": 0.2668,
+      "step": 18264
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.037650144818894e-05,
+      "loss": 1.1376,
+      "step": 18265
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0362311675371604e-05,
+      "loss": 0.9116,
+      "step": 18266
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.034812349810236e-05,
+      "loss": 0.1334,
+      "step": 18267
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.03339369166084e-05,
+      "loss": 0.4204,
+      "step": 18268
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.031975193111678e-05,
+      "loss": 0.2782,
+      "step": 18269
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0305568541854777e-05,
+      "loss": 0.397,
+      "step": 18270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.029138674904953e-05,
+      "loss": 0.5541,
+      "step": 18271
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.027720655292798e-05,
+      "loss": 0.4138,
+      "step": 18272
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.02630279537174e-05,
+      "loss": 0.2137,
+      "step": 18273
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.024885095164469e-05,
+      "loss": 0.0022,
+      "step": 18274
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.023467554693688e-05,
+      "loss": 0.005,
+      "step": 18275
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0220501739821115e-05,
+      "loss": 0.4694,
+      "step": 18276
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.020632953052422e-05,
+      "loss": 0.4156,
+      "step": 18277
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.019215891927314e-05,
+      "loss": 0.6982,
+      "step": 18278
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.017798990629493e-05,
+      "loss": 1.5457,
+      "step": 18279
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0163822491816355e-05,
+      "loss": 0.2817,
+      "step": 18280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.014965667606434e-05,
+      "loss": 0.9593,
+      "step": 18281
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.01354924592657e-05,
+      "loss": 0.4924,
+      "step": 18282
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.012132984164729e-05,
+      "loss": 0.2114,
+      "step": 18283
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.010716882343588e-05,
+      "loss": 0.8955,
+      "step": 18284
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0093009404858256e-05,
+      "loss": 0.2847,
+      "step": 18285
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.007885158614113e-05,
+      "loss": 0.4042,
+      "step": 18286
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.006469536751124e-05,
+      "loss": 0.0572,
+      "step": 18287
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.005054074919527e-05,
+      "loss": 0.6895,
+      "step": 18288
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0036387731419894e-05,
+      "loss": 0.1865,
+      "step": 18289
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.0022236314411734e-05,
+      "loss": 0.4099,
+      "step": 18290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.000808649839741e-05,
+      "loss": 0.3215,
+      "step": 18291
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9993938283603585e-05,
+      "loss": 0.2032,
+      "step": 18292
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.997979167025663e-05,
+      "loss": 0.3508,
+      "step": 18293
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.996564665858327e-05,
+      "loss": 0.3276,
+      "step": 18294
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9951503248809985e-05,
+      "loss": 0.0181,
+      "step": 18295
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.993736144116312e-05,
+      "loss": 0.3898,
+      "step": 18296
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.99232212358693e-05,
+      "loss": 0.0038,
+      "step": 18297
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.990908263315494e-05,
+      "loss": 0.6786,
+      "step": 18298
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9894945633246316e-05,
+      "loss": 0.2211,
+      "step": 18299
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9880810236369944e-05,
+      "loss": 0.4008,
+      "step": 18300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.986667644275219e-05,
+      "loss": 1.0807,
+      "step": 18301
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.985254425261927e-05,
+      "loss": 0.5588,
+      "step": 18302
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.983841366619757e-05,
+      "loss": 0.3241,
+      "step": 18303
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.982428468371335e-05,
+      "loss": 0.628,
+      "step": 18304
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9810157305392866e-05,
+      "loss": 0.64,
+      "step": 18305
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.979603153146235e-05,
+      "loss": 0.2826,
+      "step": 18306
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.978190736214801e-05,
+      "loss": 0.0018,
+      "step": 18307
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9767784797676e-05,
+      "loss": 0.3151,
+      "step": 18308
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9753663838272516e-05,
+      "loss": 0.4908,
+      "step": 18309
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.973954448416363e-05,
+      "loss": 1.1524,
+      "step": 18310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.972542673557549e-05,
+      "loss": 0.4426,
+      "step": 18311
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.971131059273414e-05,
+      "loss": 0.9512,
+      "step": 18312
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.969719605586564e-05,
+      "loss": 0.0028,
+      "step": 18313
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9683083125196e-05,
+      "loss": 0.5102,
+      "step": 18314
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.9668971800951254e-05,
+      "loss": 0.3376,
+      "step": 18315
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.965486208335733e-05,
+      "loss": 0.0132,
+      "step": 18316
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.964075397264025e-05,
+      "loss": 0.5554,
+      "step": 18317
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.962664746902578e-05,
+      "loss": 0.0044,
+      "step": 18318
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.961254257273994e-05,
+      "loss": 0.6646,
+      "step": 18319
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.959843928400863e-05,
+      "loss": 0.0018,
+      "step": 18320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.958433760305755e-05,
+      "loss": 0.0054,
+      "step": 18321
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.957023753011269e-05,
+      "loss": 0.2461,
+      "step": 18322
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.95561390653997e-05,
+      "loss": 0.0134,
+      "step": 18323
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9542042209144364e-05,
+      "loss": 0.5092,
+      "step": 18324
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.952794696157253e-05,
+      "loss": 0.2506,
+      "step": 18325
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.951385332290981e-05,
+      "loss": 0.0019,
+      "step": 18326
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.949976129338185e-05,
+      "loss": 0.9354,
+      "step": 18327
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.948567087321449e-05,
+      "loss": 0.0059,
+      "step": 18328
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.947158206263321e-05,
+      "loss": 0.0072,
+      "step": 18329
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9457494861863677e-05,
+      "loss": 0.6315,
+      "step": 18330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9443409271131465e-05,
+      "loss": 0.2097,
+      "step": 18331
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.942932529066215e-05,
+      "loss": 0.0017,
+      "step": 18332
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.941524292068125e-05,
+      "loss": 0.0074,
+      "step": 18333
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.940116216141427e-05,
+      "loss": 0.242,
+      "step": 18334
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.938708301308671e-05,
+      "loss": 0.0028,
+      "step": 18335
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9373005475924005e-05,
+      "loss": 0.2842,
+      "step": 18336
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.935892955015161e-05,
+      "loss": 0.4816,
+      "step": 18337
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9344855235994905e-05,
+      "loss": 0.2418,
+      "step": 18338
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.93307825336793e-05,
+      "loss": 0.0186,
+      "step": 18339
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9316711443430115e-05,
+      "loss": 0.231,
+      "step": 18340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.930264196547274e-05,
+      "loss": 0.0008,
+      "step": 18341
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.928857410003233e-05,
+      "loss": 0.341,
+      "step": 18342
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.927450784733431e-05,
+      "loss": 0.5976,
+      "step": 18343
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.926044320760393e-05,
+      "loss": 0.5546,
+      "step": 18344
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.924638018106627e-05,
+      "loss": 0.6864,
+      "step": 18345
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9232318767946655e-05,
+      "loss": 0.2051,
+      "step": 18346
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9218258968470293e-05,
+      "loss": 0.0055,
+      "step": 18347
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.920420078286215e-05,
+      "loss": 0.2336,
+      "step": 18348
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.91901442113475e-05,
+      "loss": 1.2343,
+      "step": 18349
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.917608925415147e-05,
+      "loss": 0.0446,
+      "step": 18350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9162035911498935e-05,
+      "loss": 0.8852,
+      "step": 18351
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.914798418361516e-05,
+      "loss": 0.2302,
+      "step": 18352
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9133934070725015e-05,
+      "loss": 0.888,
+      "step": 18353
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.911988557305355e-05,
+      "loss": 0.7131,
+      "step": 18354
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.91058386908257e-05,
+      "loss": 0.0147,
+      "step": 18355
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.909179342426643e-05,
+      "loss": 0.4989,
+      "step": 18356
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.907774977360064e-05,
+      "loss": 0.3096,
+      "step": 18357
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9063707739053236e-05,
+      "loss": 0.0046,
+      "step": 18358
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.904966732084906e-05,
+      "loss": 0.4384,
+      "step": 18359
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.903562851921295e-05,
+      "loss": 0.3697,
+      "step": 18360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.902159133436974e-05,
+      "loss": 0.3977,
+      "step": 18361
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.9007555766544175e-05,
+      "loss": 0.0023,
+      "step": 18362
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8993521815961057e-05,
+      "loss": 0.0035,
+      "step": 18363
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.897948948284508e-05,
+      "loss": 0.6562,
+      "step": 18364
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.896545876742099e-05,
+      "loss": 0.152,
+      "step": 18365
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8951429669913424e-05,
+      "loss": 0.8938,
+      "step": 18366
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8937402190547064e-05,
+      "loss": 0.0478,
+      "step": 18367
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.892337632954655e-05,
+      "loss": 0.0034,
+      "step": 18368
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8909352087136514e-05,
+      "loss": 0.8183,
+      "step": 18369
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8895329463541374e-05,
+      "loss": 0.9476,
+      "step": 18370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.88813084589859e-05,
+      "loss": 0.6978,
+      "step": 18371
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.886728907369447e-05,
+      "loss": 0.9099,
+      "step": 18372
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8853271307891587e-05,
+      "loss": 0.3142,
+      "step": 18373
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.883925516180184e-05,
+      "loss": 0.4284,
+      "step": 18374
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.882524063564957e-05,
+      "loss": 0.3133,
+      "step": 18375
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.881122772965917e-05,
+      "loss": 0.7095,
+      "step": 18376
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.879721644405519e-05,
+      "loss": 0.6478,
+      "step": 18377
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.878320677906186e-05,
+      "loss": 0.5919,
+      "step": 18378
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.876919873490356e-05,
+      "loss": 0.3884,
+      "step": 18379
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8755192311804634e-05,
+      "loss": 0.0027,
+      "step": 18380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.874118750998935e-05,
+      "loss": 0.2258,
+      "step": 18381
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8727184329681973e-05,
+      "loss": 0.6366,
+      "step": 18382
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.871318277110676e-05,
+      "loss": 0.0612,
+      "step": 18383
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.86991828344879e-05,
+      "loss": 0.5482,
+      "step": 18384
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.86851845200496e-05,
+      "loss": 0.7161,
+      "step": 18385
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.867118782801603e-05,
+      "loss": 0.416,
+      "step": 18386
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8657192758611304e-05,
+      "loss": 0.5718,
+      "step": 18387
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.864319931205954e-05,
+      "loss": 0.3681,
+      "step": 18388
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8629207488584824e-05,
+      "loss": 1.0744,
+      "step": 18389
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.861521728841127e-05,
+      "loss": 0.3578,
+      "step": 18390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8601228711762756e-05,
+      "loss": 0.0096,
+      "step": 18391
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.858724175886343e-05,
+      "loss": 0.0908,
+      "step": 18392
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.857325642993728e-05,
+      "loss": 0.3455,
+      "step": 18393
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8559272725208106e-05,
+      "loss": 0.3434,
+      "step": 18394
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.854529064489999e-05,
+      "loss": 0.3252,
+      "step": 18395
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.853131018923682e-05,
+      "loss": 0.1442,
+      "step": 18396
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.851733135844235e-05,
+      "loss": 0.1737,
+      "step": 18397
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8503354152740553e-05,
+      "loss": 0.1824,
+      "step": 18398
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.848937857235528e-05,
+      "loss": 0.3414,
+      "step": 18399
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.847540461751015e-05,
+      "loss": 0.3162,
+      "step": 18400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.846143228842915e-05,
+      "loss": 0.0018,
+      "step": 18401
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.844746158533587e-05,
+      "loss": 0.9025,
+      "step": 18402
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.84334925084541e-05,
+      "loss": 0.5098,
+      "step": 18403
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8419525058007494e-05,
+      "loss": 0.7799,
+      "step": 18404
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.840555923421976e-05,
+      "loss": 0.537,
+      "step": 18405
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8391595037314524e-05,
+      "loss": 0.3305,
+      "step": 18406
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8377632467515387e-05,
+      "loss": 0.0022,
+      "step": 18407
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8363671525045956e-05,
+      "loss": 0.0014,
+      "step": 18408
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.83497122101298e-05,
+      "loss": 0.9283,
+      "step": 18409
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.833575452299044e-05,
+      "loss": 0.636,
+      "step": 18410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.83217984638514e-05,
+      "loss": 0.4747,
+      "step": 18411
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8307844032936164e-05,
+      "loss": 1.3417,
+      "step": 18412
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.829389123046817e-05,
+      "loss": 0.5589,
+      "step": 18413
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8279940056670876e-05,
+      "loss": 0.4713,
+      "step": 18414
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8265990511767685e-05,
+      "loss": 0.0152,
+      "step": 18415
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.825204259598198e-05,
+      "loss": 0.6199,
+      "step": 18416
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8238096309537095e-05,
+      "loss": 0.0033,
+      "step": 18417
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.822415165265644e-05,
+      "loss": 0.7695,
+      "step": 18418
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.821020862556314e-05,
+      "loss": 0.0023,
+      "step": 18419
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.819626722848068e-05,
+      "loss": 0.0096,
+      "step": 18420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.818232746163219e-05,
+      "loss": 0.4088,
+      "step": 18421
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.816838932524085e-05,
+      "loss": 0.4495,
+      "step": 18422
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.815445281953002e-05,
+      "loss": 0.001,
+      "step": 18423
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.814051794472275e-05,
+      "loss": 0.1242,
+      "step": 18424
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.812658470104215e-05,
+      "loss": 0.3898,
+      "step": 18425
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.811265308871152e-05,
+      "loss": 0.4389,
+      "step": 18426
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.809872310795376e-05,
+      "loss": 0.7772,
+      "step": 18427
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.808479475899204e-05,
+      "loss": 0.2077,
+      "step": 18428
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8070868042049385e-05,
+      "loss": 0.8167,
+      "step": 18429
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8056942957348796e-05,
+      "loss": 0.4643,
+      "step": 18430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8043019505113264e-05,
+      "loss": 0.2584,
+      "step": 18431
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.802909768556576e-05,
+      "loss": 0.2759,
+      "step": 18432
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.8015177498929224e-05,
+      "loss": 0.1749,
+      "step": 18433
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.800125894542656e-05,
+      "loss": 1.6328,
+      "step": 18434
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.798734202528065e-05,
+      "loss": 0.0739,
+      "step": 18435
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7973426738714376e-05,
+      "loss": 0.8204,
+      "step": 18436
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.795951308595053e-05,
+      "loss": 0.3301,
+      "step": 18437
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.794560106721196e-05,
+      "loss": 0.0023,
+      "step": 18438
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.793169068272147e-05,
+      "loss": 1.3651,
+      "step": 18439
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.791778193270167e-05,
+      "loss": 0.4211,
+      "step": 18440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7903874817375426e-05,
+      "loss": 0.8202,
+      "step": 18441
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7889969336965474e-05,
+      "loss": 0.2002,
+      "step": 18442
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7876065491694296e-05,
+      "loss": 0.2676,
+      "step": 18443
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.786216328178472e-05,
+      "loss": 0.9688,
+      "step": 18444
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.784826270745936e-05,
+      "loss": 0.3537,
+      "step": 18445
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7834363768940686e-05,
+      "loss": 0.0026,
+      "step": 18446
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7820466466451375e-05,
+      "loss": 0.161,
+      "step": 18447
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7806570800213996e-05,
+      "loss": 0.2897,
+      "step": 18448
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7792676770450925e-05,
+      "loss": 0.33,
+      "step": 18449
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7778784377384823e-05,
+      "loss": 0.8144,
+      "step": 18450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.776489362123805e-05,
+      "loss": 0.3619,
+      "step": 18451
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.775100450223301e-05,
+      "loss": 0.4177,
+      "step": 18452
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.773711702059228e-05,
+      "loss": 0.7527,
+      "step": 18453
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.77232311765381e-05,
+      "loss": 0.5461,
+      "step": 18454
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.770934697029286e-05,
+      "loss": 0.6918,
+      "step": 18455
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.769546440207893e-05,
+      "loss": 0.7923,
+      "step": 18456
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.76815834721186e-05,
+      "loss": 0.4254,
+      "step": 18457
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.766770418063414e-05,
+      "loss": 0.1945,
+      "step": 18458
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7653826527847824e-05,
+      "loss": 0.001,
+      "step": 18459
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.763995051398188e-05,
+      "loss": 0.3178,
+      "step": 18460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.76260761392585e-05,
+      "loss": 0.148,
+      "step": 18461
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.761220340389988e-05,
+      "loss": 0.3068,
+      "step": 18462
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.759833230812815e-05,
+      "loss": 0.6935,
+      "step": 18463
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.758446285216545e-05,
+      "loss": 0.7734,
+      "step": 18464
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.757059503623387e-05,
+      "loss": 0.2796,
+      "step": 18465
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7556728860555484e-05,
+      "loss": 0.5352,
+      "step": 18466
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7542864325352345e-05,
+      "loss": 0.2501,
+      "step": 18467
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7529001430846474e-05,
+      "loss": 0.4396,
+      "step": 18468
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.751514017725989e-05,
+      "loss": 0.2555,
+      "step": 18469
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.750128056481448e-05,
+      "loss": 0.3898,
+      "step": 18470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.74874225937322e-05,
+      "loss": 0.1799,
+      "step": 18471
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7473566264235085e-05,
+      "loss": 0.332,
+      "step": 18472
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.745971157654488e-05,
+      "loss": 0.4581,
+      "step": 18473
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.744585853088347e-05,
+      "loss": 0.3455,
+      "step": 18474
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.743200712747281e-05,
+      "loss": 0.9498,
+      "step": 18475
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.741815736653458e-05,
+      "loss": 0.0015,
+      "step": 18476
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.740430924829059e-05,
+      "loss": 0.5591,
+      "step": 18477
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.739046277296264e-05,
+      "loss": 1.6195,
+      "step": 18478
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.737661794077241e-05,
+      "loss": 0.51,
+      "step": 18479
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.736277475194164e-05,
+      "loss": 0.2859,
+      "step": 18480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7348933206691984e-05,
+      "loss": 0.0362,
+      "step": 18481
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.733509330524511e-05,
+      "loss": 0.0014,
+      "step": 18482
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.732125504782265e-05,
+      "loss": 0.391,
+      "step": 18483
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.730741843464618e-05,
+      "loss": 0.011,
+      "step": 18484
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7293583465937284e-05,
+      "loss": 1.1684,
+      "step": 18485
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.727975014191751e-05,
+      "loss": 2.4756,
+      "step": 18486
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.726591846280838e-05,
+      "loss": 0.4176,
+      "step": 18487
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7252088428831424e-05,
+      "loss": 2.1177,
+      "step": 18488
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.723826004020798e-05,
+      "loss": 0.0991,
+      "step": 18489
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.722443329715962e-05,
+      "loss": 0.0374,
+      "step": 18490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7210608199907765e-05,
+      "loss": 0.2046,
+      "step": 18491
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7196784748673663e-05,
+      "loss": 0.476,
+      "step": 18492
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7182962943678805e-05,
+      "loss": 0.7862,
+      "step": 18493
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.716914278514453e-05,
+      "loss": 0.0016,
+      "step": 18494
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7155324273292034e-05,
+      "loss": 0.5557,
+      "step": 18495
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.71415074083427e-05,
+      "loss": 0.1007,
+      "step": 18496
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.712769219051782e-05,
+      "loss": 0.561,
+      "step": 18497
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.711387862003845e-05,
+      "loss": 0.7385,
+      "step": 18498
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.710006669712601e-05,
+      "loss": 0.4116,
+      "step": 18499
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7086256422001514e-05,
+      "loss": 0.001,
+      "step": 18500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7072447794886125e-05,
+      "loss": 0.1515,
+      "step": 18501
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.705864081600111e-05,
+      "loss": 3.4769,
+      "step": 18502
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.704483548556742e-05,
+      "loss": 0.1668,
+      "step": 18503
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7031031803806175e-05,
+      "loss": 0.7605,
+      "step": 18504
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.701722977093842e-05,
+      "loss": 0.0029,
+      "step": 18505
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.7003429387185175e-05,
+      "loss": 0.1345,
+      "step": 18506
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6989630652767435e-05,
+      "loss": 0.0025,
+      "step": 18507
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.697583356790617e-05,
+      "loss": 0.0408,
+      "step": 18508
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.69620381328223e-05,
+      "loss": 1.6384,
+      "step": 18509
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.694824434773675e-05,
+      "loss": 0.0027,
+      "step": 18510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6934452212870416e-05,
+      "loss": 0.4511,
+      "step": 18511
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6920661728444146e-05,
+      "loss": 0.3991,
+      "step": 18512
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6906872894678774e-05,
+      "loss": 3.2297,
+      "step": 18513
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6893085711795124e-05,
+      "loss": 0.195,
+      "step": 18514
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.687930018001395e-05,
+      "loss": 0.2845,
+      "step": 18515
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6865516299556025e-05,
+      "loss": 0.3079,
+      "step": 18516
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.685173407064208e-05,
+      "loss": 1.9938,
+      "step": 18517
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.683795349349286e-05,
+      "loss": 0.0021,
+      "step": 18518
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6824174568328886e-05,
+      "loss": 0.4616,
+      "step": 18519
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.681039729537095e-05,
+      "loss": 0.2221,
+      "step": 18520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.67966216748397e-05,
+      "loss": 0.0011,
+      "step": 18521
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6782847706955616e-05,
+      "loss": 0.5627,
+      "step": 18522
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.676907539193927e-05,
+      "loss": 0.0035,
+      "step": 18523
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.675530473001135e-05,
+      "loss": 0.3363,
+      "step": 18524
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6741535721392234e-05,
+      "loss": 0.0027,
+      "step": 18525
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.672776836630244e-05,
+      "loss": 0.3073,
+      "step": 18526
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6714002664962455e-05,
+      "loss": 0.7388,
+      "step": 18527
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.67002386175927e-05,
+      "loss": 0.5745,
+      "step": 18528
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6686476224413595e-05,
+      "loss": 0.1457,
+      "step": 18529
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6672715485645526e-05,
+      "loss": 0.3409,
+      "step": 18530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6658956401508845e-05,
+      "loss": 0.53,
+      "step": 18531
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.664519897222389e-05,
+      "loss": 1.1641,
+      "step": 18532
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.663144319801094e-05,
+      "loss": 0.2512,
+      "step": 18533
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.661768907909031e-05,
+      "loss": 0.0022,
+      "step": 18534
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.660393661568222e-05,
+      "loss": 0.8622,
+      "step": 18535
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6590185808006926e-05,
+      "loss": 0.5588,
+      "step": 18536
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6576436656284646e-05,
+      "loss": 0.3269,
+      "step": 18537
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.656268916073542e-05,
+      "loss": 0.347,
+      "step": 18538
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6548943321579536e-05,
+      "loss": 0.0093,
+      "step": 18539
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.653519913903713e-05,
+      "loss": 0.6226,
+      "step": 18540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6521456613328125e-05,
+      "loss": 0.3374,
+      "step": 18541
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.650771574467273e-05,
+      "loss": 0.3997,
+      "step": 18542
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.649397653329101e-05,
+      "loss": 0.3544,
+      "step": 18543
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6480238979402804e-05,
+      "loss": 0.0016,
+      "step": 18544
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.646650308322827e-05,
+      "loss": 0.1826,
+      "step": 18545
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6452768844987365e-05,
+      "loss": 0.7382,
+      "step": 18546
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6439036264899856e-05,
+      "loss": 0.3966,
+      "step": 18547
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6425305343185845e-05,
+      "loss": 0.0005,
+      "step": 18548
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.641157608006509e-05,
+      "loss": 0.2973,
+      "step": 18549
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.639784847575743e-05,
+      "loss": 0.1938,
+      "step": 18550
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.638412253048284e-05,
+      "loss": 0.2839,
+      "step": 18551
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.637039824446096e-05,
+      "loss": 0.3103,
+      "step": 18552
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.635667561791159e-05,
+      "loss": 0.0272,
+      "step": 18553
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.634295465105461e-05,
+      "loss": 0.4111,
+      "step": 18554
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.632923534410958e-05,
+      "loss": 2.6788,
+      "step": 18555
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.631551769729627e-05,
+      "loss": 0.5643,
+      "step": 18556
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.630180171083435e-05,
+      "loss": 0.2529,
+      "step": 18557
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6288087384943426e-05,
+      "loss": 4.1125,
+      "step": 18558
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6274374719843145e-05,
+      "loss": 0.4921,
+      "step": 18559
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6260663715753074e-05,
+      "loss": 0.2481,
+      "step": 18560
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6246954372892784e-05,
+      "loss": 0.219,
+      "step": 18561
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.623324669148181e-05,
+      "loss": 0.3867,
+      "step": 18562
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.621954067173966e-05,
+      "loss": 0.7806,
+      "step": 18563
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6205836313885795e-05,
+      "loss": 0.2588,
+      "step": 18564
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.619213361813972e-05,
+      "loss": 0.1194,
+      "step": 18565
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.61784325847208e-05,
+      "loss": 0.0034,
+      "step": 18566
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.616473321384854e-05,
+      "loss": 0.2443,
+      "step": 18567
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.6151035505742144e-05,
+      "loss": 0.6686,
+      "step": 18568
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.613733946062109e-05,
+      "loss": 0.0252,
+      "step": 18569
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.612364507870473e-05,
+      "loss": 0.0012,
+      "step": 18570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.6109952360212246e-05,
+      "loss": 0.0241,
+      "step": 18571
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.609626130536292e-05,
+      "loss": 0.2411,
+      "step": 18572
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.6082571914376104e-05,
+      "loss": 0.5093,
+      "step": 18573
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.6068884187470914e-05,
+      "loss": 0.4888,
+      "step": 18574
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.605519812486655e-05,
+      "loss": 0.0343,
+      "step": 18575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.6041513726782206e-05,
+      "loss": 0.5116,
+      "step": 18576
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.6027830993436984e-05,
+      "loss": 0.1563,
+      "step": 18577
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.6014149925050006e-05,
+      "loss": 0.0004,
+      "step": 18578
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.600047052184038e-05,
+      "loss": 0.5381,
+      "step": 18579
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.598679278402711e-05,
+      "loss": 0.0011,
+      "step": 18580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.597311671182927e-05,
+      "loss": 0.0004,
+      "step": 18581
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5959442305465845e-05,
+      "loss": 0.2022,
+      "step": 18582
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.59457695651558e-05,
+      "loss": 0.2819,
+      "step": 18583
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.593209849111809e-05,
+      "loss": 1.2018,
+      "step": 18584
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.591842908357163e-05,
+      "loss": 0.3993,
+      "step": 18585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5904761342735336e-05,
+      "loss": 0.0669,
+      "step": 18586
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.589109526882807e-05,
+      "loss": 0.9271,
+      "step": 18587
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.587743086206866e-05,
+      "loss": 0.7831,
+      "step": 18588
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.586376812267598e-05,
+      "loss": 0.001,
+      "step": 18589
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.585010705086866e-05,
+      "loss": 0.3181,
+      "step": 18590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.583644764686563e-05,
+      "loss": 0.2808,
+      "step": 18591
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.582278991088559e-05,
+      "loss": 0.8566,
+      "step": 18592
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.580913384314714e-05,
+      "loss": 0.2491,
+      "step": 18593
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5795479443869105e-05,
+      "loss": 0.2469,
+      "step": 18594
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.57818267132701e-05,
+      "loss": 2.3126,
+      "step": 18595
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.576817565156865e-05,
+      "loss": 0.1446,
+      "step": 18596
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.575452625898354e-05,
+      "loss": 0.6578,
+      "step": 18597
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.574087853573318e-05,
+      "loss": 0.1643,
+      "step": 18598
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.572723248203612e-05,
+      "loss": 0.2941,
+      "step": 18599
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5713588098111057e-05,
+      "loss": 0.0007,
+      "step": 18600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.569994538417631e-05,
+      "loss": 3.4479,
+      "step": 18601
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.568630434045037e-05,
+      "loss": 0.3865,
+      "step": 18602
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.56726649671518e-05,
+      "loss": 0.0232,
+      "step": 18603
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5659027264498884e-05,
+      "loss": 0.5168,
+      "step": 18604
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.564539123271004e-05,
+      "loss": 0.1298,
+      "step": 18605
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.563175687200365e-05,
+      "loss": 0.0083,
+      "step": 18606
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5618124182598045e-05,
+      "loss": 0.7512,
+      "step": 18607
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5604493164711514e-05,
+      "loss": 0.1069,
+      "step": 18608
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.559086381856235e-05,
+      "loss": 0.0005,
+      "step": 18609
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.557723614436881e-05,
+      "loss": 0.001,
+      "step": 18610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5563610142349124e-05,
+      "loss": 0.0052,
+      "step": 18611
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5549985812721474e-05,
+      "loss": 0.3275,
+      "step": 18612
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.553636315570404e-05,
+      "loss": 0.1462,
+      "step": 18613
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.552274217151499e-05,
+      "loss": 0.7143,
+      "step": 18614
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5509122860372416e-05,
+      "loss": 0.1645,
+      "step": 18615
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.549550522249445e-05,
+      "loss": 0.0028,
+      "step": 18616
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.548188925809906e-05,
+      "loss": 0.8255,
+      "step": 18617
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.54682749674044e-05,
+      "loss": 0.4401,
+      "step": 18618
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5454662350628475e-05,
+      "loss": 0.1532,
+      "step": 18619
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.544105140798913e-05,
+      "loss": 0.2368,
+      "step": 18620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5427442139704484e-05,
+      "loss": 1.2332,
+      "step": 18621
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.541383454599246e-05,
+      "loss": 0.0054,
+      "step": 18622
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5400228627070854e-05,
+      "loss": 0.3469,
+      "step": 18623
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.538662438315761e-05,
+      "loss": 0.3809,
+      "step": 18624
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.537302181447058e-05,
+      "loss": 0.0824,
+      "step": 18625
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5359420921227564e-05,
+      "loss": 0.0011,
+      "step": 18626
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.534582170364641e-05,
+      "loss": 0.5895,
+      "step": 18627
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.533222416194484e-05,
+      "loss": 0.489,
+      "step": 18628
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.531862829634061e-05,
+      "loss": 0.0722,
+      "step": 18629
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.530503410705143e-05,
+      "loss": 0.557,
+      "step": 18630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.529144159429504e-05,
+      "loss": 3.166,
+      "step": 18631
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.527785075828903e-05,
+      "loss": 0.0237,
+      "step": 18632
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.526426159925111e-05,
+      "loss": 0.7204,
+      "step": 18633
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.525067411739883e-05,
+      "loss": 0.1495,
+      "step": 18634
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.523708831294981e-05,
+      "loss": 0.3673,
+      "step": 18635
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.522350418612158e-05,
+      "loss": 0.273,
+      "step": 18636
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5209921737131696e-05,
+      "loss": 0.6928,
+      "step": 18637
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.51963409661977e-05,
+      "loss": 0.4097,
+      "step": 18638
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5182761873536903e-05,
+      "loss": 0.8245,
+      "step": 18639
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5169184459366924e-05,
+      "loss": 0.5541,
+      "step": 18640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.515560872390517e-05,
+      "loss": 0.3888,
+      "step": 18641
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.514203466736889e-05,
+      "loss": 0.1439,
+      "step": 18642
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.512846228997563e-05,
+      "loss": 0.4355,
+      "step": 18643
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.511489159194267e-05,
+      "loss": 0.5799,
+      "step": 18644
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5101322573487236e-05,
+      "loss": 4.1854,
+      "step": 18645
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.508775523482678e-05,
+      "loss": 0.2652,
+      "step": 18646
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.507418957617843e-05,
+      "loss": 0.5465,
+      "step": 18647
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5060625597759406e-05,
+      "loss": 0.0047,
+      "step": 18648
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5047063299787074e-05,
+      "loss": 0.0017,
+      "step": 18649
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.503350268247846e-05,
+      "loss": 0.572,
+      "step": 18650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.501994374605072e-05,
+      "loss": 0.7122,
+      "step": 18651
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.5006386490721116e-05,
+      "loss": 0.2115,
+      "step": 18652
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.499283091670662e-05,
+      "loss": 0.7793,
+      "step": 18653
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4979277024224345e-05,
+      "loss": 1.2657,
+      "step": 18654
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.496572481349133e-05,
+      "loss": 1.4732,
+      "step": 18655
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.495217428472459e-05,
+      "loss": 0.3066,
+      "step": 18656
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.493862543814114e-05,
+      "loss": 0.3052,
+      "step": 18657
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.492507827395793e-05,
+      "loss": 0.3576,
+      "step": 18658
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4911532792391874e-05,
+      "loss": 0.9959,
+      "step": 18659
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.489798899365993e-05,
+      "loss": 0.4071,
+      "step": 18660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.488444687797895e-05,
+      "loss": 0.208,
+      "step": 18661
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.487090644556577e-05,
+      "loss": 0.2727,
+      "step": 18662
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4857367696637285e-05,
+      "loss": 0.3923,
+      "step": 18663
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.484383063141024e-05,
+      "loss": 0.0021,
+      "step": 18664
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.483029525010147e-05,
+      "loss": 0.2758,
+      "step": 18665
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.48167615529276e-05,
+      "loss": 0.4756,
+      "step": 18666
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.480322954010548e-05,
+      "loss": 0.5356,
+      "step": 18667
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4789699211851824e-05,
+      "loss": 0.2314,
+      "step": 18668
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.477617056838312e-05,
+      "loss": 0.3279,
+      "step": 18669
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4762643609916185e-05,
+      "loss": 0.5942,
+      "step": 18670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.474911833666762e-05,
+      "loss": 0.2467,
+      "step": 18671
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.473559474885386e-05,
+      "loss": 0.475,
+      "step": 18672
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4722072846691674e-05,
+      "loss": 0.2791,
+      "step": 18673
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.470855263039745e-05,
+      "loss": 0.3163,
+      "step": 18674
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4695034100187726e-05,
+      "loss": 0.327,
+      "step": 18675
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.468151725627899e-05,
+      "loss": 0.018,
+      "step": 18676
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.466800209888769e-05,
+      "loss": 0.2126,
+      "step": 18677
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.465448862823026e-05,
+      "loss": 0.3652,
+      "step": 18678
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4640976844523096e-05,
+      "loss": 0.5671,
+      "step": 18679
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4627466747982574e-05,
+      "loss": 0.0028,
+      "step": 18680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.461395833882501e-05,
+      "loss": 0.382,
+      "step": 18681
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.460045161726676e-05,
+      "loss": 0.0009,
+      "step": 18682
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4586946583524085e-05,
+      "loss": 0.002,
+      "step": 18683
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4573443237813255e-05,
+      "loss": 4.3065,
+      "step": 18684
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.455994158035052e-05,
+      "loss": 0.9548,
+      "step": 18685
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4546441611352064e-05,
+      "loss": 1.923,
+      "step": 18686
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.453294333103409e-05,
+      "loss": 0.4379,
+      "step": 18687
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4519446739612736e-05,
+      "loss": 0.0011,
+      "step": 18688
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.450595183730415e-05,
+      "loss": 0.1486,
+      "step": 18689
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.449245862432446e-05,
+      "loss": 0.5159,
+      "step": 18690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4478967100889615e-05,
+      "loss": 0.4015,
+      "step": 18691
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4465477267215774e-05,
+      "loss": 1.0886,
+      "step": 18692
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4451989123519e-05,
+      "loss": 0.0051,
+      "step": 18693
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.443850267001511e-05,
+      "loss": 0.2449,
+      "step": 18694
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4425017906920266e-05,
+      "loss": 0.1577,
+      "step": 18695
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4411534834450266e-05,
+      "loss": 0.2684,
+      "step": 18696
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4398053452821045e-05,
+      "loss": 0.5409,
+      "step": 18697
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4384573762248576e-05,
+      "loss": 0.2586,
+      "step": 18698
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4371095762948616e-05,
+      "loss": 0.1879,
+      "step": 18699
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.435761945513699e-05,
+      "loss": 0.7344,
+      "step": 18700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.434414483902962e-05,
+      "loss": 0.3118,
+      "step": 18701
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4330671914842176e-05,
+      "loss": 0.1844,
+      "step": 18702
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.431720068279043e-05,
+      "loss": 0.1641,
+      "step": 18703
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.430373114309009e-05,
+      "loss": 0.5562,
+      "step": 18704
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.429026329595689e-05,
+      "loss": 2.4459,
+      "step": 18705
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.427679714160645e-05,
+      "loss": 0.0027,
+      "step": 18706
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.426333268025445e-05,
+      "loss": 0.9323,
+      "step": 18707
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4249869912116485e-05,
+      "loss": 0.0011,
+      "step": 18708
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.423640883740816e-05,
+      "loss": 0.0259,
+      "step": 18709
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4222949456345e-05,
+      "loss": 0.3091,
+      "step": 18710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4209491769142555e-05,
+      "loss": 0.4217,
+      "step": 18711
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.419603577601635e-05,
+      "loss": 0.6526,
+      "step": 18712
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4182581477181826e-05,
+      "loss": 0.0019,
+      "step": 18713
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.416912887285449e-05,
+      "loss": 0.2134,
+      "step": 18714
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4155677963249654e-05,
+      "loss": 0.3759,
+      "step": 18715
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.414222874858282e-05,
+      "loss": 0.4971,
+      "step": 18716
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.412878122906935e-05,
+      "loss": 0.4077,
+      "step": 18717
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4115335404924504e-05,
+      "loss": 0.1485,
+      "step": 18718
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.410189127636367e-05,
+      "loss": 0.4024,
+      "step": 18719
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.408844884360217e-05,
+      "loss": 0.2801,
+      "step": 18720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4075008106855106e-05,
+      "loss": 0.6526,
+      "step": 18721
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4061569066337933e-05,
+      "loss": 0.8127,
+      "step": 18722
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.404813172226567e-05,
+      "loss": 0.225,
+      "step": 18723
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.403469607485359e-05,
+      "loss": 0.1903,
+      "step": 18724
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.4021262124316804e-05,
+      "loss": 0.2298,
+      "step": 18725
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.400782987087046e-05,
+      "loss": 0.2665,
+      "step": 18726
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.399439931472964e-05,
+      "loss": 0.2082,
+      "step": 18727
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3980970456109437e-05,
+      "loss": 0.4696,
+      "step": 18728
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.396754329522487e-05,
+      "loss": 0.5902,
+      "step": 18729
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.395411783229097e-05,
+      "loss": 0.7862,
+      "step": 18730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.394069406752271e-05,
+      "loss": 1.4959,
+      "step": 18731
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3927272001135064e-05,
+      "loss": 0.6953,
+      "step": 18732
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.391385163334297e-05,
+      "loss": 2.5313,
+      "step": 18733
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.390043296436132e-05,
+      "loss": 0.4594,
+      "step": 18734
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.388701599440502e-05,
+      "loss": 0.4357,
+      "step": 18735
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.387360072368889e-05,
+      "loss": 0.3855,
+      "step": 18736
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.386018715242778e-05,
+      "loss": 0.2157,
+      "step": 18737
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3846775280836484e-05,
+      "loss": 0.2546,
+      "step": 18738
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.383336510912975e-05,
+      "loss": 0.2523,
+      "step": 18739
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.381995663752236e-05,
+      "loss": 0.2788,
+      "step": 18740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.380654986622899e-05,
+      "loss": 0.0084,
+      "step": 18741
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.379314479546441e-05,
+      "loss": 0.0184,
+      "step": 18742
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3779741425443124e-05,
+      "loss": 0.5373,
+      "step": 18743
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.376633975637996e-05,
+      "loss": 0.0007,
+      "step": 18744
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.375293978848938e-05,
+      "loss": 3.2669,
+      "step": 18745
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.373954152198599e-05,
+      "loss": 1.7844,
+      "step": 18746
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3726144957084436e-05,
+      "loss": 0.6354,
+      "step": 18747
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.371275009399914e-05,
+      "loss": 0.7405,
+      "step": 18748
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3699356932944575e-05,
+      "loss": 0.7997,
+      "step": 18749
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3685965474135385e-05,
+      "loss": 0.7291,
+      "step": 18750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.367257571778584e-05,
+      "loss": 0.5074,
+      "step": 18751
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.365918766411043e-05,
+      "loss": 0.0014,
+      "step": 18752
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.364580131332355e-05,
+      "loss": 0.5293,
+      "step": 18753
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.363241666563952e-05,
+      "loss": 0.6694,
+      "step": 18754
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3619033721272706e-05,
+      "loss": 0.4567,
+      "step": 18755
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.360565248043742e-05,
+      "loss": 0.2539,
+      "step": 18756
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.359227294334793e-05,
+      "loss": 0.562,
+      "step": 18757
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.357889511021849e-05,
+      "loss": 0.0003,
+      "step": 18758
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.356551898126334e-05,
+      "loss": 0.829,
+      "step": 18759
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3552144556696656e-05,
+      "loss": 0.0382,
+      "step": 18760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3538771836732625e-05,
+      "loss": 0.9283,
+      "step": 18761
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.35254008215854e-05,
+      "loss": 0.4743,
+      "step": 18762
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.351203151146911e-05,
+      "loss": 0.2036,
+      "step": 18763
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.349866390659774e-05,
+      "loss": 0.5981,
+      "step": 18764
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.348529800718548e-05,
+      "loss": 0.9736,
+      "step": 18765
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3471933813446346e-05,
+      "loss": 0.5801,
+      "step": 18766
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3458571325594245e-05,
+      "loss": 0.0016,
+      "step": 18767
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.344521054384325e-05,
+      "loss": 0.9083,
+      "step": 18768
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.343185146840736e-05,
+      "loss": 0.1743,
+      "step": 18769
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.341849409950033e-05,
+      "loss": 0.6535,
+      "step": 18770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3405138437336256e-05,
+      "loss": 0.0019,
+      "step": 18771
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.339178448212886e-05,
+      "loss": 0.0028,
+      "step": 18772
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3378432234092e-05,
+      "loss": 0.2346,
+      "step": 18773
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.336508169343963e-05,
+      "loss": 0.7615,
+      "step": 18774
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3351732860385386e-05,
+      "loss": 0.5029,
+      "step": 18775
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3338385735143063e-05,
+      "loss": 0.1144,
+      "step": 18776
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.332504031792645e-05,
+      "loss": 0.5073,
+      "step": 18777
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.33116966089492e-05,
+      "loss": 0.6198,
+      "step": 18778
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.329835460842502e-05,
+      "loss": 0.4901,
+      "step": 18779
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3285014316567554e-05,
+      "loss": 0.3772,
+      "step": 18780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3271675733590414e-05,
+      "loss": 1.0703,
+      "step": 18781
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3258338859707226e-05,
+      "loss": 0.8385,
+      "step": 18782
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3245003695131546e-05,
+      "loss": 0.0709,
+      "step": 18783
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.323167024007691e-05,
+      "loss": 0.9499,
+      "step": 18784
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.321833849475683e-05,
+      "loss": 0.0015,
+      "step": 18785
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.320500845938481e-05,
+      "loss": 0.5577,
+      "step": 18786
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.319168013417431e-05,
+      "loss": 0.5483,
+      "step": 18787
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.317835351933874e-05,
+      "loss": 0.0005,
+      "step": 18788
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3165028615091515e-05,
+      "loss": 0.7572,
+      "step": 18789
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3151705421646045e-05,
+      "loss": 0.2942,
+      "step": 18790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.31383839392157e-05,
+      "loss": 1.0123,
+      "step": 18791
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.312506416801365e-05,
+      "loss": 0.0049,
+      "step": 18792
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.31117461082534e-05,
+      "loss": 0.2833,
+      "step": 18793
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.309842976014807e-05,
+      "loss": 0.0019,
+      "step": 18794
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.308511512391092e-05,
+      "loss": 0.7167,
+      "step": 18795
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.307180219975526e-05,
+      "loss": 0.0009,
+      "step": 18796
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.305849098789418e-05,
+      "loss": 0.5548,
+      "step": 18797
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.304518148854083e-05,
+      "loss": 0.292,
+      "step": 18798
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3031873701908465e-05,
+      "loss": 0.0005,
+      "step": 18799
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.301856762821006e-05,
+      "loss": 1.3156,
+      "step": 18800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.3005263267658744e-05,
+      "loss": 0.0026,
+      "step": 18801
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.299196062046755e-05,
+      "loss": 0.2576,
+      "step": 18802
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.297865968684953e-05,
+      "loss": 0.2598,
+      "step": 18803
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2965360467017644e-05,
+      "loss": 0.4066,
+      "step": 18804
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2952062961184875e-05,
+      "loss": 0.0096,
+      "step": 18805
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.293876716956416e-05,
+      "loss": 0.0006,
+      "step": 18806
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2925473092368415e-05,
+      "loss": 0.5622,
+      "step": 18807
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.291218072981052e-05,
+      "loss": 0.2008,
+      "step": 18808
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2898890082103327e-05,
+      "loss": 0.5795,
+      "step": 18809
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.288560114945968e-05,
+      "loss": 2.4329,
+      "step": 18810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.287231393209237e-05,
+      "loss": 0.2797,
+      "step": 18811
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2859028430214235e-05,
+      "loss": 0.5724,
+      "step": 18812
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.284574464403787e-05,
+      "loss": 0.6805,
+      "step": 18813
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2832462573776124e-05,
+      "loss": 0.0673,
+      "step": 18814
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.281918221964171e-05,
+      "loss": 0.0006,
+      "step": 18815
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.280590358184714e-05,
+      "loss": 0.7375,
+      "step": 18816
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.279262666060518e-05,
+      "loss": 0.0017,
+      "step": 18817
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.2779351456128474e-05,
+      "loss": 0.0238,
+      "step": 18818
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.276607796862944e-05,
+      "loss": 0.7233,
+      "step": 18819
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.275280619832083e-05,
+      "loss": 0.2804,
+      "step": 18820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.273953614541503e-05,
+      "loss": 0.3682,
+      "step": 18821
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.272626781012452e-05,
+      "loss": 1.1587,
+      "step": 18822
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2713001192661925e-05,
+      "loss": 0.4498,
+      "step": 18823
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.269973629323958e-05,
+      "loss": 0.0072,
+      "step": 18824
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.268647311206985e-05,
+      "loss": 0.1198,
+      "step": 18825
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2673211649365314e-05,
+      "loss": 0.3378,
+      "step": 18826
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2659951905338156e-05,
+      "loss": 0.0153,
+      "step": 18827
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2646693880200786e-05,
+      "loss": 0.0038,
+      "step": 18828
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.263343757416548e-05,
+      "loss": 1.3156,
+      "step": 18829
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.262018298744456e-05,
+      "loss": 0.0021,
+      "step": 18830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.260693012025024e-05,
+      "loss": 0.5264,
+      "step": 18831
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.259367897279476e-05,
+      "loss": 0.5631,
+      "step": 18832
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2580429545290324e-05,
+      "loss": 0.2253,
+      "step": 18833
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.25671818379491e-05,
+      "loss": 0.3225,
+      "step": 18834
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.25539358509832e-05,
+      "loss": 0.662,
+      "step": 18835
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.254069158460479e-05,
+      "loss": 1.1045,
+      "step": 18836
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.252744903902592e-05,
+      "loss": 0.265,
+      "step": 18837
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.251420821445866e-05,
+      "loss": 0.6878,
+      "step": 18838
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2500969111115036e-05,
+      "loss": 0.5751,
+      "step": 18839
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.248773172920707e-05,
+      "loss": 0.4997,
+      "step": 18840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.247449606894671e-05,
+      "loss": 0.4247,
+      "step": 18841
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2461262130545994e-05,
+      "loss": 0.0011,
+      "step": 18842
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.244802991421671e-05,
+      "loss": 0.4074,
+      "step": 18843
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2434799420170754e-05,
+      "loss": 1.1575,
+      "step": 18844
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.242157064862017e-05,
+      "loss": 0.2363,
+      "step": 18845
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2408343599776626e-05,
+      "loss": 0.3594,
+      "step": 18846
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.239511827385194e-05,
+      "loss": 0.0352,
+      "step": 18847
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.238189467105802e-05,
+      "loss": 0.3511,
+      "step": 18848
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.236867279160651e-05,
+      "loss": 1.4784,
+      "step": 18849
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.235545263570918e-05,
+      "loss": 0.0049,
+      "step": 18850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.234223420357771e-05,
+      "loss": 0.1921,
+      "step": 18851
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.23290174954238e-05,
+      "loss": 0.1243,
+      "step": 18852
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2315802511459084e-05,
+      "loss": 0.7275,
+      "step": 18853
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.230258925189518e-05,
+      "loss": 0.2703,
+      "step": 18854
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.228937771694367e-05,
+      "loss": 0.4443,
+      "step": 18855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.227616790681613e-05,
+      "loss": 1.239,
+      "step": 18856
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2262959821724083e-05,
+      "loss": 0.1662,
+      "step": 18857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.224975346187904e-05,
+      "loss": 0.0118,
+      "step": 18858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.22365488274925e-05,
+      "loss": 0.001,
+      "step": 18859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.222334591877589e-05,
+      "loss": 0.1788,
+      "step": 18860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.221014473594069e-05,
+      "loss": 0.4189,
+      "step": 18861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2196945279198165e-05,
+      "loss": 1.1238,
+      "step": 18862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2183747548759825e-05,
+      "loss": 0.2827,
+      "step": 18863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.217055154483698e-05,
+      "loss": 0.3271,
+      "step": 18864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.215735726764085e-05,
+      "loss": 0.408,
+      "step": 18865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.214416471738284e-05,
+      "loss": 0.5038,
+      "step": 18866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.213097389427421e-05,
+      "loss": 0.3982,
+      "step": 18867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.211778479852605e-05,
+      "loss": 0.4377,
+      "step": 18868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2104597430349755e-05,
+      "loss": 0.21,
+      "step": 18869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.209141178995635e-05,
+      "loss": 0.3549,
+      "step": 18870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2078227877556986e-05,
+      "loss": 0.4179,
+      "step": 18871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.206504569336294e-05,
+      "loss": 0.8213,
+      "step": 18872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2051865237585146e-05,
+      "loss": 0.4114,
+      "step": 18873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.203868651043468e-05,
+      "loss": 0.3492,
+      "step": 18874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.2025509512122704e-05,
+      "loss": 1.2701,
+      "step": 18875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.201233424286012e-05,
+      "loss": 0.0151,
+      "step": 18876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.199916070285792e-05,
+      "loss": 0.3925,
+      "step": 18877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.198598889232708e-05,
+      "loss": 0.1379,
+      "step": 18878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.197281881147852e-05,
+      "loss": 0.0004,
+      "step": 18879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.195965046052314e-05,
+      "loss": 0.2729,
+      "step": 18880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1946483839671807e-05,
+      "loss": 0.0045,
+      "step": 18881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.193331894913538e-05,
+      "loss": 0.1656,
+      "step": 18882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.192015578912466e-05,
+      "loss": 0.8136,
+      "step": 18883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.190699435985044e-05,
+      "loss": 0.7936,
+      "step": 18884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.189383466152348e-05,
+      "loss": 0.2152,
+      "step": 18885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1880676694354515e-05,
+      "loss": 1.0528,
+      "step": 18886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.186752045855425e-05,
+      "loss": 0.3083,
+      "step": 18887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.185436595433336e-05,
+      "loss": 0.1742,
+      "step": 18888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.184121318190249e-05,
+      "loss": 0.0008,
+      "step": 18889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.182806214147227e-05,
+      "loss": 0.2633,
+      "step": 18890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.181491283325333e-05,
+      "loss": 0.5676,
+      "step": 18891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.180176525745611e-05,
+      "loss": 0.4,
+      "step": 18892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.17886194142913e-05,
+      "loss": 0.2096,
+      "step": 18893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.177547530396937e-05,
+      "loss": 0.7964,
+      "step": 18894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1762332926700745e-05,
+      "loss": 0.403,
+      "step": 18895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1749192282695886e-05,
+      "loss": 0.3768,
+      "step": 18896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1736053372165325e-05,
+      "loss": 0.178,
+      "step": 18897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.172291619531934e-05,
+      "loss": 0.5835,
+      "step": 18898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.170978075236835e-05,
+      "loss": 0.2902,
+      "step": 18899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.169664704352272e-05,
+      "loss": 0.0016,
+      "step": 18900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.168351506899273e-05,
+      "loss": 0.4077,
+      "step": 18901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.167038482898868e-05,
+      "loss": 0.3018,
+      "step": 18902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.165725632372085e-05,
+      "loss": 0.2335,
+      "step": 18903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.164412955339946e-05,
+      "loss": 0.2302,
+      "step": 18904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.16310045182347e-05,
+      "loss": 0.0021,
+      "step": 18905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1617881218436775e-05,
+      "loss": 0.6075,
+      "step": 18906
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.160475965421584e-05,
+      "loss": 0.3839,
+      "step": 18907
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.159163982578198e-05,
+      "loss": 0.4157,
+      "step": 18908
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.157852173334531e-05,
+      "loss": 0.391,
+      "step": 18909
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.156540537711594e-05,
+      "loss": 0.6519,
+      "step": 18910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.155229075730378e-05,
+      "loss": 0.8638,
+      "step": 18911
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.153917787411898e-05,
+      "loss": 0.1766,
+      "step": 18912
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1526066727771504e-05,
+      "loss": 0.5285,
+      "step": 18913
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.151295731847118e-05,
+      "loss": 0.2984,
+      "step": 18914
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.149984964642808e-05,
+      "loss": 1.061,
+      "step": 18915
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.14867437118521e-05,
+      "loss": 0.2981,
+      "step": 18916
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.147363951495297e-05,
+      "loss": 0.0004,
+      "step": 18917
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1460537055940714e-05,
+      "loss": 0.3633,
+      "step": 18918
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.144743633502502e-05,
+      "loss": 0.0505,
+      "step": 18919
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.143433735241567e-05,
+      "loss": 0.9988,
+      "step": 18920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1421240108322577e-05,
+      "loss": 0.6555,
+      "step": 18921
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.140814460295531e-05,
+      "loss": 1.7529,
+      "step": 18922
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.139505083652359e-05,
+      "loss": 0.0861,
+      "step": 18923
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.138195880923725e-05,
+      "loss": 0.4171,
+      "step": 18924
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1368868521305745e-05,
+      "loss": 0.5032,
+      "step": 18925
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1355779972938754e-05,
+      "loss": 0.3482,
+      "step": 18926
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.134269316434599e-05,
+      "loss": 0.2924,
+      "step": 18927
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.132960809573687e-05,
+      "loss": 1.6626,
+      "step": 18928
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.131652476732098e-05,
+      "loss": 0.2969,
+      "step": 18929
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.130344317930785e-05,
+      "loss": 0.0004,
+      "step": 18930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1290363331906925e-05,
+      "loss": 0.4469,
+      "step": 18931
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.12772852253277e-05,
+      "loss": 0.4317,
+      "step": 18932
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.126420885977958e-05,
+      "loss": 0.5728,
+      "step": 18933
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.125113423547196e-05,
+      "loss": 0.2633,
+      "step": 18934
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.123806135261421e-05,
+      "loss": 0.2448,
+      "step": 18935
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1224990211415675e-05,
+      "loss": 0.0188,
+      "step": 18936
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.121192081208569e-05,
+      "loss": 0.4825,
+      "step": 18937
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1198853154833524e-05,
+      "loss": 0.003,
+      "step": 18938
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1185787239868434e-05,
+      "loss": 0.5263,
+      "step": 18939
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1172723067399695e-05,
+      "loss": 2.9531,
+      "step": 18940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.115966063763638e-05,
+      "loss": 0.5938,
+      "step": 18941
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.11465999507878e-05,
+      "loss": 0.0023,
+      "step": 18942
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.11335410070631e-05,
+      "loss": 0.5124,
+      "step": 18943
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1120483806671306e-05,
+      "loss": 1.507,
+      "step": 18944
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1107428349821526e-05,
+      "loss": 0.729,
+      "step": 18945
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.109437463672294e-05,
+      "loss": 1.3244,
+      "step": 18946
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1081322667584446e-05,
+      "loss": 0.4895,
+      "step": 18947
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1068272442615106e-05,
+      "loss": 0.0017,
+      "step": 18948
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.105522396202391e-05,
+      "loss": 0.3348,
+      "step": 18949
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1042177226019793e-05,
+      "loss": 0.2423,
+      "step": 18950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.1029132234811686e-05,
+      "loss": 0.1898,
+      "step": 18951
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.101608898860848e-05,
+      "loss": 0.3051,
+      "step": 18952
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.100304748761905e-05,
+      "loss": 1.3498,
+      "step": 18953
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.099000773205224e-05,
+      "loss": 0.4872,
+      "step": 18954
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0976969722116844e-05,
+      "loss": 0.0005,
+      "step": 18955
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.096393345802166e-05,
+      "loss": 0.3801,
+      "step": 18956
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.095089893997544e-05,
+      "loss": 2.0039,
+      "step": 18957
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.093786616818692e-05,
+      "loss": 0.3543,
+      "step": 18958
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.092483514286479e-05,
+      "loss": 0.2855,
+      "step": 18959
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.091180586421773e-05,
+      "loss": 1.6519,
+      "step": 18960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.089877833245438e-05,
+      "loss": 0.7946,
+      "step": 18961
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.088575254778339e-05,
+      "loss": 0.0012,
+      "step": 18962
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.087272851041325e-05,
+      "loss": 0.3999,
+      "step": 18963
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.085970622055261e-05,
+      "loss": 0.6621,
+      "step": 18964
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0846685678410035e-05,
+      "loss": 0.1821,
+      "step": 18965
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0833666884193884e-05,
+      "loss": 0.5558,
+      "step": 18966
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0820649838112803e-05,
+      "loss": 0.449,
+      "step": 18967
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.080763454037511e-05,
+      "loss": 0.3799,
+      "step": 18968
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.079462099118923e-05,
+      "loss": 1.0815,
+      "step": 18969
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0781609190763706e-05,
+      "loss": 0.5166,
+      "step": 18970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.076859913930674e-05,
+      "loss": 0.917,
+      "step": 18971
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.075559083702669e-05,
+      "loss": 0.5535,
+      "step": 18972
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.074258428413198e-05,
+      "loss": 0.0039,
+      "step": 18973
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.072957948083077e-05,
+      "loss": 0.0219,
+      "step": 18974
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.07165764273313e-05,
+      "loss": 0.4504,
+      "step": 18975
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.070357512384195e-05,
+      "loss": 0.4054,
+      "step": 18976
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.069057557057076e-05,
+      "loss": 0.3289,
+      "step": 18977
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.067757776772595e-05,
+      "loss": 0.0009,
+      "step": 18978
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0664581715515664e-05,
+      "loss": 0.1714,
+      "step": 18979
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.065158741414803e-05,
+      "loss": 0.9221,
+      "step": 18980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.06385948638311e-05,
+      "loss": 0.434,
+      "step": 18981
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.062560406477294e-05,
+      "loss": 0.0025,
+      "step": 18982
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.061261501718158e-05,
+      "loss": 0.2909,
+      "step": 18983
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.059962772126503e-05,
+      "loss": 0.8345,
+      "step": 18984
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.058664217723125e-05,
+      "loss": 0.001,
+      "step": 18985
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.05736583852882e-05,
+      "loss": 1.2701,
+      "step": 18986
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.056067634564378e-05,
+      "loss": 0.0005,
+      "step": 18987
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0547696058505874e-05,
+      "loss": 0.2197,
+      "step": 18988
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.053471752408241e-05,
+      "loss": 0.2708,
+      "step": 18989
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.052174074258106e-05,
+      "loss": 0.3605,
+      "step": 18990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.050876571420979e-05,
+      "loss": 0.2807,
+      "step": 18991
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.049579243917634e-05,
+      "loss": 0.6099,
+      "step": 18992
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.048282091768834e-05,
+      "loss": 0.8876,
+      "step": 18993
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.046985114995366e-05,
+      "loss": 0.0418,
+      "step": 18994
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.045688313617996e-05,
+      "loss": 2.0254,
+      "step": 18995
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.044391687657485e-05,
+      "loss": 0.0013,
+      "step": 18996
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0430952371345975e-05,
+      "loss": 0.0884,
+      "step": 18997
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0417989620700975e-05,
+      "loss": 1.2071,
+      "step": 18998
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.040502862484738e-05,
+      "loss": 0.2606,
+      "step": 18999
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.039206938399281e-05,
+      "loss": 0.5271,
+      "step": 19000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.037911189834473e-05,
+      "loss": 0.4859,
+      "step": 19001
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.036615616811066e-05,
+      "loss": 0.2758,
+      "step": 19002
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.035320219349806e-05,
+      "loss": 0.6214,
+      "step": 19003
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.034024997471436e-05,
+      "loss": 0.0432,
+      "step": 19004
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0327299511966995e-05,
+      "loss": 0.726,
+      "step": 19005
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0314350805463326e-05,
+      "loss": 0.4813,
+      "step": 19006
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.030140385541072e-05,
+      "loss": 0.7348,
+      "step": 19007
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.028845866201649e-05,
+      "loss": 0.4121,
+      "step": 19008
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.027551522548795e-05,
+      "loss": 1.0207,
+      "step": 19009
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.026257354603234e-05,
+      "loss": 0.0651,
+      "step": 19010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.024963362385698e-05,
+      "loss": 0.9315,
+      "step": 19011
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.023669545916894e-05,
+      "loss": 0.4799,
+      "step": 19012
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.022375905217553e-05,
+      "loss": 0.9056,
+      "step": 19013
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0210824403083915e-05,
+      "loss": 0.3154,
+      "step": 19014
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.019789151210109e-05,
+      "loss": 0.165,
+      "step": 19015
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0184960379434324e-05,
+      "loss": 0.285,
+      "step": 19016
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.017203100529057e-05,
+      "loss": 0.1594,
+      "step": 19017
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.015910338987686e-05,
+      "loss": 0.6226,
+      "step": 19018
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.014617753340036e-05,
+      "loss": 0.2916,
+      "step": 19019
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0133253436067905e-05,
+      "loss": 0.604,
+      "step": 19020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.012033109808647e-05,
+      "loss": 0.0639,
+      "step": 19021
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0107410519663104e-05,
+      "loss": 0.8631,
+      "step": 19022
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.00944917010046e-05,
+      "loss": 0.5559,
+      "step": 19023
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.008157464231782e-05,
+      "loss": 0.3051,
+      "step": 19024
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.006865934380975e-05,
+      "loss": 0.1659,
+      "step": 19025
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0055745805687073e-05,
+      "loss": 0.7096,
+      "step": 19026
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.004283402815662e-05,
+      "loss": 0.0868,
+      "step": 19027
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.002992401142515e-05,
+      "loss": 0.7681,
+      "step": 19028
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.001701575569941e-05,
+      "loss": 0.0856,
+      "step": 19029
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.0004109261186096e-05,
+      "loss": 0.6891,
+      "step": 19030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.999120452809191e-05,
+      "loss": 0.0006,
+      "step": 19031
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9978301556623465e-05,
+      "loss": 0.8228,
+      "step": 19032
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.99654003469874e-05,
+      "loss": 0.3848,
+      "step": 19033
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9952500899390325e-05,
+      "loss": 0.2367,
+      "step": 19034
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.993960321403879e-05,
+      "loss": 0.5669,
+      "step": 19035
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.992670729113931e-05,
+      "loss": 0.302,
+      "step": 19036
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.991381313089843e-05,
+      "loss": 0.107,
+      "step": 19037
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9900920733522676e-05,
+      "loss": 0.339,
+      "step": 19038
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.988803009921832e-05,
+      "loss": 0.013,
+      "step": 19039
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9875141228191976e-05,
+      "loss": 0.5576,
+      "step": 19040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9862254120650005e-05,
+      "loss": 0.344,
+      "step": 19041
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.984936877679866e-05,
+      "loss": 0.3278,
+      "step": 19042
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9836485196844395e-05,
+      "loss": 1.1099,
+      "step": 19043
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.982360338099355e-05,
+      "loss": 0.4182,
+      "step": 19044
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.981072332945223e-05,
+      "loss": 0.4219,
+      "step": 19045
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.979784504242691e-05,
+      "loss": 0.5454,
+      "step": 19046
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.978496852012367e-05,
+      "loss": 0.3422,
+      "step": 19047
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.977209376274874e-05,
+      "loss": 0.0008,
+      "step": 19048
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9759220770508296e-05,
+      "loss": 0.154,
+      "step": 19049
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9746349543608497e-05,
+      "loss": 0.2622,
+      "step": 19050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.973348008225544e-05,
+      "loss": 0.1558,
+      "step": 19051
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9720612386655206e-05,
+      "loss": 1.0574,
+      "step": 19052
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9707746457013854e-05,
+      "loss": 0.268,
+      "step": 19053
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.969488229353743e-05,
+      "loss": 0.7726,
+      "step": 19054
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.968201989643192e-05,
+      "loss": 0.4771,
+      "step": 19055
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9669159265903284e-05,
+      "loss": 0.3113,
+      "step": 19056
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.965630040215748e-05,
+      "loss": 0.6694,
+      "step": 19057
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.964344330540044e-05,
+      "loss": 0.2021,
+      "step": 19058
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.963058797583801e-05,
+      "loss": 0.0006,
+      "step": 19059
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.96177344136761e-05,
+      "loss": 0.1895,
+      "step": 19060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.96048826191205e-05,
+      "loss": 0.4946,
+      "step": 19061
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.959203259237701e-05,
+      "loss": 0.7782,
+      "step": 19062
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.957918433365149e-05,
+      "loss": 0.863,
+      "step": 19063
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.95663378431495e-05,
+      "loss": 0.264,
+      "step": 19064
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.955349312107698e-05,
+      "loss": 2.4027,
+      "step": 19065
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9540650167639464e-05,
+      "loss": 0.29,
+      "step": 19066
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9527808983042606e-05,
+      "loss": 0.2765,
+      "step": 19067
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.951496956749218e-05,
+      "loss": 0.3821,
+      "step": 19068
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.950213192119366e-05,
+      "loss": 0.0089,
+      "step": 19069
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.948929604435262e-05,
+      "loss": 0.0011,
+      "step": 19070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.947646193717475e-05,
+      "loss": 0.5306,
+      "step": 19071
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.946362959986541e-05,
+      "loss": 0.0073,
+      "step": 19072
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.945079903263011e-05,
+      "loss": 0.7488,
+      "step": 19073
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.943797023567443e-05,
+      "loss": 0.0056,
+      "step": 19074
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9425143209203695e-05,
+      "loss": 0.3802,
+      "step": 19075
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.941231795342334e-05,
+      "loss": 0.7273,
+      "step": 19076
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9399494468538746e-05,
+      "loss": 0.3356,
+      "step": 19077
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9386672754755264e-05,
+      "loss": 0.623,
+      "step": 19078
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9373852812278204e-05,
+      "loss": 0.3724,
+      "step": 19079
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.936103464131287e-05,
+      "loss": 0.7492,
+      "step": 19080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.934821824206452e-05,
+      "loss": 0.3367,
+      "step": 19081
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9335403614738374e-05,
+      "loss": 1.62,
+      "step": 19082
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.932259075953966e-05,
+      "loss": 0.2737,
+      "step": 19083
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9309779676673565e-05,
+      "loss": 0.2027,
+      "step": 19084
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.929697036634522e-05,
+      "loss": 0.3966,
+      "step": 19085
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.928416282875975e-05,
+      "loss": 0.2784,
+      "step": 19086
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.927135706412229e-05,
+      "loss": 0.2021,
+      "step": 19087
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9258553072637793e-05,
+      "loss": 0.0023,
+      "step": 19088
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.924575085451141e-05,
+      "loss": 0.8684,
+      "step": 19089
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.923295040994814e-05,
+      "loss": 0.5517,
+      "step": 19090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.922015173915285e-05,
+      "loss": 0.001,
+      "step": 19091
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.920735484233061e-05,
+      "loss": 0.1623,
+      "step": 19092
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9194559719686355e-05,
+      "loss": 0.6347,
+      "step": 19093
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.918176637142484e-05,
+      "loss": 0.4717,
+      "step": 19094
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.916897479775112e-05,
+      "loss": 0.3065,
+      "step": 19095
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.915618499886988e-05,
+      "loss": 0.3603,
+      "step": 19096
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9143396974985986e-05,
+      "loss": 0.4858,
+      "step": 19097
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.913061072630423e-05,
+      "loss": 0.2146,
+      "step": 19098
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9117826253029325e-05,
+      "loss": 0.2198,
+      "step": 19099
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.910504355536605e-05,
+      "loss": 0.7223,
+      "step": 19100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.909226263351905e-05,
+      "loss": 1.1595,
+      "step": 19101
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9079483487693034e-05,
+      "loss": 0.1915,
+      "step": 19102
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.906670611809261e-05,
+      "loss": 0.0005,
+      "step": 19103
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9053930524922395e-05,
+      "loss": 0.3429,
+      "step": 19104
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9041156708386984e-05,
+      "loss": 0.2753,
+      "step": 19105
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9028384668690916e-05,
+      "loss": 0.001,
+      "step": 19106
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9015614406038726e-05,
+      "loss": 0.8614,
+      "step": 19107
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9002845920634915e-05,
+      "loss": 0.0012,
+      "step": 19108
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.899007921268393e-05,
+      "loss": 0.207,
+      "step": 19109
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.897731428239024e-05,
+      "loss": 0.2204,
+      "step": 19110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8964551129958234e-05,
+      "loss": 0.4967,
+      "step": 19111
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.895178975559233e-05,
+      "loss": 0.1166,
+      "step": 19112
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.893903015949678e-05,
+      "loss": 0.0008,
+      "step": 19113
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.892627234187606e-05,
+      "loss": 0.2479,
+      "step": 19114
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8913516302934364e-05,
+      "loss": 0.3852,
+      "step": 19115
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.890076204287593e-05,
+      "loss": 1.136,
+      "step": 19116
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8888009561905146e-05,
+      "loss": 0.3827,
+      "step": 19117
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8875258860226084e-05,
+      "loss": 0.0222,
+      "step": 19118
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.886250993804294e-05,
+      "loss": 0.5653,
+      "step": 19119
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.884976279555999e-05,
+      "loss": 0.9372,
+      "step": 19120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.883701743298124e-05,
+      "loss": 0.6063,
+      "step": 19121
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.882427385051076e-05,
+      "loss": 0.8542,
+      "step": 19122
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.881153204835278e-05,
+      "loss": 0.4483,
+      "step": 19123
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.879879202671121e-05,
+      "loss": 0.8055,
+      "step": 19124
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.878605378579008e-05,
+      "loss": 0.6899,
+      "step": 19125
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.87733173257934e-05,
+      "loss": 0.001,
+      "step": 19126
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8760582646925114e-05,
+      "loss": 0.0019,
+      "step": 19127
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.874784974938914e-05,
+      "loss": 0.1832,
+      "step": 19128
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.873511863338941e-05,
+      "loss": 0.4314,
+      "step": 19129
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.872238929912974e-05,
+      "loss": 0.2816,
+      "step": 19130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.870966174681402e-05,
+      "loss": 0.3909,
+      "step": 19131
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8696935976646045e-05,
+      "loss": 0.5448,
+      "step": 19132
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.86842119888296e-05,
+      "loss": 0.2771,
+      "step": 19133
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.867148978356843e-05,
+      "loss": 0.0014,
+      "step": 19134
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.865876936106626e-05,
+      "loss": 0.0007,
+      "step": 19135
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8646050721526855e-05,
+      "loss": 0.7453,
+      "step": 19136
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8633333865153745e-05,
+      "loss": 0.735,
+      "step": 19137
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.862061879215069e-05,
+      "loss": 0.1095,
+      "step": 19138
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8607905502721326e-05,
+      "loss": 0.3468,
+      "step": 19139
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8595193997069066e-05,
+      "loss": 0.2436,
+      "step": 19140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.858248427539764e-05,
+      "loss": 0.1591,
+      "step": 19141
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8569776337910566e-05,
+      "loss": 0.2811,
+      "step": 19142
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8557070184811186e-05,
+      "loss": 0.0003,
+      "step": 19143
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8544365816303175e-05,
+      "loss": 1.2906,
+      "step": 19144
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.853166323258982e-05,
+      "loss": 0.5748,
+      "step": 19145
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.851896243387455e-05,
+      "loss": 0.6113,
+      "step": 19146
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.850626342036088e-05,
+      "loss": 0.3261,
+      "step": 19147
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8493566192252024e-05,
+      "loss": 0.3863,
+      "step": 19148
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.848087074975136e-05,
+      "loss": 0.7978,
+      "step": 19149
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.846817709306218e-05,
+      "loss": 0.2998,
+      "step": 19150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8455485222387745e-05,
+      "loss": 0.0304,
+      "step": 19151
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.844279513793132e-05,
+      "loss": 0.7851,
+      "step": 19152
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8430106839896125e-05,
+      "loss": 0.0035,
+      "step": 19153
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.841742032848531e-05,
+      "loss": 0.2991,
+      "step": 19154
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.840473560390203e-05,
+      "loss": 0.2701,
+      "step": 19155
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.839205266634946e-05,
+      "loss": 0.3124,
+      "step": 19156
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.837937151603065e-05,
+      "loss": 0.292,
+      "step": 19157
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8366692153148684e-05,
+      "loss": 0.4622,
+      "step": 19158
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.835401457790661e-05,
+      "loss": 0.0054,
+      "step": 19159
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8341338790507426e-05,
+      "loss": 0.9625,
+      "step": 19160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8328664791154136e-05,
+      "loss": 0.421,
+      "step": 19161
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8315992580049675e-05,
+      "loss": 0.3705,
+      "step": 19162
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8303322157397034e-05,
+      "loss": 0.4863,
+      "step": 19163
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8290653523399006e-05,
+      "loss": 0.3823,
+      "step": 19164
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.827798667825847e-05,
+      "loss": 0.5217,
+      "step": 19165
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.826532162217839e-05,
+      "loss": 0.8624,
+      "step": 19166
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.825265835536146e-05,
+      "loss": 0.2715,
+      "step": 19167
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.823999687801043e-05,
+      "loss": 0.2639,
+      "step": 19168
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.822733719032822e-05,
+      "loss": 0.0035,
+      "step": 19169
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8214679292517415e-05,
+      "loss": 0.6462,
+      "step": 19170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.820202318478071e-05,
+      "loss": 0.0016,
+      "step": 19171
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.818936886732092e-05,
+      "loss": 0.9883,
+      "step": 19172
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.817671634034052e-05,
+      "loss": 0.2603,
+      "step": 19173
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.81640656040422e-05,
+      "loss": 0.0234,
+      "step": 19174
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.815141665862851e-05,
+      "loss": 1.5002,
+      "step": 19175
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.813876950430204e-05,
+      "loss": 0.3149,
+      "step": 19176
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.812612414126529e-05,
+      "loss": 0.1826,
+      "step": 19177
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.811348056972075e-05,
+      "loss": 0.0043,
+      "step": 19178
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8100838789870905e-05,
+      "loss": 0.564,
+      "step": 19179
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.80881988019182e-05,
+      "loss": 0.8347,
+      "step": 19180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.807556060606504e-05,
+      "loss": 0.3892,
+      "step": 19181
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8062924202513784e-05,
+      "loss": 0.9058,
+      "step": 19182
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.805028959146681e-05,
+      "loss": 0.6546,
+      "step": 19183
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.803765677312643e-05,
+      "loss": 0.242,
+      "step": 19184
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.8025025747695e-05,
+      "loss": 0.6551,
+      "step": 19185
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.801239651537464e-05,
+      "loss": 0.5101,
+      "step": 19186
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.799976907636773e-05,
+      "loss": 0.199,
+      "step": 19187
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.798714343087646e-05,
+      "loss": 0.0231,
+      "step": 19188
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.79745195791029e-05,
+      "loss": 0.5344,
+      "step": 19189
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.796189752124932e-05,
+      "loss": 0.9015,
+      "step": 19190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.794927725751785e-05,
+      "loss": 0.2096,
+      "step": 19191
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.793665878811045e-05,
+      "loss": 0.0334,
+      "step": 19192
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.792404211322938e-05,
+      "loss": 0.1271,
+      "step": 19193
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.791142723307652e-05,
+      "loss": 0.2825,
+      "step": 19194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.789881414785389e-05,
+      "loss": 0.4784,
+      "step": 19195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.788620285776358e-05,
+      "loss": 0.304,
+      "step": 19196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.787359336300745e-05,
+      "loss": 0.0034,
+      "step": 19197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.78609856637874e-05,
+      "loss": 0.3432,
+      "step": 19198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7848379760305445e-05,
+      "loss": 0.0028,
+      "step": 19199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.783577565276333e-05,
+      "loss": 0.372,
+      "step": 19200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7823173341362945e-05,
+      "loss": 0.5635,
+      "step": 19201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.781057282630607e-05,
+      "loss": 0.3269,
+      "step": 19202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7797974107794506e-05,
+      "loss": 0.0012,
+      "step": 19203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7785377186029993e-05,
+      "loss": 1.4379,
+      "step": 19204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7772782061214254e-05,
+      "loss": 0.4238,
+      "step": 19205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7760188733548976e-05,
+      "loss": 0.5054,
+      "step": 19206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.774759720323583e-05,
+      "loss": 0.4818,
+      "step": 19207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7735007470476444e-05,
+      "loss": 0.3026,
+      "step": 19208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.772241953547243e-05,
+      "loss": 0.6344,
+      "step": 19209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.770983339842537e-05,
+      "loss": 1.5969,
+      "step": 19210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.76972490595368e-05,
+      "loss": 0.0033,
+      "step": 19211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.768466651900829e-05,
+      "loss": 0.0017,
+      "step": 19212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.767208577704119e-05,
+      "loss": 2.9375,
+      "step": 19213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7659506833837106e-05,
+      "loss": 0.0103,
+      "step": 19214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.764692968959746e-05,
+      "loss": 0.2365,
+      "step": 19215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7634354344523564e-05,
+      "loss": 0.4331,
+      "step": 19216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.762178079881681e-05,
+      "loss": 0.7858,
+      "step": 19217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.760920905267868e-05,
+      "loss": 3.3242,
+      "step": 19218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.759663910631033e-05,
+      "loss": 0.2486,
+      "step": 19219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7584070959913074e-05,
+      "loss": 0.9383,
+      "step": 19220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7571504613688295e-05,
+      "loss": 0.7441,
+      "step": 19221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.75589400678371e-05,
+      "loss": 0.2157,
+      "step": 19222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7546377322560725e-05,
+      "loss": 0.6541,
+      "step": 19223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.753381637806035e-05,
+      "loss": 0.0031,
+      "step": 19224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7521257234537106e-05,
+      "loss": 0.2575,
+      "step": 19225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7508699892192136e-05,
+      "loss": 0.0022,
+      "step": 19226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.749614435122651e-05,
+      "loss": 0.1362,
+      "step": 19227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.748359061184129e-05,
+      "loss": 0.3762,
+      "step": 19228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7471038674237486e-05,
+      "loss": 0.3586,
+      "step": 19229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.745848853861613e-05,
+      "loss": 2.7501,
+      "step": 19230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.744594020517816e-05,
+      "loss": 0.206,
+      "step": 19231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7433393674124555e-05,
+      "loss": 0.8371,
+      "step": 19232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.742084894565621e-05,
+      "loss": 0.403,
+      "step": 19233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7408306019974055e-05,
+      "loss": 0.0108,
+      "step": 19234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.739576489727883e-05,
+      "loss": 0.2471,
+      "step": 19235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.738322557777144e-05,
+      "loss": 0.6192,
+      "step": 19236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.737068806165275e-05,
+      "loss": 0.5329,
+      "step": 19237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7358152349123365e-05,
+      "loss": 0.2494,
+      "step": 19238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7345618440384165e-05,
+      "loss": 0.4443,
+      "step": 19239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.733308633563586e-05,
+      "loss": 0.2103,
+      "step": 19240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7320556035079e-05,
+      "loss": 0.522,
+      "step": 19241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.73080275389144e-05,
+      "loss": 0.3055,
+      "step": 19242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7295500847342584e-05,
+      "loss": 0.2433,
+      "step": 19243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7282975960564134e-05,
+      "loss": 0.1864,
+      "step": 19244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.727045287877973e-05,
+      "loss": 0.0722,
+      "step": 19245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.725793160218981e-05,
+      "loss": 0.29,
+      "step": 19246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.724541213099487e-05,
+      "loss": 0.526,
+      "step": 19247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.723289446539551e-05,
+      "loss": 0.4202,
+      "step": 19248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7220378605592074e-05,
+      "loss": 0.8728,
+      "step": 19249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7207864551785005e-05,
+      "loss": 1.1719,
+      "step": 19250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.719535230417472e-05,
+      "loss": 0.228,
+      "step": 19251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7182841862961556e-05,
+      "loss": 1.7006,
+      "step": 19252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.717033322834588e-05,
+      "loss": 0.2768,
+      "step": 19253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7157826400527975e-05,
+      "loss": 0.9042,
+      "step": 19254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.714532137970813e-05,
+      "loss": 0.637,
+      "step": 19255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.71328181660866e-05,
+      "loss": 0.2947,
+      "step": 19256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7120316759863595e-05,
+      "loss": 0.1934,
+      "step": 19257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.710781716123929e-05,
+      "loss": 0.3624,
+      "step": 19258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.709531937041389e-05,
+      "loss": 0.3139,
+      "step": 19259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7082823387587484e-05,
+      "loss": 0.2399,
+      "step": 19260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7070329212960254e-05,
+      "loss": 1.312,
+      "step": 19261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7057836846732116e-05,
+      "loss": 0.0008,
+      "step": 19262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.704534628910329e-05,
+      "loss": 0.4452,
+      "step": 19263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.7032857540273736e-05,
+      "loss": 0.8145,
+      "step": 19264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.70203706004434e-05,
+      "loss": 1.5082,
+      "step": 19265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.700788546981222e-05,
+      "loss": 0.6538,
+      "step": 19266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.699540214858027e-05,
+      "loss": 0.1942,
+      "step": 19267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.698292063694731e-05,
+      "loss": 0.7925,
+      "step": 19268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6970440935113224e-05,
+      "loss": 0.0327,
+      "step": 19269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6957963043277976e-05,
+      "loss": 0.4278,
+      "step": 19270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.694548696164125e-05,
+      "loss": 0.7329,
+      "step": 19271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6933012690402894e-05,
+      "loss": 0.1697,
+      "step": 19272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6920540229762636e-05,
+      "loss": 1.6111,
+      "step": 19273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.690806957992021e-05,
+      "loss": 0.0049,
+      "step": 19274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.689560074107533e-05,
+      "loss": 0.491,
+      "step": 19275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6883133713427665e-05,
+      "loss": 0.2836,
+      "step": 19276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.687066849717685e-05,
+      "loss": 0.8523,
+      "step": 19277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6858205092522485e-05,
+      "loss": 0.4738,
+      "step": 19278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.684574349966417e-05,
+      "loss": 0.058,
+      "step": 19279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.683328371880146e-05,
+      "loss": 0.3442,
+      "step": 19280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.682082575013386e-05,
+      "loss": 0.5417,
+      "step": 19281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.680836959386088e-05,
+      "loss": 0.0473,
+      "step": 19282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6795915250182015e-05,
+      "loss": 0.8633,
+      "step": 19283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6783462719296605e-05,
+      "loss": 0.3937,
+      "step": 19284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6771012001404154e-05,
+      "loss": 0.6162,
+      "step": 19285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.675856309670408e-05,
+      "loss": 0.3782,
+      "step": 19286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.674611600539556e-05,
+      "loss": 2.4285,
+      "step": 19287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6733670727678065e-05,
+      "loss": 0.6212,
+      "step": 19288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.672122726375089e-05,
+      "loss": 0.251,
+      "step": 19289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6708785613813175e-05,
+      "loss": 2.8318,
+      "step": 19290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6696345778064305e-05,
+      "loss": 0.4131,
+      "step": 19291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.668390775670338e-05,
+      "loss": 0.0007,
+      "step": 19292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.667147154992955e-05,
+      "loss": 0.6917,
+      "step": 19293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6659037157942124e-05,
+      "loss": 1.5145,
+      "step": 19294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.664660458094006e-05,
+      "loss": 0.0154,
+      "step": 19295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.663417381912248e-05,
+      "loss": 0.1473,
+      "step": 19296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.662174487268853e-05,
+      "loss": 0.0003,
+      "step": 19297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.660931774183716e-05,
+      "loss": 0.1833,
+      "step": 19298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.659689242676734e-05,
+      "loss": 0.3509,
+      "step": 19299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.658446892767816e-05,
+      "loss": 0.1346,
+      "step": 19300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6572047244768476e-05,
+      "loss": 0.0027,
+      "step": 19301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6559627378237225e-05,
+      "loss": 0.2738,
+      "step": 19302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.654720932828327e-05,
+      "loss": 0.7576,
+      "step": 19303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.653479309510552e-05,
+      "loss": 0.001,
+      "step": 19304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6522378678902756e-05,
+      "loss": 0.3258,
+      "step": 19305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.650996607987378e-05,
+      "loss": 0.3405,
+      "step": 19306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.649755529821738e-05,
+      "loss": 0.1421,
+      "step": 19307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6485146334132285e-05,
+      "loss": 0.0097,
+      "step": 19308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.647273918781723e-05,
+      "loss": 0.5378,
+      "step": 19309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.64603338594709e-05,
+      "loss": 0.1439,
+      "step": 19310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6447930349291846e-05,
+      "loss": 0.4833,
+      "step": 19311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.643552865747881e-05,
+      "loss": 0.017,
+      "step": 19312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6423128784230384e-05,
+      "loss": 0.2361,
+      "step": 19313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6410730729745016e-05,
+      "loss": 0.4443,
+      "step": 19314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.639833449422136e-05,
+      "loss": 0.4757,
+      "step": 19315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.638594007785793e-05,
+      "loss": 0.6031,
+      "step": 19316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.6373547480853125e-05,
+      "loss": 0.521,
+      "step": 19317
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6361156703405365e-05,
+      "loss": 1.043,
+      "step": 19318
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6348767745713234e-05,
+      "loss": 0.0005,
+      "step": 19319
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6336380607974966e-05,
+      "loss": 0.1164,
+      "step": 19320
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6323995290389e-05,
+      "loss": 0.2941,
+      "step": 19321
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6311611793153625e-05,
+      "loss": 0.6592,
+      "step": 19322
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.629923011646717e-05,
+      "loss": 0.3023,
+      "step": 19323
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.62868502605279e-05,
+      "loss": 0.0022,
+      "step": 19324
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6274472225534065e-05,
+      "loss": 0.1628,
+      "step": 19325
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6262096011683886e-05,
+      "loss": 0.1504,
+      "step": 19326
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6249721619175545e-05,
+      "loss": 0.0047,
+      "step": 19327
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6237349048207186e-05,
+      "loss": 0.6162,
+      "step": 19328
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.622497829897696e-05,
+      "loss": 0.7145,
+      "step": 19329
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6212609371682934e-05,
+      "loss": 0.3247,
+      "step": 19330
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6200242266523214e-05,
+      "loss": 0.4998,
+      "step": 19331
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6187876983695866e-05,
+      "loss": 0.3973,
+      "step": 19332
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6175513523398783e-05,
+      "loss": 0.6112,
+      "step": 19333
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6163151885830043e-05,
+      "loss": 0.0359,
+      "step": 19334
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6150792071187635e-05,
+      "loss": 0.6787,
+      "step": 19335
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.613843407966934e-05,
+      "loss": 0.375,
+      "step": 19336
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.61260779114732e-05,
+      "loss": 1.2885,
+      "step": 19337
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6113723566797046e-05,
+      "loss": 0.1527,
+      "step": 19338
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.610137104583861e-05,
+      "loss": 0.1889,
+      "step": 19339
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6089020348795876e-05,
+      "loss": 0.3454,
+      "step": 19340
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.607667147586647e-05,
+      "loss": 0.0018,
+      "step": 19341
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.606432442724815e-05,
+      "loss": 0.3066,
+      "step": 19342
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.605197920313877e-05,
+      "loss": 0.3426,
+      "step": 19343
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.603963580373588e-05,
+      "loss": 0.2208,
+      "step": 19344
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6027294229237165e-05,
+      "loss": 0.586,
+      "step": 19345
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.601495447984035e-05,
+      "loss": 0.5679,
+      "step": 19346
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.6002616555742955e-05,
+      "loss": 0.5896,
+      "step": 19347
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.599028045714251e-05,
+      "loss": 0.2419,
+      "step": 19348
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5977946184236706e-05,
+      "loss": 1.3055,
+      "step": 19349
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.596561373722293e-05,
+      "loss": 0.0077,
+      "step": 19350
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.595328311629871e-05,
+      "loss": 2.7409,
+      "step": 19351
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5940954321661495e-05,
+      "loss": 0.1322,
+      "step": 19352
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.592862735350871e-05,
+      "loss": 0.2601,
+      "step": 19353
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.591630221203777e-05,
+      "loss": 0.9885,
+      "step": 19354
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5903978897446045e-05,
+      "loss": 0.4102,
+      "step": 19355
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5891657409930854e-05,
+      "loss": 0.0025,
+      "step": 19356
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.587933774968952e-05,
+      "loss": 0.3293,
+      "step": 19357
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5867019916919304e-05,
+      "loss": 0.0011,
+      "step": 19358
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.585470391181754e-05,
+      "loss": 0.2841,
+      "step": 19359
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5842389734581284e-05,
+      "loss": 0.3283,
+      "step": 19360
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.583007738540788e-05,
+      "loss": 0.2096,
+      "step": 19361
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5817766864494486e-05,
+      "loss": 0.001,
+      "step": 19362
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.580545817203811e-05,
+      "loss": 1.0508,
+      "step": 19363
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.579315130823598e-05,
+      "loss": 0.8122,
+      "step": 19364
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5780846273285166e-05,
+      "loss": 0.3174,
+      "step": 19365
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.576854306738259e-05,
+      "loss": 0.2773,
+      "step": 19366
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.575624169072541e-05,
+      "loss": 0.2856,
+      "step": 19367
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.574394214351062e-05,
+      "loss": 0.0597,
+      "step": 19368
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.573164442593506e-05,
+      "loss": 2.1059,
+      "step": 19369
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.571934853819572e-05,
+      "loss": 0.2714,
+      "step": 19370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5707054480489526e-05,
+      "loss": 0.3085,
+      "step": 19371
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.56947622530133e-05,
+      "loss": 1.1815,
+      "step": 19372
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.568247185596391e-05,
+      "loss": 0.7788,
+      "step": 19373
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.567018328953818e-05,
+      "loss": 0.3824,
+      "step": 19374
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5657896553932875e-05,
+      "loss": 0.222,
+      "step": 19375
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5645611649344754e-05,
+      "loss": 0.2631,
+      "step": 19376
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.563332857597054e-05,
+      "loss": 1.1202,
+      "step": 19377
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.562104733400692e-05,
+      "loss": 0.0007,
+      "step": 19378
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.560876792365058e-05,
+      "loss": 0.4011,
+      "step": 19379
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.559649034509814e-05,
+      "loss": 0.3278,
+      "step": 19380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.558421459854622e-05,
+      "loss": 0.3089,
+      "step": 19381
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.557194068419137e-05,
+      "loss": 0.4176,
+      "step": 19382
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.555966860223017e-05,
+      "loss": 0.7008,
+      "step": 19383
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.554739835285917e-05,
+      "loss": 0.2777,
+      "step": 19384
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.553512993627472e-05,
+      "loss": 0.5175,
+      "step": 19385
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5522863352673434e-05,
+      "loss": 0.2636,
+      "step": 19386
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5510598602251714e-05,
+      "loss": 0.0021,
+      "step": 19387
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5498335685205854e-05,
+      "loss": 0.0456,
+      "step": 19388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5486074601732403e-05,
+      "loss": 0.3829,
+      "step": 19389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.547381535202755e-05,
+      "loss": 0.7692,
+      "step": 19390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5461557936287616e-05,
+      "loss": 0.234,
+      "step": 19391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.544930235470902e-05,
+      "loss": 0.3926,
+      "step": 19392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.543704860748791e-05,
+      "loss": 0.1827,
+      "step": 19393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5424796694820465e-05,
+      "loss": 0.191,
+      "step": 19394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.541254661690303e-05,
+      "loss": 0.7357,
+      "step": 19395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5400298373931655e-05,
+      "loss": 0.2033,
+      "step": 19396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.538805196610247e-05,
+      "loss": 0.5768,
+      "step": 19397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.537580739361172e-05,
+      "loss": 0.6906,
+      "step": 19398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5363564656655336e-05,
+      "loss": 0.0067,
+      "step": 19399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.535132375542941e-05,
+      "loss": 2.6895,
+      "step": 19400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5339084690129985e-05,
+      "loss": 0.9685,
+      "step": 19401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.532684746095302e-05,
+      "loss": 0.0119,
+      "step": 19402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5314612068094505e-05,
+      "loss": 0.5862,
+      "step": 19403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.530237851175036e-05,
+      "loss": 0.2594,
+      "step": 19404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.529014679211648e-05,
+      "loss": 0.0038,
+      "step": 19405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5277916909388755e-05,
+      "loss": 0.5687,
+      "step": 19406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.526568886376301e-05,
+      "loss": 0.4191,
+      "step": 19407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.52534626554351e-05,
+      "loss": 0.5297,
+      "step": 19408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.524123828460071e-05,
+      "loss": 0.4871,
+      "step": 19409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.522901575145568e-05,
+      "loss": 0.5302,
+      "step": 19410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.521679505619577e-05,
+      "loss": 0.0007,
+      "step": 19411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.520457619901653e-05,
+      "loss": 0.0022,
+      "step": 19412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5192359180113763e-05,
+      "loss": 0.0548,
+      "step": 19413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5180143999683114e-05,
+      "loss": 0.8452,
+      "step": 19414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.516793065792004e-05,
+      "loss": 0.2928,
+      "step": 19415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5155719155020275e-05,
+      "loss": 0.6535,
+      "step": 19416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.514350949117934e-05,
+      "loss": 0.6906,
+      "step": 19417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5131301666592686e-05,
+      "loss": 2.3828,
+      "step": 19418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.511909568145584e-05,
+      "loss": 0.265,
+      "step": 19419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.510689153596427e-05,
+      "loss": 0.2857,
+      "step": 19420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.509468923031339e-05,
+      "loss": 0.053,
+      "step": 19421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.508248876469862e-05,
+      "loss": 0.0074,
+      "step": 19422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.507029013931534e-05,
+      "loss": 0.1921,
+      "step": 19423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.5058093354358866e-05,
+      "loss": 0.6612,
+      "step": 19424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.504589841002451e-05,
+      "loss": 0.0051,
+      "step": 19425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.503370530650759e-05,
+      "loss": 0.308,
+      "step": 19426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.502151404400332e-05,
+      "loss": 0.588,
+      "step": 19427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.500932462270697e-05,
+      "loss": 0.338,
+      "step": 19428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4997137042813685e-05,
+      "loss": 0.5086,
+      "step": 19429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.498495130451866e-05,
+      "loss": 0.6514,
+      "step": 19430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.497276740801704e-05,
+      "loss": 0.705,
+      "step": 19431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.496058535350391e-05,
+      "loss": 0.0126,
+      "step": 19432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.494840514117436e-05,
+      "loss": 1.2291,
+      "step": 19433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.493622677122342e-05,
+      "loss": 0.6562,
+      "step": 19434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4924050243846125e-05,
+      "loss": 0.3138,
+      "step": 19435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.491187555923752e-05,
+      "loss": 0.2144,
+      "step": 19436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.489970271759241e-05,
+      "loss": 0.6832,
+      "step": 19437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.488753171910589e-05,
+      "loss": 0.2221,
+      "step": 19438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.487536256397276e-05,
+      "loss": 1.0065,
+      "step": 19439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.486319525238788e-05,
+      "loss": 0.2584,
+      "step": 19440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4851029784546224e-05,
+      "loss": 0.5825,
+      "step": 19441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.483886616064245e-05,
+      "loss": 0.625,
+      "step": 19442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.482670438087138e-05,
+      "loss": 0.858,
+      "step": 19443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.481454444542786e-05,
+      "loss": 0.9994,
+      "step": 19444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4802386354506496e-05,
+      "loss": 1.4703,
+      "step": 19445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.479023010830199e-05,
+      "loss": 0.3537,
+      "step": 19446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.477807570700913e-05,
+      "loss": 0.1582,
+      "step": 19447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.476592315082242e-05,
+      "loss": 0.7611,
+      "step": 19448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.47537724399365e-05,
+      "loss": 0.0025,
+      "step": 19449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4741623574545974e-05,
+      "loss": 0.7932,
+      "step": 19450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.472947655484535e-05,
+      "loss": 0.0011,
+      "step": 19451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.471733138102917e-05,
+      "loss": 0.138,
+      "step": 19452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.470518805329191e-05,
+      "loss": 0.0029,
+      "step": 19453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.469304657182804e-05,
+      "loss": 0.0029,
+      "step": 19454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.468090693683196e-05,
+      "loss": 1.2485,
+      "step": 19455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.466876914849811e-05,
+      "loss": 0.0745,
+      "step": 19456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.465663320702085e-05,
+      "loss": 0.9928,
+      "step": 19457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4644499112594423e-05,
+      "loss": 0.2701,
+      "step": 19458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4632366865413275e-05,
+      "loss": 0.0767,
+      "step": 19459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.462023646567166e-05,
+      "loss": 0.2538,
+      "step": 19460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4608107913563723e-05,
+      "loss": 0.0076,
+      "step": 19461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.459598120928379e-05,
+      "loss": 0.6117,
+      "step": 19462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4583856353026065e-05,
+      "loss": 0.5615,
+      "step": 19463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.457173334498458e-05,
+      "loss": 0.6498,
+      "step": 19464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.455961218535359e-05,
+      "loss": 0.0349,
+      "step": 19465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.454749287432722e-05,
+      "loss": 0.5064,
+      "step": 19466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4535375412099384e-05,
+      "loss": 0.004,
+      "step": 19467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.45232597988643e-05,
+      "loss": 1.0407,
+      "step": 19468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4511146034815875e-05,
+      "loss": 0.4803,
+      "step": 19469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.449903412014813e-05,
+      "loss": 1.0116,
+      "step": 19470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4486924055054996e-05,
+      "loss": 0.1168,
+      "step": 19471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.447481583973041e-05,
+      "loss": 0.5024,
+      "step": 19472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.446270947436827e-05,
+      "loss": 0.4937,
+      "step": 19473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4450604959162447e-05,
+      "loss": 0.0009,
+      "step": 19474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.443850229430677e-05,
+      "loss": 0.3672,
+      "step": 19475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.442640147999504e-05,
+      "loss": 1.3757,
+      "step": 19476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4414302516421045e-05,
+      "loss": 0.236,
+      "step": 19477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.440220540377852e-05,
+      "loss": 0.5264,
+      "step": 19478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4390110142261184e-05,
+      "loss": 0.0058,
+      "step": 19479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.437801673206273e-05,
+      "loss": 0.233,
+      "step": 19480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.436592517337682e-05,
+      "loss": 0.0014,
+      "step": 19481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.435383546639708e-05,
+      "loss": 0.8295,
+      "step": 19482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.434174761131709e-05,
+      "loss": 0.1862,
+      "step": 19483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.432966160833045e-05,
+      "loss": 0.002,
+      "step": 19484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.431757745763071e-05,
+      "loss": 1.2735,
+      "step": 19485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4305495159411274e-05,
+      "loss": 0.0022,
+      "step": 19486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4293414713865786e-05,
+      "loss": 0.1275,
+      "step": 19487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.428133612118756e-05,
+      "loss": 2.3633,
+      "step": 19488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.426925938157002e-05,
+      "loss": 0.3504,
+      "step": 19489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.42571844952067e-05,
+      "loss": 0.8295,
+      "step": 19490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.424511146229082e-05,
+      "loss": 0.6304,
+      "step": 19491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4233040283015715e-05,
+      "loss": 0.2194,
+      "step": 19492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4220970957574814e-05,
+      "loss": 0.3557,
+      "step": 19493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.420890348616125e-05,
+      "loss": 0.4073,
+      "step": 19494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4196837868968294e-05,
+      "loss": 0.171,
+      "step": 19495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4184774106189255e-05,
+      "loss": 0.5856,
+      "step": 19496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4172712198017206e-05,
+      "loss": 0.4441,
+      "step": 19497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.416065214464533e-05,
+      "loss": 0.7401,
+      "step": 19498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4148593946266775e-05,
+      "loss": 2.5234,
+      "step": 19499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.413653760307461e-05,
+      "loss": 0.0018,
+      "step": 19500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.412448311526191e-05,
+      "loss": 0.3228,
+      "step": 19501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4112430483021715e-05,
+      "loss": 0.7401,
+      "step": 19502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.410037970654701e-05,
+      "loss": 0.028,
+      "step": 19503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.408833078603079e-05,
+      "loss": 0.1666,
+      "step": 19504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.407628372166599e-05,
+      "loss": 0.2639,
+      "step": 19505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.406423851364557e-05,
+      "loss": 0.4458,
+      "step": 19506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.40521951621623e-05,
+      "loss": 1.0397,
+      "step": 19507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.404015366740914e-05,
+      "loss": 0.473,
+      "step": 19508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.4028114029578945e-05,
+      "loss": 0.3427,
+      "step": 19509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.401607624886436e-05,
+      "loss": 0.7219,
+      "step": 19510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.40040403254583e-05,
+      "loss": 0.1769,
+      "step": 19511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.399200625955347e-05,
+      "loss": 0.8129,
+      "step": 19512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3979974051342494e-05,
+      "loss": 0.2179,
+      "step": 19513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.396794370101815e-05,
+      "loss": 0.0026,
+      "step": 19514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.395591520877307e-05,
+      "loss": 0.521,
+      "step": 19515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.394388857479978e-05,
+      "loss": 0.6934,
+      "step": 19516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3931863799291024e-05,
+      "loss": 0.0013,
+      "step": 19517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.391984088243923e-05,
+      "loss": 1.0403,
+      "step": 19518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.390781982443695e-05,
+      "loss": 0.6075,
+      "step": 19519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.389580062547676e-05,
+      "loss": 0.175,
+      "step": 19520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.388378328575104e-05,
+      "loss": 0.3339,
+      "step": 19521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.387176780545227e-05,
+      "loss": 0.1539,
+      "step": 19522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3859754184772856e-05,
+      "loss": 0.0091,
+      "step": 19523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.384774242390518e-05,
+      "loss": 0.2638,
+      "step": 19524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.383573252304158e-05,
+      "loss": 0.2131,
+      "step": 19525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.38237244823744e-05,
+      "loss": 0.5841,
+      "step": 19526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3811718302095896e-05,
+      "loss": 1.1012,
+      "step": 19527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.379971398239835e-05,
+      "loss": 0.0025,
+      "step": 19528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.378771152347401e-05,
+      "loss": 0.4826,
+      "step": 19529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3775710925515046e-05,
+      "loss": 0.142,
+      "step": 19530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.376371218871365e-05,
+      "loss": 1.089,
+      "step": 19531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3751715313261954e-05,
+      "loss": 0.0008,
+      "step": 19532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.373972029935209e-05,
+      "loss": 0.3778,
+      "step": 19533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.37277271471761e-05,
+      "loss": 0.0024,
+      "step": 19534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.371573585692606e-05,
+      "loss": 0.003,
+      "step": 19535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3703746428794044e-05,
+      "loss": 0.0102,
+      "step": 19536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.369175886297195e-05,
+      "loss": 0.0893,
+      "step": 19537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3679773159651726e-05,
+      "loss": 2.1108,
+      "step": 19538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.366778931902547e-05,
+      "loss": 0.2505,
+      "step": 19539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.365580734128493e-05,
+      "loss": 3.4289,
+      "step": 19540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3643827226621976e-05,
+      "loss": 0.0024,
+      "step": 19541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3631848975228595e-05,
+      "loss": 0.2762,
+      "step": 19542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.361987258729646e-05,
+      "loss": 0.2273,
+      "step": 19543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3607898063017356e-05,
+      "loss": 0.0084,
+      "step": 19544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.359592540258319e-05,
+      "loss": 0.1807,
+      "step": 19545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3583954606185515e-05,
+      "loss": 0.1954,
+      "step": 19546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.357198567401611e-05,
+      "loss": 0.0045,
+      "step": 19547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.356001860626661e-05,
+      "loss": 0.1244,
+      "step": 19548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.354805340312866e-05,
+      "loss": 0.0096,
+      "step": 19549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3536090064793865e-05,
+      "loss": 0.0023,
+      "step": 19550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3524128591453796e-05,
+      "loss": 0.8054,
+      "step": 19551
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3512168983300016e-05,
+      "loss": 0.4203,
+      "step": 19552
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.350021124052401e-05,
+      "loss": 0.3216,
+      "step": 19553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.34882553633173e-05,
+      "loss": 0.2286,
+      "step": 19554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3476301351871336e-05,
+      "loss": 0.0174,
+      "step": 19555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.346434920637747e-05,
+      "loss": 0.5004,
+      "step": 19556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3452398927027186e-05,
+      "loss": 0.2921,
+      "step": 19557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.344045051401186e-05,
+      "loss": 0.0015,
+      "step": 19558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.3428503967522695e-05,
+      "loss": 0.6028,
+      "step": 19559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.341655928775115e-05,
+      "loss": 0.0763,
+      "step": 19560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.340461647488848e-05,
+      "loss": 0.3518,
+      "step": 19561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.33926755291258e-05,
+      "loss": 0.4612,
+      "step": 19562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.338073645065446e-05,
+      "loss": 0.0169,
+      "step": 19563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.336879923966564e-05,
+      "loss": 0.6494,
+      "step": 19564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.33568638963504e-05,
+      "loss": 3.8834,
+      "step": 19565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.334493042089999e-05,
+      "loss": 0.8325,
+      "step": 19566
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.333299881350541e-05,
+      "loss": 0.2676,
+      "step": 19567
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.332106907435773e-05,
+      "loss": 0.1483,
+      "step": 19568
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3309141203648094e-05,
+      "loss": 0.576,
+      "step": 19569
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.329721520156737e-05,
+      "loss": 0.4559,
+      "step": 19570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3285291068306634e-05,
+      "loss": 0.6521,
+      "step": 19571
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.327336880405678e-05,
+      "loss": 0.3717,
+      "step": 19572
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.326144840900875e-05,
+      "loss": 0.2741,
+      "step": 19573
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.32495298833534e-05,
+      "loss": 0.4038,
+      "step": 19574
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.323761322728164e-05,
+      "loss": 0.3435,
+      "step": 19575
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.322569844098425e-05,
+      "loss": 0.0844,
+      "step": 19576
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.321378552465205e-05,
+      "loss": 0.4643,
+      "step": 19577
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.320187447847581e-05,
+      "loss": 0.4074,
+      "step": 19578
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3189965302646265e-05,
+      "loss": 0.0021,
+      "step": 19579
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.31780579973541e-05,
+      "loss": 0.1043,
+      "step": 19580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.316615256279003e-05,
+      "loss": 0.659,
+      "step": 19581
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3154248999144675e-05,
+      "loss": 0.0017,
+      "step": 19582
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3142347306608666e-05,
+      "loss": 1.8832,
+      "step": 19583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.313044748537257e-05,
+      "loss": 0.7045,
+      "step": 19584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.311854953562702e-05,
+      "loss": 0.0104,
+      "step": 19585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3106653457562384e-05,
+      "loss": 0.6253,
+      "step": 19586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.309475925136932e-05,
+      "loss": 0.2011,
+      "step": 19587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.308286691723828e-05,
+      "loss": 0.3296,
+      "step": 19588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.30709764553596e-05,
+      "loss": 1.3496,
+      "step": 19589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3059087865923714e-05,
+      "loss": 0.1786,
+      "step": 19590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.304720114912111e-05,
+      "loss": 1.884,
+      "step": 19591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3035316305142016e-05,
+      "loss": 0.3785,
+      "step": 19592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.3023433334176754e-05,
+      "loss": 0.6018,
+      "step": 19593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.301155223641574e-05,
+      "loss": 2.8519,
+      "step": 19594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.299967301204909e-05,
+      "loss": 0.4973,
+      "step": 19595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.29877956612671e-05,
+      "loss": 1.3647,
+      "step": 19596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2975920184259934e-05,
+      "loss": 0.2033,
+      "step": 19597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2964046581217777e-05,
+      "loss": 0.4166,
+      "step": 19598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2952174852330774e-05,
+      "loss": 0.79,
+      "step": 19599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.294030499778902e-05,
+      "loss": 0.2342,
+      "step": 19600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.292843701778262e-05,
+      "loss": 0.213,
+      "step": 19601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.291657091250158e-05,
+      "loss": 1.1767,
+      "step": 19602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.290470668213594e-05,
+      "loss": 0.572,
+      "step": 19603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.289284432687575e-05,
+      "loss": 0.6846,
+      "step": 19604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.28809838469108e-05,
+      "loss": 0.1756,
+      "step": 19605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.286912524243118e-05,
+      "loss": 0.1996,
+      "step": 19606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.285726851362675e-05,
+      "loss": 1.8219,
+      "step": 19607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2845413660687296e-05,
+      "loss": 0.0227,
+      "step": 19608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2833560683802753e-05,
+      "loss": 0.7485,
+      "step": 19609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.282170958316293e-05,
+      "loss": 0.0143,
+      "step": 19610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.280986035895748e-05,
+      "loss": 0.207,
+      "step": 19611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.279801301137629e-05,
+      "loss": 1.0009,
+      "step": 19612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.278616754060908e-05,
+      "loss": 1.0015,
+      "step": 19613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.277432394684538e-05,
+      "loss": 0.275,
+      "step": 19614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.276248223027504e-05,
+      "loss": 0.6845,
+      "step": 19615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.275064239108756e-05,
+      "loss": 0.2149,
+      "step": 19616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.273880442947253e-05,
+      "loss": 0.0026,
+      "step": 19617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.272696834561964e-05,
+      "loss": 1.128,
+      "step": 19618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.271513413971832e-05,
+      "loss": 0.1652,
+      "step": 19619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.270330181195806e-05,
+      "loss": 0.0096,
+      "step": 19620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.269147136252846e-05,
+      "loss": 0.3814,
+      "step": 19621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.267964279161885e-05,
+      "loss": 0.0145,
+      "step": 19622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2667816099418706e-05,
+      "loss": 0.654,
+      "step": 19623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.265599128611739e-05,
+      "loss": 0.1642,
+      "step": 19624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.264416835190424e-05,
+      "loss": 0.2665,
+      "step": 19625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.263234729696863e-05,
+      "loss": 0.0407,
+      "step": 19626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2620528121499836e-05,
+      "loss": 0.2902,
+      "step": 19627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.260871082568712e-05,
+      "loss": 0.3616,
+      "step": 19628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2596895409719716e-05,
+      "loss": 0.9494,
+      "step": 19629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.258508187378684e-05,
+      "loss": 0.2468,
+      "step": 19630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.257327021807766e-05,
+      "loss": 0.0039,
+      "step": 19631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.256146044278132e-05,
+      "loss": 0.0024,
+      "step": 19632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.254965254808693e-05,
+      "loss": 1.4186,
+      "step": 19633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2537846534183644e-05,
+      "loss": 0.0187,
+      "step": 19634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.252604240126037e-05,
+      "loss": 0.3674,
+      "step": 19635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.251424014950627e-05,
+      "loss": 0.8036,
+      "step": 19636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.250243977911034e-05,
+      "loss": 2.6097,
+      "step": 19637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.249064129026144e-05,
+      "loss": 0.2911,
+      "step": 19638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.247884468314853e-05,
+      "loss": 0.4893,
+      "step": 19639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.246704995796063e-05,
+      "loss": 1.0471,
+      "step": 19640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2455257114886504e-05,
+      "loss": 0.2729,
+      "step": 19641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2443466154114964e-05,
+      "loss": 0.1795,
+      "step": 19642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2431677075834975e-05,
+      "loss": 0.2986,
+      "step": 19643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.24198898802352e-05,
+      "loss": 1.0239,
+      "step": 19644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2408104567504414e-05,
+      "loss": 0.7581,
+      "step": 19645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.239632113783136e-05,
+      "loss": 0.0058,
+      "step": 19646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2384539591404714e-05,
+      "loss": 1.0208,
+      "step": 19647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.237275992841317e-05,
+      "loss": 0.8133,
+      "step": 19648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.236098214904532e-05,
+      "loss": 0.0015,
+      "step": 19649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.23492062534898e-05,
+      "loss": 0.2592,
+      "step": 19650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2337432241935175e-05,
+      "loss": 0.9779,
+      "step": 19651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.232566011456999e-05,
+      "loss": 0.2025,
+      "step": 19652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2313889871582746e-05,
+      "loss": 0.2689,
+      "step": 19653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2302121513161936e-05,
+      "loss": 0.5349,
+      "step": 19654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.229035503949599e-05,
+      "loss": 0.2366,
+      "step": 19655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.227859045077342e-05,
+      "loss": 0.5884,
+      "step": 19656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.226682774718246e-05,
+      "loss": 0.0018,
+      "step": 19657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2255066928911586e-05,
+      "loss": 2.2314,
+      "step": 19658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.224330799614916e-05,
+      "loss": 0.3563,
+      "step": 19659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2231550949083356e-05,
+      "loss": 0.2546,
+      "step": 19660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2219795787902545e-05,
+      "loss": 0.0015,
+      "step": 19661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.220804251279498e-05,
+      "loss": 0.0037,
+      "step": 19662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2196291123948754e-05,
+      "loss": 0.0004,
+      "step": 19663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2184541621552216e-05,
+      "loss": 0.1245,
+      "step": 19664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.217279400579338e-05,
+      "loss": 0.2605,
+      "step": 19665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2161048276860366e-05,
+      "loss": 0.0017,
+      "step": 19666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.214930443494142e-05,
+      "loss": 0.4458,
+      "step": 19667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.213756248022443e-05,
+      "loss": 0.7593,
+      "step": 19668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2125822412897455e-05,
+      "loss": 0.3539,
+      "step": 19669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2114084233148605e-05,
+      "loss": 1.958,
+      "step": 19670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.210234794116575e-05,
+      "loss": 0.0044,
+      "step": 19671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2090613537136805e-05,
+      "loss": 0.4488,
+      "step": 19672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2078881021249804e-05,
+      "loss": 0.6291,
+      "step": 19673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.206715039369251e-05,
+      "loss": 0.5077,
+      "step": 19674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.205542165465281e-05,
+      "loss": 1.2792,
+      "step": 19675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2043694804318515e-05,
+      "loss": 0.0066,
+      "step": 19676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2031969842877434e-05,
+      "loss": 0.1862,
+      "step": 19677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.2020246770517305e-05,
+      "loss": 0.7707,
+      "step": 19678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.200852558742585e-05,
+      "loss": 0.2705,
+      "step": 19679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.19968062937908e-05,
+      "loss": 1.4805,
+      "step": 19680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1985088889799777e-05,
+      "loss": 0.2502,
+      "step": 19681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.197337337564045e-05,
+      "loss": 0.8663,
+      "step": 19682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.196165975150046e-05,
+      "loss": 0.2875,
+      "step": 19683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.194994801756725e-05,
+      "loss": 0.0017,
+      "step": 19684
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.19382381740285e-05,
+      "loss": 0.1924,
+      "step": 19685
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.192653022107172e-05,
+      "loss": 0.6394,
+      "step": 19686
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.191482415888426e-05,
+      "loss": 0.3469,
+      "step": 19687
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1903119987653733e-05,
+      "loss": 0.49,
+      "step": 19688
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1891417707567527e-05,
+      "loss": 0.6216,
+      "step": 19689
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1879717318812975e-05,
+      "loss": 0.4214,
+      "step": 19690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1868018821577435e-05,
+      "loss": 1.2425,
+      "step": 19691
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.185632221604835e-05,
+      "loss": 0.3324,
+      "step": 19692
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.184462750241295e-05,
+      "loss": 0.5436,
+      "step": 19693
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.183293468085848e-05,
+      "loss": 0.354,
+      "step": 19694
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.182124375157222e-05,
+      "loss": 0.9358,
+      "step": 19695
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.180955471474139e-05,
+      "loss": 0.1866,
+      "step": 19696
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.179786757055317e-05,
+      "loss": 0.1741,
+      "step": 19697
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.178618231919469e-05,
+      "loss": 0.4594,
+      "step": 19698
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.177449896085309e-05,
+      "loss": 0.5752,
+      "step": 19699
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.176281749571546e-05,
+      "loss": 1.6484,
+      "step": 19700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1751137923968865e-05,
+      "loss": 1.246,
+      "step": 19701
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1739460245800315e-05,
+      "loss": 0.0031,
+      "step": 19702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.172778446139684e-05,
+      "loss": 2.1403,
+      "step": 19703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.171611057094538e-05,
+      "loss": 0.3636,
+      "step": 19704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.170443857463294e-05,
+      "loss": 0.3942,
+      "step": 19705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1692768472646304e-05,
+      "loss": 0.021,
+      "step": 19706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1681100265172443e-05,
+      "loss": 0.0023,
+      "step": 19707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.166943395239825e-05,
+      "loss": 0.0106,
+      "step": 19708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.16577695345104e-05,
+      "loss": 0.0978,
+      "step": 19709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1646107011695805e-05,
+      "loss": 0.006,
+      "step": 19710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.163444638414122e-05,
+      "loss": 0.1717,
+      "step": 19711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.162278765203325e-05,
+      "loss": 0.6253,
+      "step": 19712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.161113081555877e-05,
+      "loss": 0.0654,
+      "step": 19713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.159947587490431e-05,
+      "loss": 0.2169,
+      "step": 19714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1587822830256494e-05,
+      "loss": 0.088,
+      "step": 19715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1576171681802066e-05,
+      "loss": 0.3669,
+      "step": 19716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.156452242972749e-05,
+      "loss": 0.0011,
+      "step": 19717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.15528750742193e-05,
+      "loss": 0.3792,
+      "step": 19718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.154122961546412e-05,
+      "loss": 0.6772,
+      "step": 19719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.152958605364834e-05,
+      "loss": 0.3807,
+      "step": 19720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.151794438895839e-05,
+      "loss": 0.2709,
+      "step": 19721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.150630462158082e-05,
+      "loss": 0.2665,
+      "step": 19722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.149466675170192e-05,
+      "loss": 0.8104,
+      "step": 19723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1483030779508076e-05,
+      "loss": 0.6458,
+      "step": 19724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.147139670518564e-05,
+      "loss": 0.2625,
+      "step": 19725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1459764528920876e-05,
+      "loss": 0.2523,
+      "step": 19726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1448134250900084e-05,
+      "loss": 0.5381,
+      "step": 19727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.14365058713095e-05,
+      "loss": 0.9416,
+      "step": 19728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.142487939033535e-05,
+      "loss": 0.339,
+      "step": 19729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.141325480816379e-05,
+      "loss": 0.391,
+      "step": 19730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.140163212498097e-05,
+      "loss": 0.7541,
+      "step": 19731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1390011340973065e-05,
+      "loss": 0.008,
+      "step": 19732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1378392456326047e-05,
+      "loss": 1.1569,
+      "step": 19733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1366775471226097e-05,
+      "loss": 0.1898,
+      "step": 19734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1355160385859217e-05,
+      "loss": 0.0438,
+      "step": 19735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.134354720041131e-05,
+      "loss": 0.3883,
+      "step": 19736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.133193591506844e-05,
+      "loss": 0.2685,
+      "step": 19737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1320326530016586e-05,
+      "loss": 3.3967,
+      "step": 19738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.130871904544151e-05,
+      "loss": 0.8013,
+      "step": 19739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.129711346152918e-05,
+      "loss": 0.004,
+      "step": 19740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.128550977846549e-05,
+      "loss": 0.6006,
+      "step": 19741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.127390799643616e-05,
+      "loss": 0.3552,
+      "step": 19742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.126230811562699e-05,
+      "loss": 0.8413,
+      "step": 19743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.125071013622378e-05,
+      "loss": 0.5078,
+      "step": 19744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1239114058412214e-05,
+      "loss": 0.2091,
+      "step": 19745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1227519882377996e-05,
+      "loss": 0.263,
+      "step": 19746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.121592760830681e-05,
+      "loss": 0.0046,
+      "step": 19747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.120433723638425e-05,
+      "loss": 0.1974,
+      "step": 19748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.119274876679597e-05,
+      "loss": 0.1685,
+      "step": 19749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.11811621997275e-05,
+      "loss": 0.2795,
+      "step": 19750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.116957753536439e-05,
+      "loss": 0.3785,
+      "step": 19751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.115799477389217e-05,
+      "loss": 0.3469,
+      "step": 19752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1146413915496295e-05,
+      "loss": 0.2429,
+      "step": 19753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.113483496036224e-05,
+      "loss": 0.006,
+      "step": 19754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.11232579086754e-05,
+      "loss": 0.4053,
+      "step": 19755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.111168276062118e-05,
+      "loss": 0.5611,
+      "step": 19756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1100109516384965e-05,
+      "loss": 0.1096,
+      "step": 19757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.108853817615198e-05,
+      "loss": 0.0007,
+      "step": 19758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.107696874010764e-05,
+      "loss": 0.3695,
+      "step": 19759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1065401208437195e-05,
+      "loss": 0.6971,
+      "step": 19760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.105383558132579e-05,
+      "loss": 1.2906,
+      "step": 19761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.1042271858958764e-05,
+      "loss": 0.4887,
+      "step": 19762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.103071004152119e-05,
+      "loss": 0.3151,
+      "step": 19763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.101915012919819e-05,
+      "loss": 0.2701,
+      "step": 19764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.100759212217503e-05,
+      "loss": 0.1771,
+      "step": 19765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.099603602063666e-05,
+      "loss": 0.0077,
+      "step": 19766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.098448182476814e-05,
+      "loss": 0.6597,
+      "step": 19767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0972929534754596e-05,
+      "loss": 0.0019,
+      "step": 19768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.096137915078091e-05,
+      "loss": 0.021,
+      "step": 19769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.094983067303209e-05,
+      "loss": 0.0017,
+      "step": 19770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.093828410169306e-05,
+      "loss": 0.4212,
+      "step": 19771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.092673943694872e-05,
+      "loss": 0.6814,
+      "step": 19772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0915196678983936e-05,
+      "loss": 0.2391,
+      "step": 19773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.090365582798358e-05,
+      "loss": 0.2973,
+      "step": 19774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.089211688413241e-05,
+      "loss": 0.3597,
+      "step": 19775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.088057984761525e-05,
+      "loss": 1.7612,
+      "step": 19776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.086904471861683e-05,
+      "loss": 0.0051,
+      "step": 19777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0857511497321874e-05,
+      "loss": 0.0019,
+      "step": 19778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0845980183915045e-05,
+      "loss": 0.001,
+      "step": 19779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.083445077858104e-05,
+      "loss": 0.1438,
+      "step": 19780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.082292328150448e-05,
+      "loss": 0.3244,
+      "step": 19781
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0811397692869874e-05,
+      "loss": 0.9897,
+      "step": 19782
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0799874012861896e-05,
+      "loss": 0.6373,
+      "step": 19783
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.078835224166509e-05,
+      "loss": 0.0178,
+      "step": 19784
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0776832379463815e-05,
+      "loss": 0.2877,
+      "step": 19785
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.076531442644269e-05,
+      "loss": 0.0182,
+      "step": 19786
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.075379838278616e-05,
+      "loss": 0.0067,
+      "step": 19787
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.074228424867849e-05,
+      "loss": 0.2935,
+      "step": 19788
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.07307720243042e-05,
+      "loss": 0.8233,
+      "step": 19789
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.071926170984763e-05,
+      "loss": 0.0035,
+      "step": 19790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0707753305493047e-05,
+      "loss": 0.0036,
+      "step": 19791
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.069624681142475e-05,
+      "loss": 0.6942,
+      "step": 19792
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.068474222782701e-05,
+      "loss": 0.6965,
+      "step": 19793
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.067323955488406e-05,
+      "loss": 0.806,
+      "step": 19794
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0661738792780096e-05,
+      "loss": 0.6497,
+      "step": 19795
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.065023994169928e-05,
+      "loss": 0.3864,
+      "step": 19796
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0638743001825755e-05,
+      "loss": 0.5148,
+      "step": 19797
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0627247973343624e-05,
+      "loss": 0.4459,
+      "step": 19798
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0615754856436965e-05,
+      "loss": 0.7893,
+      "step": 19799
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.060426365128982e-05,
+      "loss": 0.7661,
+      "step": 19800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.059277435808619e-05,
+      "loss": 0.3154,
+      "step": 19801
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.058128697701009e-05,
+      "loss": 0.1968,
+      "step": 19802
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.056980150824546e-05,
+      "loss": 0.1087,
+      "step": 19803
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.055831795197621e-05,
+      "loss": 0.6174,
+      "step": 19804
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0546836308386247e-05,
+      "loss": 0.4095,
+      "step": 19805
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.053535657765941e-05,
+      "loss": 0.3292,
+      "step": 19806
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.052387875997957e-05,
+      "loss": 0.8078,
+      "step": 19807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.051240285553048e-05,
+      "loss": 0.281,
+      "step": 19808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.050092886449599e-05,
+      "loss": 0.2347,
+      "step": 19809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.048945678705969e-05,
+      "loss": 0.5728,
+      "step": 19810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0477986623405475e-05,
+      "loss": 0.1682,
+      "step": 19811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0466518373716882e-05,
+      "loss": 0.3445,
+      "step": 19812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.045505203817756e-05,
+      "loss": 0.6727,
+      "step": 19813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0443587616971242e-05,
+      "loss": 0.3075,
+      "step": 19814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.043212511028139e-05,
+      "loss": 0.7621,
+      "step": 19815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.0420664518291586e-05,
+      "loss": 0.5927,
+      "step": 19816
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0409205841185452e-05,
+      "loss": 1.0275,
+      "step": 19817
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0397749079146372e-05,
+      "loss": 0.3155,
+      "step": 19818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0386294232357832e-05,
+      "loss": 0.2674,
+      "step": 19819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0374841301003273e-05,
+      "loss": 0.2359,
+      "step": 19820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0363390285266108e-05,
+      "loss": 1.7601,
+      "step": 19821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0351941185329696e-05,
+      "loss": 0.5359,
+      "step": 19822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0340494001377376e-05,
+      "loss": 0.4982,
+      "step": 19823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.032904873359246e-05,
+      "loss": 0.3468,
+      "step": 19824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0317605382158223e-05,
+      "loss": 0.358,
+      "step": 19825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0306163947257934e-05,
+      "loss": 0.0259,
+      "step": 19826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.029472442907478e-05,
+      "loss": 0.4311,
+      "step": 19827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0283286827791964e-05,
+      "loss": 1.0468,
+      "step": 19828
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.027185114359264e-05,
+      "loss": 0.4285,
+      "step": 19829
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.026041737665998e-05,
+      "loss": 1.1252,
+      "step": 19830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.024898552717695e-05,
+      "loss": 0.1767,
+      "step": 19831
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0237555595326723e-05,
+      "loss": 0.5875,
+      "step": 19832
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0226127581292354e-05,
+      "loss": 0.3571,
+      "step": 19833
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0214701485256725e-05,
+      "loss": 0.3007,
+      "step": 19834
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0203277307402913e-05,
+      "loss": 0.2401,
+      "step": 19835
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0191855047913867e-05,
+      "loss": 0.7238,
+      "step": 19836
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0180434706972356e-05,
+      "loss": 0.3849,
+      "step": 19837
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.016901628476142e-05,
+      "loss": 0.1941,
+      "step": 19838
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.015759978146388e-05,
+      "loss": 0.1647,
+      "step": 19839
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.014618519726244e-05,
+      "loss": 0.2356,
+      "step": 19840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0134772532340067e-05,
+      "loss": 0.352,
+      "step": 19841
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.012336178687935e-05,
+      "loss": 1.0666,
+      "step": 19842
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0111952961063112e-05,
+      "loss": 0.3353,
+      "step": 19843
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0100546055074004e-05,
+      "loss": 0.3053,
+      "step": 19844
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0089141069094712e-05,
+      "loss": 0.5398,
+      "step": 19845
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.007773800330786e-05,
+      "loss": 0.1165,
+      "step": 19846
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0066336857896052e-05,
+      "loss": 0.0086,
+      "step": 19847
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.005493763304187e-05,
+      "loss": 0.2857,
+      "step": 19848
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0043540328927858e-05,
+      "loss": 0.4251,
+      "step": 19849
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0032144945736515e-05,
+      "loss": 0.8442,
+      "step": 19850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0020751483650312e-05,
+      "loss": 0.0014,
+      "step": 19851
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.0009359942851725e-05,
+      "loss": 0.1927,
+      "step": 19852
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.999797032352315e-05,
+      "loss": 1.0811,
+      "step": 19853
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9986582625846982e-05,
+      "loss": 0.3939,
+      "step": 19854
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9975196850005572e-05,
+      "loss": 0.2004,
+      "step": 19855
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.996381299618126e-05,
+      "loss": 0.0032,
+      "step": 19856
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.995243106455634e-05,
+      "loss": 0.4286,
+      "step": 19857
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9941051055313098e-05,
+      "loss": 0.2344,
+      "step": 19858
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.992967296863366e-05,
+      "loss": 0.7633,
+      "step": 19859
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9918296804700402e-05,
+      "loss": 0.4436,
+      "step": 19860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.990692256369536e-05,
+      "loss": 0.0005,
+      "step": 19861
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9895550245800676e-05,
+      "loss": 0.007,
+      "step": 19862
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.988417985119858e-05,
+      "loss": 0.893,
+      "step": 19863
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9872811380071032e-05,
+      "loss": 0.1735,
+      "step": 19864
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.98614448326001e-05,
+      "loss": 0.6314,
+      "step": 19865
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9850080208967893e-05,
+      "loss": 1.3673,
+      "step": 19866
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9838717509356293e-05,
+      "loss": 0.656,
+      "step": 19867
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9827356733947287e-05,
+      "loss": 0.1541,
+      "step": 19868
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.981599788292282e-05,
+      "loss": 2.2998,
+      "step": 19869
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9804640956464775e-05,
+      "loss": 0.0014,
+      "step": 19870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9793285954755023e-05,
+      "loss": 0.1361,
+      "step": 19871
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.978193287797539e-05,
+      "loss": 1.0014,
+      "step": 19872
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9770581726307673e-05,
+      "loss": 0.2941,
+      "step": 19873
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.975923249993365e-05,
+      "loss": 0.6492,
+      "step": 19874
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.974788519903507e-05,
+      "loss": 0.2267,
+      "step": 19875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9736539823793625e-05,
+      "loss": 0.3729,
+      "step": 19876
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.972519637439102e-05,
+      "loss": 1.3832,
+      "step": 19877
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9713854851008874e-05,
+      "loss": 0.5726,
+      "step": 19878
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9702515253828856e-05,
+      "loss": 0.3508,
+      "step": 19879
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9691177583032444e-05,
+      "loss": 1.0659,
+      "step": 19880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9679841838801304e-05,
+      "loss": 1.1494,
+      "step": 19881
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.966850802131696e-05,
+      "loss": 0.2456,
+      "step": 19882
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9657176130760785e-05,
+      "loss": 0.4176,
+      "step": 19883
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9645846167314363e-05,
+      "loss": 0.1994,
+      "step": 19884
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9634518131159137e-05,
+      "loss": 0.0016,
+      "step": 19885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.962319202247637e-05,
+      "loss": 0.374,
+      "step": 19886
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9611867841447578e-05,
+      "loss": 0.6153,
+      "step": 19887
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9600545588254077e-05,
+      "loss": 0.023,
+      "step": 19888
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9589225263077056e-05,
+      "loss": 0.1791,
+      "step": 19889
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9577906866097968e-05,
+      "loss": 0.2651,
+      "step": 19890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9566590397497934e-05,
+      "loss": 0.8762,
+      "step": 19891
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9555275857458167e-05,
+      "loss": 0.6278,
+      "step": 19892
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.954396324615998e-05,
+      "loss": 0.1776,
+      "step": 19893
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.95326525637844e-05,
+      "loss": 0.364,
+      "step": 19894
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9521343810512592e-05,
+      "loss": 0.011,
+      "step": 19895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9510036986525652e-05,
+      "loss": 1.4344,
+      "step": 19896
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9498732092004634e-05,
+      "loss": 0.3635,
+      "step": 19897
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9487429127130573e-05,
+      "loss": 0.858,
+      "step": 19898
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.947612809208448e-05,
+      "loss": 0.6943,
+      "step": 19899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.94648289870473e-05,
+      "loss": 0.2642,
+      "step": 19900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9453531812199998e-05,
+      "loss": 0.4692,
+      "step": 19901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9442236567723476e-05,
+      "loss": 0.0016,
+      "step": 19902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.943094325379859e-05,
+      "loss": 0.2143,
+      "step": 19903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9419651870606208e-05,
+      "loss": 0.8162,
+      "step": 19904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9408362418327123e-05,
+      "loss": 0.0748,
+      "step": 19905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.939707489714214e-05,
+      "loss": 0.8432,
+      "step": 19906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9385789307231998e-05,
+      "loss": 0.3763,
+      "step": 19907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9374505648777434e-05,
+      "loss": 0.5322,
+      "step": 19908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9363223921959157e-05,
+      "loss": 0.4398,
+      "step": 19909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9351944126957767e-05,
+      "loss": 0.7427,
+      "step": 19910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9340666263953873e-05,
+      "loss": 0.2466,
+      "step": 19911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9329390333128212e-05,
+      "loss": 1.6866,
+      "step": 19912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9318116334661223e-05,
+      "loss": 0.4413,
+      "step": 19913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.930684426873345e-05,
+      "loss": 0.3007,
+      "step": 19914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9295574135525505e-05,
+      "loss": 0.3557,
+      "step": 19915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9284305935217738e-05,
+      "loss": 0.2023,
+      "step": 19916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9273039667990666e-05,
+      "loss": 0.4524,
+      "step": 19917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.926177533402466e-05,
+      "loss": 0.581,
+      "step": 19918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9250512933500114e-05,
+      "loss": 0.0245,
+      "step": 19919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9239252466597394e-05,
+      "loss": 0.495,
+      "step": 19920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9227993933496813e-05,
+      "loss": 0.6687,
+      "step": 19921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9216737334378655e-05,
+      "loss": 0.148,
+      "step": 19922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9205482669423173e-05,
+      "loss": 0.4549,
+      "step": 19923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.91942299388106e-05,
+      "loss": 0.0149,
+      "step": 19924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.918297914272113e-05,
+      "loss": 0.2759,
+      "step": 19925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9171730281334925e-05,
+      "loss": 0.3483,
+      "step": 19926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9160483354832127e-05,
+      "loss": 1.0542,
+      "step": 19927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9149238363392868e-05,
+      "loss": 1.7409,
+      "step": 19928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.913799530719711e-05,
+      "loss": 0.351,
+      "step": 19929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.912675418642502e-05,
+      "loss": 0.1899,
+      "step": 19930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9115515001256577e-05,
+      "loss": 0.4483,
+      "step": 19931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9104277751871668e-05,
+      "loss": 2.138,
+      "step": 19932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9093042438450348e-05,
+      "loss": 1.3108,
+      "step": 19933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.908180906117254e-05,
+      "loss": 0.0031,
+      "step": 19934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9070577620217998e-05,
+      "loss": 0.6162,
+      "step": 19935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9059348115766716e-05,
+      "loss": 0.3664,
+      "step": 19936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9048120547998514e-05,
+      "loss": 0.2279,
+      "step": 19937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.903689491709305e-05,
+      "loss": 0.0014,
+      "step": 19938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.9025671223230247e-05,
+      "loss": 0.3715,
+      "step": 19939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.901444946658974e-05,
+      "loss": 0.7215,
+      "step": 19940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.900322964735119e-05,
+      "loss": 0.5855,
+      "step": 19941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8992011765694416e-05,
+      "loss": 0.2913,
+      "step": 19942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.898079582179893e-05,
+      "loss": 0.3314,
+      "step": 19943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8969581815844367e-05,
+      "loss": 0.2171,
+      "step": 19944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.895836974801032e-05,
+      "loss": 0.359,
+      "step": 19945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8947159618476306e-05,
+      "loss": 0.2475,
+      "step": 19946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8935951427421873e-05,
+      "loss": 0.0177,
+      "step": 19947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8924745175026488e-05,
+      "loss": 0.2003,
+      "step": 19948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8913540861469593e-05,
+      "loss": 0.4204,
+      "step": 19949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8902338486930608e-05,
+      "loss": 0.0011,
+      "step": 19950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.889113805158893e-05,
+      "loss": 0.0895,
+      "step": 19951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8879939555623928e-05,
+      "loss": 0.1604,
+      "step": 19952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8868742999214924e-05,
+      "loss": 0.0483,
+      "step": 19953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.885754838254119e-05,
+      "loss": 0.3252,
+      "step": 19954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.884635570578202e-05,
+      "loss": 0.5586,
+      "step": 19955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8835164969116636e-05,
+      "loss": 0.0023,
+      "step": 19956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8823976172724246e-05,
+      "loss": 0.6885,
+      "step": 19957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.881278931678405e-05,
+      "loss": 0.3751,
+      "step": 19958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.880160440147508e-05,
+      "loss": 0.0009,
+      "step": 19959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8790421426976558e-05,
+      "loss": 0.0029,
+      "step": 19960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8779240393467566e-05,
+      "loss": 0.2755,
+      "step": 19961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8768061301127087e-05,
+      "loss": 0.6611,
+      "step": 19962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.875688415013411e-05,
+      "loss": 0.0009,
+      "step": 19963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.874570894066775e-05,
+      "loss": 0.5961,
+      "step": 19964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8734535672906856e-05,
+      "loss": 0.6531,
+      "step": 19965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.872336434703038e-05,
+      "loss": 0.5743,
+      "step": 19966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.87121949632172e-05,
+      "loss": 0.3064,
+      "step": 19967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8701027521646202e-05,
+      "loss": 0.4978,
+      "step": 19968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8689862022496196e-05,
+      "loss": 0.5678,
+      "step": 19969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.867869846594599e-05,
+      "loss": 5.5248,
+      "step": 19970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.866753685217435e-05,
+      "loss": 0.6461,
+      "step": 19971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8656377181360006e-05,
+      "loss": 0.0017,
+      "step": 19972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8645219453681685e-05,
+      "loss": 0.3157,
+      "step": 19973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8634063669318035e-05,
+      "loss": 1.9306,
+      "step": 19974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.862290982844772e-05,
+      "loss": 0.0019,
+      "step": 19975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.861175793124933e-05,
+      "loss": 0.0005,
+      "step": 19976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8600607977901517e-05,
+      "loss": 2.6134,
+      "step": 19977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.858945996858268e-05,
+      "loss": 3.7516,
+      "step": 19978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8578313903471467e-05,
+      "loss": 0.0863,
+      "step": 19979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8567169782746374e-05,
+      "loss": 0.2401,
+      "step": 19980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8556027606585724e-05,
+      "loss": 0.7393,
+      "step": 19981
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.854488737516809e-05,
+      "loss": 0.0121,
+      "step": 19982
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8533749088671832e-05,
+      "loss": 0.3182,
+      "step": 19983
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8522612747275204e-05,
+      "loss": 0.9817,
+      "step": 19984
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8511478351156664e-05,
+      "loss": 0.0096,
+      "step": 19985
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8500345900494504e-05,
+      "loss": 0.3765,
+      "step": 19986
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8489215395466896e-05,
+      "loss": 0.1873,
+      "step": 19987
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.84780868362522e-05,
+      "loss": 0.5209,
+      "step": 19988
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.846696022302854e-05,
+      "loss": 0.262,
+      "step": 19989
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.845583555597407e-05,
+      "loss": 0.2254,
+      "step": 19990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.844471283526707e-05,
+      "loss": 3.9147,
+      "step": 19991
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.843359206108551e-05,
+      "loss": 0.0056,
+      "step": 19992
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8422473233607496e-05,
+      "loss": 2.2696,
+      "step": 19993
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.841135635301119e-05,
+      "loss": 0.7484,
+      "step": 19994
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8400241419474495e-05,
+      "loss": 0.7286,
+      "step": 19995
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.838912843317544e-05,
+      "loss": 0.4844,
+      "step": 19996
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.837801739429196e-05,
+      "loss": 0.0011,
+      "step": 19997
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.836690830300202e-05,
+      "loss": 0.8593,
+      "step": 19998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.835580115948351e-05,
+      "loss": 0.5803,
+      "step": 19999
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8344695963914264e-05,
+      "loss": 0.5017,
+      "step": 20000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.833359271647213e-05,
+      "loss": 0.3253,
+      "step": 20001
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.832249141733493e-05,
+      "loss": 0.4854,
+      "step": 20002
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8311392066680405e-05,
+      "loss": 0.3954,
+      "step": 20003
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.830029466468631e-05,
+      "loss": 0.3225,
+      "step": 20004
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8289199211530362e-05,
+      "loss": 1.7285,
+      "step": 20005
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8278105707390208e-05,
+      "loss": 0.2498,
+      "step": 20006
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8267014152443563e-05,
+      "loss": 0.445,
+      "step": 20007
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8255924546867897e-05,
+      "loss": 0.1246,
+      "step": 20008
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.824483689084094e-05,
+      "loss": 1.0984,
+      "step": 20009
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8233751184540225e-05,
+      "loss": 0.0524,
+      "step": 20010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8222667428143204e-05,
+      "loss": 0.0011,
+      "step": 20011
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.821158562182736e-05,
+      "loss": 0.6444,
+      "step": 20012
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8200505765770286e-05,
+      "loss": 1.0397,
+      "step": 20013
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.818942786014926e-05,
+      "loss": 0.0075,
+      "step": 20014
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.817835190514174e-05,
+      "loss": 1.3832,
+      "step": 20015
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8167277900925077e-05,
+      "loss": 0.0009,
+      "step": 20016
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.815620584767662e-05,
+      "loss": 0.4812,
+      "step": 20017
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8145135745573676e-05,
+      "loss": 0.0145,
+      "step": 20018
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8134067594793487e-05,
+      "loss": 0.4128,
+      "step": 20019
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.812300139551332e-05,
+      "loss": 0.3179,
+      "step": 20020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.811193714791038e-05,
+      "loss": 2.5179,
+      "step": 20021
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8100874852161816e-05,
+      "loss": 0.67,
+      "step": 20022
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.808981450844482e-05,
+      "loss": 0.4762,
+      "step": 20023
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.807875611693647e-05,
+      "loss": 0.8208,
+      "step": 20024
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.806769967781387e-05,
+      "loss": 0.3874,
+      "step": 20025
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8056645191254085e-05,
+      "loss": 0.5134,
+      "step": 20026
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.804559265743405e-05,
+      "loss": 0.1434,
+      "step": 20027
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8034542076530862e-05,
+      "loss": 0.2686,
+      "step": 20028
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.802349344872148e-05,
+      "loss": 0.3469,
+      "step": 20029
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.8012446774182705e-05,
+      "loss": 0.002,
+      "step": 20030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.800140205309157e-05,
+      "loss": 0.6211,
+      "step": 20031
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7990359285624914e-05,
+      "loss": 0.4835,
+      "step": 20032
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.797931847195946e-05,
+      "loss": 2.4228,
+      "step": 20033
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7968279612272156e-05,
+      "loss": 0.4236,
+      "step": 20034
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.795724270673972e-05,
+      "loss": 0.1384,
+      "step": 20035
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7946207755538813e-05,
+      "loss": 0.6753,
+      "step": 20036
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7935174758846296e-05,
+      "loss": 3.752,
+      "step": 20037
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7924143716838725e-05,
+      "loss": 0.1854,
+      "step": 20038
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.791311462969273e-05,
+      "loss": 0.1265,
+      "step": 20039
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7902087497585063e-05,
+      "loss": 0.505,
+      "step": 20040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7891062320692187e-05,
+      "loss": 0.0046,
+      "step": 20041
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7880039099190642e-05,
+      "loss": 0.728,
+      "step": 20042
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7869017833257075e-05,
+      "loss": 0.4321,
+      "step": 20043
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7857998523067854e-05,
+      "loss": 0.3643,
+      "step": 20044
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7846981168799474e-05,
+      "loss": 0.2859,
+      "step": 20045
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7835965770628366e-05,
+      "loss": 0.7803,
+      "step": 20046
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7824952328730933e-05,
+      "loss": 0.242,
+      "step": 20047
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7813940843283516e-05,
+      "loss": 0.6286,
+      "step": 20048
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.780293131446248e-05,
+      "loss": 0.3696,
+      "step": 20049
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7791923742444077e-05,
+      "loss": 0.3177,
+      "step": 20050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7780918127404637e-05,
+      "loss": 0.3481,
+      "step": 20051
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7769914469520356e-05,
+      "loss": 1.1159,
+      "step": 20052
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7758912768967452e-05,
+      "loss": 0.3915,
+      "step": 20053
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7747913025922103e-05,
+      "loss": 0.3217,
+      "step": 20054
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7736915240560437e-05,
+      "loss": 0.6499,
+      "step": 20055
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7725919413058646e-05,
+      "loss": 0.2075,
+      "step": 20056
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7714925543592653e-05,
+      "loss": 0.4865,
+      "step": 20057
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7703933632338655e-05,
+      "loss": 0.0049,
+      "step": 20058
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.769294367947264e-05,
+      "loss": 0.2311,
+      "step": 20059
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.768195568517052e-05,
+      "loss": 0.5933,
+      "step": 20060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7670969649608322e-05,
+      "loss": 0.0008,
+      "step": 20061
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.765998557296202e-05,
+      "loss": 0.3224,
+      "step": 20062
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.7649003455407387e-05,
+      "loss": 0.341,
+      "step": 20063
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.763802329712034e-05,
+      "loss": 0.7078,
+      "step": 20064
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.762704509827671e-05,
+      "loss": 1.0571,
+      "step": 20065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7616068859052305e-05,
+      "loss": 0.3348,
+      "step": 20066
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7605094579622872e-05,
+      "loss": 0.0009,
+      "step": 20067
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7594122260164158e-05,
+      "loss": 0.3474,
+      "step": 20068
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7583151900851892e-05,
+      "loss": 0.6278,
+      "step": 20069
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7572183501861712e-05,
+      "loss": 0.2772,
+      "step": 20070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.756121706336928e-05,
+      "loss": 0.1235,
+      "step": 20071
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7550252585550185e-05,
+      "loss": 0.1451,
+      "step": 20072
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7539290068580045e-05,
+      "loss": 0.4301,
+      "step": 20073
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7528329512634377e-05,
+      "loss": 0.0061,
+      "step": 20074
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7517370917888713e-05,
+      "loss": 1.8524,
+      "step": 20075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.750641428451853e-05,
+      "loss": 0.2274,
+      "step": 20076
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7495459612699282e-05,
+      "loss": 0.2059,
+      "step": 20077
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7484506902606434e-05,
+      "loss": 1.2197,
+      "step": 20078
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7473556154415262e-05,
+      "loss": 0.208,
+      "step": 20079
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7462607368301238e-05,
+      "loss": 0.528,
+      "step": 20080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7451660544439708e-05,
+      "loss": 0.668,
+      "step": 20081
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7440715683005814e-05,
+      "loss": 0.9606,
+      "step": 20082
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.742977278417497e-05,
+      "loss": 0.1197,
+      "step": 20083
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7418831848122412e-05,
+      "loss": 0.2666,
+      "step": 20084
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.740789287502321e-05,
+      "loss": 0.729,
+      "step": 20085
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7396955865052695e-05,
+      "loss": 0.5064,
+      "step": 20086
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7386020818385883e-05,
+      "loss": 0.9433,
+      "step": 20087
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7375087735197902e-05,
+      "loss": 0.2303,
+      "step": 20088
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.736415661566393e-05,
+      "loss": 0.4416,
+      "step": 20089
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7353227459958906e-05,
+      "loss": 0.1779,
+      "step": 20090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.734230026825784e-05,
+      "loss": 0.0024,
+      "step": 20091
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.733137504073581e-05,
+      "loss": 0.2769,
+      "step": 20092
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7320451777567694e-05,
+      "loss": 0.6663,
+      "step": 20093
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7309530478928416e-05,
+      "loss": 0.5497,
+      "step": 20094
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.729861114499287e-05,
+      "loss": 0.399,
+      "step": 20095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.728769377593592e-05,
+      "loss": 0.5983,
+      "step": 20096
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.727677837193239e-05,
+      "loss": 0.5287,
+      "step": 20097
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7265864933157073e-05,
+      "loss": 0.831,
+      "step": 20098
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7254953459784728e-05,
+      "loss": 0.0818,
+      "step": 20099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7244043951990104e-05,
+      "loss": 1.7503,
+      "step": 20100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.723313640994787e-05,
+      "loss": 1.5055,
+      "step": 20101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7222230833832708e-05,
+      "loss": 0.9833,
+      "step": 20102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7211327223819263e-05,
+      "loss": 0.3772,
+      "step": 20103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7200425580082125e-05,
+      "loss": 0.3446,
+      "step": 20104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.718952590279592e-05,
+      "loss": 0.2193,
+      "step": 20105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7178628192135066e-05,
+      "loss": 0.1817,
+      "step": 20106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7167732448274198e-05,
+      "loss": 0.328,
+      "step": 20107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7156838671387782e-05,
+      "loss": 0.3699,
+      "step": 20108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7145946861650157e-05,
+      "loss": 0.6888,
+      "step": 20109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7135057019235857e-05,
+      "loss": 0.0012,
+      "step": 20110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.712416914431925e-05,
+      "loss": 1.6558,
+      "step": 20111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7113283237074622e-05,
+      "loss": 0.8187,
+      "step": 20112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7102399297676347e-05,
+      "loss": 0.2465,
+      "step": 20113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7091517326298717e-05,
+      "loss": 1.0139,
+      "step": 20114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7080637323115962e-05,
+      "loss": 0.6371,
+      "step": 20115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.706975928830232e-05,
+      "loss": 0.734,
+      "step": 20116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7058883222031997e-05,
+      "loss": 0.1558,
+      "step": 20117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7048009124479143e-05,
+      "loss": 0.3229,
+      "step": 20118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7037136995817904e-05,
+      "loss": 0.5037,
+      "step": 20119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7026266836222377e-05,
+      "loss": 0.9434,
+      "step": 20120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.701539864586662e-05,
+      "loss": 0.0019,
+      "step": 20121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.7004532424924696e-05,
+      "loss": 0.0008,
+      "step": 20122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.699366817357058e-05,
+      "loss": 1.2003,
+      "step": 20123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.698280589197827e-05,
+      "loss": 0.2008,
+      "step": 20124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.697194558032169e-05,
+      "loss": 0.3017,
+      "step": 20125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.696108723877477e-05,
+      "loss": 0.1903,
+      "step": 20126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6950230867511385e-05,
+      "loss": 0.9061,
+      "step": 20127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6939376466705382e-05,
+      "loss": 0.0011,
+      "step": 20128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.692852403653057e-05,
+      "loss": 0.272,
+      "step": 20129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6917673577160777e-05,
+      "loss": 0.4199,
+      "step": 20130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6906825088769662e-05,
+      "loss": 0.2458,
+      "step": 20131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6895978571531024e-05,
+      "loss": 0.2019,
+      "step": 20132
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6885134025618587e-05,
+      "loss": 0.1948,
+      "step": 20133
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6874291451205865e-05,
+      "loss": 0.8765,
+      "step": 20134
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6863450848466684e-05,
+      "loss": 0.0171,
+      "step": 20135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.685261221757447e-05,
+      "loss": 0.4628,
+      "step": 20136
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.684177555870284e-05,
+      "loss": 0.273,
+      "step": 20137
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6830940872025397e-05,
+      "loss": 0.0033,
+      "step": 20138
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6820108157715542e-05,
+      "loss": 0.7135,
+      "step": 20139
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.680927741594677e-05,
+      "loss": 0.452,
+      "step": 20140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.67984486468926e-05,
+      "loss": 0.154,
+      "step": 20141
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6787621850726338e-05,
+      "loss": 0.4536,
+      "step": 20142
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6776797027621405e-05,
+      "loss": 0.4761,
+      "step": 20143
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.676597417775113e-05,
+      "loss": 0.0352,
+      "step": 20144
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.675515330128883e-05,
+      "loss": 2.8809,
+      "step": 20145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6744334398407775e-05,
+      "loss": 0.2001,
+      "step": 20146
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.673351746928123e-05,
+      "loss": 0.0015,
+      "step": 20147
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6722702514082422e-05,
+      "loss": 0.3169,
+      "step": 20148
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.67118895329845e-05,
+      "loss": 0.3952,
+      "step": 20149
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6701078526160652e-05,
+      "loss": 1.0183,
+      "step": 20150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.669026949378398e-05,
+      "loss": 0.5229,
+      "step": 20151
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.667946243602757e-05,
+      "loss": 0.8566,
+      "step": 20152
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6668657353064495e-05,
+      "loss": 0.3983,
+      "step": 20153
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6657854245067823e-05,
+      "loss": 0.0794,
+      "step": 20154
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.664705311221042e-05,
+      "loss": 2.6134,
+      "step": 20155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6636253954665382e-05,
+      "loss": 0.9419,
+      "step": 20156
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6625456772605637e-05,
+      "loss": 0.6722,
+      "step": 20157
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6614661566203942e-05,
+      "loss": 0.0014,
+      "step": 20158
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6603868335633333e-05,
+      "loss": 0.0008,
+      "step": 20159
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.65930770810666e-05,
+      "loss": 0.2458,
+      "step": 20160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6582287802676457e-05,
+      "loss": 0.0791,
+      "step": 20161
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.657150050063583e-05,
+      "loss": 0.5017,
+      "step": 20162
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.656071517511735e-05,
+      "loss": 0.25,
+      "step": 20163
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.654993182629376e-05,
+      "loss": 0.3201,
+      "step": 20164
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.653915045433772e-05,
+      "loss": 1.2649,
+      "step": 20165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6528371059421915e-05,
+      "loss": 0.5916,
+      "step": 20166
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6517593641718936e-05,
+      "loss": 0.0046,
+      "step": 20167
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.650681820140138e-05,
+      "loss": 0.5691,
+      "step": 20168
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6496044738641792e-05,
+      "loss": 0.5265,
+      "step": 20169
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.648527325361268e-05,
+      "loss": 0.3989,
+      "step": 20170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6474503746486566e-05,
+      "loss": 3.2798,
+      "step": 20171
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6463736217435865e-05,
+      "loss": 0.1492,
+      "step": 20172
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.645297066663305e-05,
+      "loss": 0.1833,
+      "step": 20173
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.644220709425048e-05,
+      "loss": 3.6643,
+      "step": 20174
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.643144550046053e-05,
+      "loss": 2.7852,
+      "step": 20175
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6420685885435532e-05,
+      "loss": 0.0135,
+      "step": 20176
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6409928249347783e-05,
+      "loss": 0.1384,
+      "step": 20177
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.639917259236956e-05,
+      "loss": 0.0009,
+      "step": 20178
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.638841891467311e-05,
+      "loss": 0.4274,
+      "step": 20179
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6377667216430558e-05,
+      "loss": 0.6209,
+      "step": 20180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6366917497814155e-05,
+      "loss": 0.3693,
+      "step": 20181
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6356169758996058e-05,
+      "loss": 1.42,
+      "step": 20182
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6345424000148284e-05,
+      "loss": 0.0009,
+      "step": 20183
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6334680221443028e-05,
+      "loss": 0.265,
+      "step": 20184
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6323938423052233e-05,
+      "loss": 0.445,
+      "step": 20185
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6313198605147912e-05,
+      "loss": 0.6298,
+      "step": 20186
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6302460767902162e-05,
+      "loss": 0.001,
+      "step": 20187
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.629172491148683e-05,
+      "loss": 1.0747,
+      "step": 20188
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6280991036073822e-05,
+      "loss": 0.6768,
+      "step": 20189
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.627025914183514e-05,
+      "loss": 0.7983,
+      "step": 20190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6259529228942534e-05,
+      "loss": 0.6167,
+      "step": 20191
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6248801297567856e-05,
+      "loss": 0.0081,
+      "step": 20192
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6238075347882893e-05,
+      "loss": 0.0017,
+      "step": 20193
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.622735138005942e-05,
+      "loss": 0.2858,
+      "step": 20194
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6216629394269166e-05,
+      "loss": 0.5427,
+      "step": 20195
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.62059093906838e-05,
+      "loss": 0.5361,
+      "step": 20196
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6195191369475028e-05,
+      "loss": 0.0006,
+      "step": 20197
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6184475330814448e-05,
+      "loss": 1.0815,
+      "step": 20198
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6173761274873674e-05,
+      "loss": 0.529,
+      "step": 20199
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6163049201824283e-05,
+      "loss": 3.1904,
+      "step": 20200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.615233911183779e-05,
+      "loss": 0.3896,
+      "step": 20201
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.614163100508574e-05,
+      "loss": 0.4123,
+      "step": 20202
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6130924881739597e-05,
+      "loss": 0.5364,
+      "step": 20203
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6120220741970737e-05,
+      "loss": 0.3343,
+      "step": 20204
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6109518585950644e-05,
+      "loss": 0.305,
+      "step": 20205
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.609881841385073e-05,
+      "loss": 0.9379,
+      "step": 20206
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6088120225842206e-05,
+      "loss": 0.7734,
+      "step": 20207
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6077424022096506e-05,
+      "loss": 0.4234,
+      "step": 20208
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.606672980278492e-05,
+      "loss": 0.3453,
+      "step": 20209
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.605603756807858e-05,
+      "loss": 0.6309,
+      "step": 20210
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.604534731814887e-05,
+      "loss": 0.463,
+      "step": 20211
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6034659053166846e-05,
+      "loss": 2.328,
+      "step": 20212
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6023972773303676e-05,
+      "loss": 0.0009,
+      "step": 20213
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6013288478730605e-05,
+      "loss": 0.4534,
+      "step": 20214
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.6002606169618595e-05,
+      "loss": 0.7562,
+      "step": 20215
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5991925846138744e-05,
+      "loss": 0.4337,
+      "step": 20216
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.59812475084621e-05,
+      "loss": 0.2654,
+      "step": 20217
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5970571156759655e-05,
+      "loss": 0.3038,
+      "step": 20218
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.595989679120236e-05,
+      "loss": 0.1146,
+      "step": 20219
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.594922441196116e-05,
+      "loss": 0.5144,
+      "step": 20220
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.593855401920697e-05,
+      "loss": 0.9133,
+      "step": 20221
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.592788561311063e-05,
+      "loss": 0.3486,
+      "step": 20222
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5917219193843014e-05,
+      "loss": 0.5945,
+      "step": 20223
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5906554761574893e-05,
+      "loss": 0.0054,
+      "step": 20224
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5895892316477068e-05,
+      "loss": 0.2627,
+      "step": 20225
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5885231858720283e-05,
+      "loss": 0.0184,
+      "step": 20226
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5874573388475223e-05,
+      "loss": 0.1393,
+      "step": 20227
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5863916905912595e-05,
+      "loss": 1.6116,
+      "step": 20228
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5853262411203035e-05,
+      "loss": 0.1859,
+      "step": 20229
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5842609904517153e-05,
+      "loss": 0.0005,
+      "step": 20230
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.583195938602558e-05,
+      "loss": 0.4361,
+      "step": 20231
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5821310855898747e-05,
+      "loss": 0.5644,
+      "step": 20232
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5810664314307335e-05,
+      "loss": 0.0046,
+      "step": 20233
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.580001976142172e-05,
+      "loss": 0.3284,
+      "step": 20234
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5789377197412336e-05,
+      "loss": 0.4256,
+      "step": 20235
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5778736622449747e-05,
+      "loss": 0.0102,
+      "step": 20236
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.576809803670423e-05,
+      "loss": 0.7629,
+      "step": 20237
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.575746144034614e-05,
+      "loss": 0.2397,
+      "step": 20238
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5746826833545913e-05,
+      "loss": 0.172,
+      "step": 20239
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5736194216473737e-05,
+      "loss": 0.6257,
+      "step": 20240
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.572556358929991e-05,
+      "loss": 0.0149,
+      "step": 20241
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.571493495219469e-05,
+      "loss": 0.2786,
+      "step": 20242
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.570430830532824e-05,
+      "loss": 0.4644,
+      "step": 20243
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5693683648870766e-05,
+      "loss": 0.2769,
+      "step": 20244
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5683060982992376e-05,
+      "loss": 0.0042,
+      "step": 20245
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5672440307863203e-05,
+      "loss": 2.1567,
+      "step": 20246
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5661821623653306e-05,
+      "loss": 0.7048,
+      "step": 20247
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5651204930532737e-05,
+      "loss": 0.4858,
+      "step": 20248
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.564059022867149e-05,
+      "loss": 0.49,
+      "step": 20249
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.562997751823957e-05,
+      "loss": 0.3142,
+      "step": 20250
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.561936679940688e-05,
+      "loss": 0.0017,
+      "step": 20251
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.560875807234341e-05,
+      "loss": 1.1258,
+      "step": 20252
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5598151337218935e-05,
+      "loss": 0.0007,
+      "step": 20253
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.558754659420339e-05,
+      "loss": 0.677,
+      "step": 20254
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5576943843466603e-05,
+      "loss": 0.0078,
+      "step": 20255
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5566343085178254e-05,
+      "loss": 0.6521,
+      "step": 20256
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5555744319508207e-05,
+      "loss": 0.0015,
+      "step": 20257
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.55451475466262e-05,
+      "loss": 0.0012,
+      "step": 20258
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.553455276670179e-05,
+      "loss": 0.4745,
+      "step": 20259
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.552395997990479e-05,
+      "loss": 0.2316,
+      "step": 20260
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5513369186404735e-05,
+      "loss": 0.2112,
+      "step": 20261
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5502780386371207e-05,
+      "loss": 0.002,
+      "step": 20262
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5492193579973868e-05,
+      "loss": 0.475,
+      "step": 20263
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5481608767382165e-05,
+      "loss": 0.4016,
+      "step": 20264
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5471025948765618e-05,
+      "loss": 0.009,
+      "step": 20265
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5460445124293683e-05,
+      "loss": 1.1536,
+      "step": 20266
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.544986629413583e-05,
+      "loss": 2.3313,
+      "step": 20267
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5439289458461437e-05,
+      "loss": 0.2213,
+      "step": 20268
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5428714617439887e-05,
+      "loss": 0.0188,
+      "step": 20269
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5418141771240512e-05,
+      "loss": 0.9433,
+      "step": 20270
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5407570920032627e-05,
+      "loss": 0.3719,
+      "step": 20271
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5397002063985512e-05,
+      "loss": 0.4122,
+      "step": 20272
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5386435203268413e-05,
+      "loss": 0.5931,
+      "step": 20273
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5375870338050514e-05,
+      "loss": 0.7059,
+      "step": 20274
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.536530746850103e-05,
+      "loss": 0.2785,
+      "step": 20275
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5354746594789106e-05,
+      "loss": 0.0005,
+      "step": 20276
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5344187717083824e-05,
+      "loss": 0.3631,
+      "step": 20277
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.533363083555432e-05,
+      "loss": 0.0039,
+      "step": 20278
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5323075950369605e-05,
+      "loss": 0.2934,
+      "step": 20279
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5312523061698708e-05,
+      "loss": 0.4415,
+      "step": 20280
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5301972169710633e-05,
+      "loss": 0.0029,
+      "step": 20281
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5291423274574372e-05,
+      "loss": 0.4478,
+      "step": 20282
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5280876376458764e-05,
+      "loss": 0.0006,
+      "step": 20283
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.52703314755327e-05,
+      "loss": 0.2845,
+      "step": 20284
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5259788571965166e-05,
+      "loss": 0.0354,
+      "step": 20285
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5249247665924876e-05,
+      "loss": 1.6381,
+      "step": 20286
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.523870875758061e-05,
+      "loss": 0.0013,
+      "step": 20287
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.522817184710127e-05,
+      "loss": 0.9634,
+      "step": 20288
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.521763693465545e-05,
+      "loss": 0.2279,
+      "step": 20289
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5207104020411917e-05,
+      "loss": 0.3321,
+      "step": 20290
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.519657310453932e-05,
+      "loss": 0.0007,
+      "step": 20291
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.518604418720629e-05,
+      "loss": 0.0007,
+      "step": 20292
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.517551726858146e-05,
+      "loss": 0.3024,
+      "step": 20293
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5164992348833374e-05,
+      "loss": 0.8673,
+      "step": 20294
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5154469428130587e-05,
+      "loss": 0.0004,
+      "step": 20295
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.514394850664161e-05,
+      "loss": 0.8946,
+      "step": 20296
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5133429584534908e-05,
+      "loss": 0.8289,
+      "step": 20297
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5122912661978934e-05,
+      "loss": 0.7005,
+      "step": 20298
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5112397739142114e-05,
+      "loss": 0.2973,
+      "step": 20299
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5101884816192803e-05,
+      "loss": 0.0018,
+      "step": 20300
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5091373893299392e-05,
+      "loss": 0.2701,
+      "step": 20301
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5080864970630105e-05,
+      "loss": 1.0898,
+      "step": 20302
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5070358048353318e-05,
+      "loss": 0.4036,
+      "step": 20303
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.505985312663728e-05,
+      "loss": 0.0029,
+      "step": 20304
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5049350205650127e-05,
+      "loss": 0.5086,
+      "step": 20305
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5038849285560135e-05,
+      "loss": 0.007,
+      "step": 20306
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5028350366535475e-05,
+      "loss": 0.4119,
+      "step": 20307
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5017853448744145e-05,
+      "loss": 0.0018,
+      "step": 20308
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.5007358532354392e-05,
+      "loss": 0.0844,
+      "step": 20309
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4996865617534183e-05,
+      "loss": 0.0896,
+      "step": 20310
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4986374704451516e-05,
+      "loss": 0.3726,
+      "step": 20311
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.4975885793274526e-05,
+      "loss": 0.247,
+      "step": 20312
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.496539888417104e-05,
+      "loss": 0.1561,
+      "step": 20313
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.495491397730901e-05,
+      "loss": 0.5758,
+      "step": 20314
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4944431072856437e-05,
+      "loss": 0.1849,
+      "step": 20315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4933950170981083e-05,
+      "loss": 0.0154,
+      "step": 20316
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4923471271850813e-05,
+      "loss": 0.3305,
+      "step": 20317
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.491299437563345e-05,
+      "loss": 0.0006,
+      "step": 20318
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4902519482496753e-05,
+      "loss": 0.6751,
+      "step": 20319
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4892046592608455e-05,
+      "loss": 0.0029,
+      "step": 20320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4881575706136258e-05,
+      "loss": 0.3521,
+      "step": 20321
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.487110682324788e-05,
+      "loss": 2.627,
+      "step": 20322
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.486063994411091e-05,
+      "loss": 0.1832,
+      "step": 20323
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.485017506889299e-05,
+      "loss": 1.092,
+      "step": 20324
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4839712197761686e-05,
+      "loss": 0.4192,
+      "step": 20325
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.482925133088456e-05,
+      "loss": 0.0008,
+      "step": 20326
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.481879246842911e-05,
+      "loss": 0.0086,
+      "step": 20327
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.480833561056284e-05,
+      "loss": 0.9669,
+      "step": 20328
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4797880757453183e-05,
+      "loss": 0.2048,
+      "step": 20329
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4787427909267565e-05,
+      "loss": 0.2769,
+      "step": 20330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.477697706617341e-05,
+      "loss": 0.8423,
+      "step": 20331
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4766528228337996e-05,
+      "loss": 0.4877,
+      "step": 20332
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.475608139592864e-05,
+      "loss": 3.0041,
+      "step": 20333
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.474563656911274e-05,
+      "loss": 0.3863,
+      "step": 20334
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.473519374805747e-05,
+      "loss": 0.0096,
+      "step": 20335
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4724752932930036e-05,
+      "loss": 0.422,
+      "step": 20336
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4714314123897745e-05,
+      "loss": 0.5899,
+      "step": 20337
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.470387732112766e-05,
+      "loss": 0.4683,
+      "step": 20338
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4693442524786928e-05,
+      "loss": 0.0041,
+      "step": 20339
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4683009735042654e-05,
+      "loss": 0.0024,
+      "step": 20340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.46725789520619e-05,
+      "loss": 1.1953,
+      "step": 20341
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4662150176011704e-05,
+      "loss": 0.2079,
+      "step": 20342
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4651723407059077e-05,
+      "loss": 0.6388,
+      "step": 20343
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4641298645370972e-05,
+      "loss": 0.3448,
+      "step": 20344
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.463087589111432e-05,
+      "loss": 0.1775,
+      "step": 20345
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4620455144456058e-05,
+      "loss": 0.7568,
+      "step": 20346
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4610036405563033e-05,
+      "loss": 0.2319,
+      "step": 20347
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4599619674602077e-05,
+      "loss": 0.4167,
+      "step": 20348
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4589204951740025e-05,
+      "loss": 0.5707,
+      "step": 20349
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.457879223714368e-05,
+      "loss": 0.5958,
+      "step": 20350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.456838153097967e-05,
+      "loss": 0.0016,
+      "step": 20351
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.455797283341483e-05,
+      "loss": 0.876,
+      "step": 20352
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.454756614461581e-05,
+      "loss": 3.797,
+      "step": 20353
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4537161464749176e-05,
+      "loss": 0.5743,
+      "step": 20354
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.452675879398165e-05,
+      "loss": 0.8152,
+      "step": 20355
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.451635813247979e-05,
+      "loss": 0.1021,
+      "step": 20356
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4505959480410088e-05,
+      "loss": 0.3346,
+      "step": 20357
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4495562837939153e-05,
+      "loss": 0.5039,
+      "step": 20358
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.448516820523339e-05,
+      "loss": 0.0015,
+      "step": 20359
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4474775582459265e-05,
+      "loss": 0.191,
+      "step": 20360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.446438496978328e-05,
+      "loss": 0.5491,
+      "step": 20361
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.445399636737173e-05,
+      "loss": 0.7052,
+      "step": 20362
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4443609775390987e-05,
+      "loss": 0.4022,
+      "step": 20363
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4433225194007466e-05,
+      "loss": 0.3432,
+      "step": 20364
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.442284262338735e-05,
+      "loss": 0.0021,
+      "step": 20365
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4412462063696908e-05,
+      "loss": 0.2226,
+      "step": 20366
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.440208351510249e-05,
+      "loss": 0.5913,
+      "step": 20367
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.439170697777016e-05,
+      "loss": 0.3807,
+      "step": 20368
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4381332451866135e-05,
+      "loss": 0.3777,
+      "step": 20369
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4370959937556556e-05,
+      "loss": 0.6042,
+      "step": 20370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.43605894350075e-05,
+      "loss": 0.3405,
+      "step": 20371
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.435022094438505e-05,
+      "loss": 1.1347,
+      "step": 20372
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.433985446585523e-05,
+      "loss": 0.3148,
+      "step": 20373
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.432948999958405e-05,
+      "loss": 0.0011,
+      "step": 20374
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.43191275457375e-05,
+      "loss": 0.7502,
+      "step": 20375
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.430876710448149e-05,
+      "loss": 0.2859,
+      "step": 20376
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4298408675981928e-05,
+      "loss": 0.0024,
+      "step": 20377
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4288052260404712e-05,
+      "loss": 1.9708,
+      "step": 20378
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4277697857915672e-05,
+      "loss": 0.6201,
+      "step": 20379
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4267345468680667e-05,
+      "loss": 0.3878,
+      "step": 20380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4256995092865323e-05,
+      "loss": 0.2631,
+      "step": 20381
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4246646730635545e-05,
+      "loss": 0.7536,
+      "step": 20382
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.423630038215703e-05,
+      "loss": 0.1096,
+      "step": 20383
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.422595604759537e-05,
+      "loss": 0.3639,
+      "step": 20384
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.421561372711624e-05,
+      "loss": 0.3411,
+      "step": 20385
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4205273420885345e-05,
+      "loss": 0.8608,
+      "step": 20386
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.419493512906817e-05,
+      "loss": 1.7645,
+      "step": 20387
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.41845988518303e-05,
+      "loss": 0.1909,
+      "step": 20388
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.417426458933725e-05,
+      "loss": 0.0043,
+      "step": 20389
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4163932341754523e-05,
+      "loss": 0.0569,
+      "step": 20390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4153602109247566e-05,
+      "loss": 0.8382,
+      "step": 20391
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.41432738919818e-05,
+      "loss": 0.2649,
+      "step": 20392
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4132947690122627e-05,
+      "loss": 0.0045,
+      "step": 20393
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.412262350383538e-05,
+      "loss": 0.0007,
+      "step": 20394
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4112301333285427e-05,
+      "loss": 1.1094,
+      "step": 20395
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.410198117863802e-05,
+      "loss": 0.2493,
+      "step": 20396
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.409166304005844e-05,
+      "loss": 0.3474,
+      "step": 20397
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.408134691771191e-05,
+      "loss": 0.492,
+      "step": 20398
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4071032811763676e-05,
+      "loss": 0.1115,
+      "step": 20399
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4060720722378774e-05,
+      "loss": 0.5464,
+      "step": 20400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4050410649722474e-05,
+      "loss": 0.2264,
+      "step": 20401
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.404010259395984e-05,
+      "loss": 0.0032,
+      "step": 20402
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.4029796555255864e-05,
+      "loss": 0.3489,
+      "step": 20403
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.401949253377569e-05,
+      "loss": 0.49,
+      "step": 20404
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.400919052968429e-05,
+      "loss": 0.0005,
+      "step": 20405
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.399889054314655e-05,
+      "loss": 0.5121,
+      "step": 20406
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3988592574327547e-05,
+      "loss": 0.0016,
+      "step": 20407
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.397829662339209e-05,
+      "loss": 0.0002,
+      "step": 20408
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3968002690505044e-05,
+      "loss": 0.0042,
+      "step": 20409
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3957710775831344e-05,
+      "loss": 0.9055,
+      "step": 20410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3947420879535728e-05,
+      "loss": 0.3489,
+      "step": 20411
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3937133001782925e-05,
+      "loss": 0.2204,
+      "step": 20412
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3926847142737843e-05,
+      "loss": 0.8971,
+      "step": 20413
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3916563302565044e-05,
+      "loss": 3.1764,
+      "step": 20414
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3906281481429234e-05,
+      "loss": 0.7963,
+      "step": 20415
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.389600167949515e-05,
+      "loss": 0.0061,
+      "step": 20416
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.38857238969273e-05,
+      "loss": 0.0275,
+      "step": 20417
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.387544813389031e-05,
+      "loss": 0.0226,
+      "step": 20418
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3865174390548754e-05,
+      "loss": 0.3932,
+      "step": 20419
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3854902667067106e-05,
+      "loss": 0.105,
+      "step": 20420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3844632963609888e-05,
+      "loss": 0.418,
+      "step": 20421
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.383436528034154e-05,
+      "loss": 0.2702,
+      "step": 20422
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3824099617426474e-05,
+      "loss": 0.5353,
+      "step": 20423
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3813835975029093e-05,
+      "loss": 1.9238,
+      "step": 20424
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3803574353313752e-05,
+      "loss": 0.2484,
+      "step": 20425
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.379331475244477e-05,
+      "loss": 0.2183,
+      "step": 20426
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3783057172586444e-05,
+      "loss": 0.0037,
+      "step": 20427
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3772801613903032e-05,
+      "loss": 0.0006,
+      "step": 20428
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3762548076558795e-05,
+      "loss": 0.0106,
+      "step": 20429
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3752296560717842e-05,
+      "loss": 0.2895,
+      "step": 20430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3742047066544405e-05,
+      "loss": 0.0894,
+      "step": 20431
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3731799594202638e-05,
+      "loss": 1.5351,
+      "step": 20432
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3721554143856526e-05,
+      "loss": 0.2419,
+      "step": 20433
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3711310715670235e-05,
+      "loss": 0.0032,
+      "step": 20434
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3701069309807817e-05,
+      "loss": 0.8769,
+      "step": 20435
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3690829926433195e-05,
+      "loss": 0.326,
+      "step": 20436
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3680592565710354e-05,
+      "loss": 0.791,
+      "step": 20437
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3670357227803245e-05,
+      "loss": 0.2673,
+      "step": 20438
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3660123912875783e-05,
+      "loss": 2.1758,
+      "step": 20439
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3649892621091825e-05,
+      "loss": 0.4076,
+      "step": 20440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3639663352615207e-05,
+      "loss": 0.7948,
+      "step": 20441
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3629436107609728e-05,
+      "loss": 0.7024,
+      "step": 20442
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3619210886239192e-05,
+      "loss": 0.0008,
+      "step": 20443
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.360898768866732e-05,
+      "loss": 0.3465,
+      "step": 20444
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.359876651505781e-05,
+      "loss": 0.1346,
+      "step": 20445
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3588547365574366e-05,
+      "loss": 0.0127,
+      "step": 20446
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3578330240380628e-05,
+      "loss": 0.3917,
+      "step": 20447
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3568115139640174e-05,
+      "loss": 0.2822,
+      "step": 20448
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.355790206351663e-05,
+      "loss": 0.4849,
+      "step": 20449
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3547691012173514e-05,
+      "loss": 0.1455,
+      "step": 20450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3537481985774377e-05,
+      "loss": 0.1712,
+      "step": 20451
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3527274984482603e-05,
+      "loss": 0.1913,
+      "step": 20452
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3517070008461747e-05,
+      "loss": 0.6776,
+      "step": 20453
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3506867057875227e-05,
+      "loss": 1.1845,
+      "step": 20454
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3496666132886332e-05,
+      "loss": 0.3965,
+      "step": 20455
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.348646723365853e-05,
+      "loss": 0.1484,
+      "step": 20456
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3476270360355055e-05,
+      "loss": 0.2436,
+      "step": 20457
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3466075513139194e-05,
+      "loss": 0.3577,
+      "step": 20458
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.34558826921743e-05,
+      "loss": 0.2184,
+      "step": 20459
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.344569189762349e-05,
+      "loss": 0.3914,
+      "step": 20460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3435503129649957e-05,
+      "loss": 0.4711,
+      "step": 20461
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3425316388416964e-05,
+      "loss": 0.8577,
+      "step": 20462
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.341513167408755e-05,
+      "loss": 0.1493,
+      "step": 20463
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3404948986824766e-05,
+      "loss": 0.3168,
+      "step": 20464
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3394768326791817e-05,
+      "loss": 0.1859,
+      "step": 20465
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3384589694151608e-05,
+      "loss": 0.5826,
+      "step": 20466
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3374413089067168e-05,
+      "loss": 1.1519,
+      "step": 20467
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3364238511701456e-05,
+      "loss": 0.5465,
+      "step": 20468
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3354065962217422e-05,
+      "loss": 0.9145,
+      "step": 20469
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3343895440777953e-05,
+      "loss": 0.1299,
+      "step": 20470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3333726947545906e-05,
+      "loss": 0.5378,
+      "step": 20471
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.332356048268413e-05,
+      "loss": 0.1815,
+      "step": 20472
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3313396046355408e-05,
+      "loss": 0.0013,
+      "step": 20473
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3303233638722535e-05,
+      "loss": 0.6161,
+      "step": 20474
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3293073259948227e-05,
+      "loss": 1.3402,
+      "step": 20475
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3282914910195183e-05,
+      "loss": 0.354,
+      "step": 20476
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.327275858962608e-05,
+      "loss": 0.0013,
+      "step": 20477
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.32626042984036e-05,
+      "loss": 0.2056,
+      "step": 20478
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.325245203669021e-05,
+      "loss": 1.276,
+      "step": 20479
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3242301804648637e-05,
+      "loss": 0.0871,
+      "step": 20480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3232153602441402e-05,
+      "loss": 0.818,
+      "step": 20481
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.32220074302309e-05,
+      "loss": 0.2677,
+      "step": 20482
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.321186328817972e-05,
+      "loss": 0.0015,
+      "step": 20483
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3201721176450305e-05,
+      "loss": 0.4194,
+      "step": 20484
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3191581095204987e-05,
+      "loss": 0.0142,
+      "step": 20485
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3181443044606167e-05,
+      "loss": 2.3476,
+      "step": 20486
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3171307024816238e-05,
+      "loss": 2.4005,
+      "step": 20487
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3161173035997464e-05,
+      "loss": 0.0005,
+      "step": 20488
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3151041078312155e-05,
+      "loss": 0.2587,
+      "step": 20489
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.314091115192256e-05,
+      "loss": 0.0016,
+      "step": 20490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3130783256990866e-05,
+      "loss": 0.8064,
+      "step": 20491
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3120657393679277e-05,
+      "loss": 0.3702,
+      "step": 20492
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3110533562149948e-05,
+      "loss": 0.0029,
+      "step": 20493
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.310041176256498e-05,
+      "loss": 0.052,
+      "step": 20494
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.309029199508648e-05,
+      "loss": 1.372,
+      "step": 20495
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3080174259876483e-05,
+      "loss": 0.0656,
+      "step": 20496
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.307005855709701e-05,
+      "loss": 0.4169,
+      "step": 20497
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.305994488691005e-05,
+      "loss": 0.2555,
+      "step": 20498
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3049833249477584e-05,
+      "loss": 1.4926,
+      "step": 20499
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.303972364496149e-05,
+      "loss": 0.6689,
+      "step": 20500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3029616073523697e-05,
+      "loss": 0.3133,
+      "step": 20501
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3019510535326035e-05,
+      "loss": 0.001,
+      "step": 20502
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3009407030530387e-05,
+      "loss": 0.4363,
+      "step": 20503
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.299930555929842e-05,
+      "loss": 0.013,
+      "step": 20504
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.298920612179204e-05,
+      "loss": 0.8248,
+      "step": 20505
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.297910871817289e-05,
+      "loss": 0.3954,
+      "step": 20506
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2969013348602632e-05,
+      "loss": 0.0688,
+      "step": 20507
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2958920013243058e-05,
+      "loss": 0.4403,
+      "step": 20508
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2948828712255673e-05,
+      "loss": 0.2094,
+      "step": 20509
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2938739445802078e-05,
+      "loss": 0.1794,
+      "step": 20510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2928652214043964e-05,
+      "loss": 1.9221,
+      "step": 20511
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2918567017142748e-05,
+      "loss": 0.2309,
+      "step": 20512
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.29084838552599e-05,
+      "loss": 0.1796,
+      "step": 20513
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.289840272855703e-05,
+      "loss": 1.6563,
+      "step": 20514
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2888323637195455e-05,
+      "loss": 0.2802,
+      "step": 20515
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.287824658133661e-05,
+      "loss": 0.4326,
+      "step": 20516
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2868171561141858e-05,
+      "loss": 0.9322,
+      "step": 20517
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2858098576772542e-05,
+      "loss": 0.0094,
+      "step": 20518
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2848027628389958e-05,
+      "loss": 0.2695,
+      "step": 20519
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2837958716155392e-05,
+      "loss": 0.6106,
+      "step": 20520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.282789184023006e-05,
+      "loss": 0.0044,
+      "step": 20521
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2817827000775192e-05,
+      "loss": 0.938,
+      "step": 20522
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.280776419795195e-05,
+      "loss": 1.8309,
+      "step": 20523
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.279770343192148e-05,
+      "loss": 0.3051,
+      "step": 20524
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2787644702844875e-05,
+      "loss": 1.1247,
+      "step": 20525
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2777588010883236e-05,
+      "loss": 0.1747,
+      "step": 20526
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.276753335619761e-05,
+      "loss": 0.7008,
+      "step": 20527
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2757480738948937e-05,
+      "loss": 1.123,
+      "step": 20528
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2747430159298268e-05,
+      "loss": 0.1776,
+      "step": 20529
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.273738161740655e-05,
+      "loss": 0.7485,
+      "step": 20530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2727335113434626e-05,
+      "loss": 0.0889,
+      "step": 20531
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.271729064754344e-05,
+      "loss": 0.416,
+      "step": 20532
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.270724821989385e-05,
+      "loss": 0.2041,
+      "step": 20533
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.269720783064657e-05,
+      "loss": 3.0332,
+      "step": 20534
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2687169479962543e-05,
+      "loss": 0.4116,
+      "step": 20535
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2677133168002372e-05,
+      "loss": 0.1906,
+      "step": 20536
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2667098894926826e-05,
+      "loss": 0.2338,
+      "step": 20537
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.26570666608966e-05,
+      "loss": 0.0007,
+      "step": 20538
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2647036466072344e-05,
+      "loss": 0.0291,
+      "step": 20539
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.263700831061466e-05,
+      "loss": 0.4288,
+      "step": 20540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2626982194684146e-05,
+      "loss": 1.106,
+      "step": 20541
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2616958118441348e-05,
+      "loss": 1.5876,
+      "step": 20542
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.260693608204679e-05,
+      "loss": 0.5819,
+      "step": 20543
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.259691608566096e-05,
+      "loss": 0.3824,
+      "step": 20544
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2586898129444314e-05,
+      "loss": 0.0004,
+      "step": 20545
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2576882213557275e-05,
+      "loss": 0.703,
+      "step": 20546
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2566868338160243e-05,
+      "loss": 0.2851,
+      "step": 20547
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2556856503413545e-05,
+      "loss": 0.3063,
+      "step": 20548
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2546846709477532e-05,
+      "loss": 0.7601,
+      "step": 20549
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.25368389565125e-05,
+      "loss": 0.0008,
+      "step": 20550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2526833244678694e-05,
+      "loss": 0.0441,
+      "step": 20551
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2516829574136357e-05,
+      "loss": 0.3065,
+      "step": 20552
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2506827945045613e-05,
+      "loss": 0.9237,
+      "step": 20553
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2496828357566747e-05,
+      "loss": 0.2067,
+      "step": 20554
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.24868308118598e-05,
+      "loss": 0.3662,
+      "step": 20555
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.247683530808483e-05,
+      "loss": 0.2848,
+      "step": 20556
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2466841846402035e-05,
+      "loss": 0.4115,
+      "step": 20557
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.245685042697134e-05,
+      "loss": 0.0009,
+      "step": 20558
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2446861049952735e-05,
+      "loss": 0.3778,
+      "step": 20559
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2436873715506276e-05,
+      "loss": 0.4646,
+      "step": 20560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2426888423791818e-05,
+      "loss": 0.2387,
+      "step": 20561
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.241690517496923e-05,
+      "loss": 0.4827,
+      "step": 20562
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.240692396919852e-05,
+      "loss": 0.0441,
+      "step": 20563
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.239694480663938e-05,
+      "loss": 0.0045,
+      "step": 20564
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2386967687451673e-05,
+      "loss": 0.6623,
+      "step": 20565
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2376992611795148e-05,
+      "loss": 0.2957,
+      "step": 20566
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.236701957982957e-05,
+      "loss": 0.3426,
+      "step": 20567
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2357048591714616e-05,
+      "loss": 1.1768,
+      "step": 20568
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2347079647609957e-05,
+      "loss": 0.5373,
+      "step": 20569
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.233711274767523e-05,
+      "loss": 0.0021,
+      "step": 20570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.232714789207006e-05,
+      "loss": 0.6473,
+      "step": 20571
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2317185080953998e-05,
+      "loss": 0.0066,
+      "step": 20572
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2307224314486604e-05,
+      "loss": 0.1594,
+      "step": 20573
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2297265592827367e-05,
+      "loss": 0.3156,
+      "step": 20574
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.228730891613576e-05,
+      "loss": 0.2503,
+      "step": 20575
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2277354284571268e-05,
+      "loss": 0.8886,
+      "step": 20576
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2267401698293202e-05,
+      "loss": 0.466,
+      "step": 20577
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.225745115746103e-05,
+      "loss": 1.3222,
+      "step": 20578
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2247502662234094e-05,
+      "loss": 0.7239,
+      "step": 20579
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.22375562127716e-05,
+      "loss": 0.5958,
+      "step": 20580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2227611809232937e-05,
+      "loss": 0.422,
+      "step": 20581
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2217669451777343e-05,
+      "loss": 0.3724,
+      "step": 20582
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.220772914056394e-05,
+      "loss": 0.4484,
+      "step": 20583
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2197790875752018e-05,
+      "loss": 0.2858,
+      "step": 20584
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2187854657500654e-05,
+      "loss": 0.2275,
+      "step": 20585
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2177920485968932e-05,
+      "loss": 2.7864,
+      "step": 20586
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2167988361316058e-05,
+      "loss": 1.4441,
+      "step": 20587
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.215805828370096e-05,
+      "loss": 0.3033,
+      "step": 20588
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2148130253282686e-05,
+      "loss": 0.1908,
+      "step": 20589
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.213820427022025e-05,
+      "loss": 2.0339,
+      "step": 20590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2128280334672555e-05,
+      "loss": 0.0043,
+      "step": 20591
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2118358446798557e-05,
+      "loss": 0.712,
+      "step": 20592
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.210843860675711e-05,
+      "loss": 0.0041,
+      "step": 20593
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2098520814707094e-05,
+      "loss": 0.2178,
+      "step": 20594
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2088605070807324e-05,
+      "loss": 0.1629,
+      "step": 20595
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2078691375216562e-05,
+      "loss": 0.489,
+      "step": 20596
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.206877972809356e-05,
+      "loss": 0.2771,
+      "step": 20597
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.205887012959708e-05,
+      "loss": 0.4317,
+      "step": 20598
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.204896257988577e-05,
+      "loss": 0.5658,
+      "step": 20599
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2039057079118294e-05,
+      "loss": 0.3322,
+      "step": 20600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2029153627453273e-05,
+      "loss": 0.4553,
+      "step": 20601
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2019252225049312e-05,
+      "loss": 0.8862,
+      "step": 20602
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2009352872064985e-05,
+      "loss": 0.2006,
+      "step": 20603
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1999455568658746e-05,
+      "loss": 1.1459,
+      "step": 20604
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.198956031498909e-05,
+      "loss": 0.0054,
+      "step": 20605
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.197966711121457e-05,
+      "loss": 0.0707,
+      "step": 20606
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.196977595749353e-05,
+      "loss": 0.2089,
+      "step": 20607
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1959886853984338e-05,
+      "loss": 0.9613,
+      "step": 20608
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.194999980084547e-05,
+      "loss": 0.0532,
+      "step": 20609
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1940114798235147e-05,
+      "loss": 0.2256,
+      "step": 20610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1930231846311648e-05,
+      "loss": 0.4187,
+      "step": 20611
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1920350945233357e-05,
+      "loss": 0.5838,
+      "step": 20612
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.19104720951584e-05,
+      "loss": 0.3118,
+      "step": 20613
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1900595296244984e-05,
+      "loss": 0.688,
+      "step": 20614
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1890720548651282e-05,
+      "loss": 0.6726,
+      "step": 20615
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1880847852535433e-05,
+      "loss": 0.3069,
+      "step": 20616
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.187097720805553e-05,
+      "loss": 0.0013,
+      "step": 20617
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1861108615369628e-05,
+      "loss": 0.0027,
+      "step": 20618
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.185124207463575e-05,
+      "loss": 0.5548,
+      "step": 20619
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1841377586011922e-05,
+      "loss": 0.1467,
+      "step": 20620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1831515149656082e-05,
+      "loss": 0.2178,
+      "step": 20621
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.182165476572619e-05,
+      "loss": 1.1322,
+      "step": 20622
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1811796434380114e-05,
+      "loss": 0.1918,
+      "step": 20623
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1801940155775734e-05,
+      "loss": 0.3202,
+      "step": 20624
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1792085930070918e-05,
+      "loss": 0.1066,
+      "step": 20625
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1782233757423363e-05,
+      "loss": 0.1421,
+      "step": 20626
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1772383637990954e-05,
+      "loss": 0.481,
+      "step": 20627
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.17625355719314e-05,
+      "loss": 0.4177,
+      "step": 20628
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.175268955940231e-05,
+      "loss": 0.3401,
+      "step": 20629
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1742845600561472e-05,
+      "loss": 0.0011,
+      "step": 20630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1733003695566494e-05,
+      "loss": 0.3587,
+      "step": 20631
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.17231638445749e-05,
+      "loss": 0.7226,
+      "step": 20632
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1713326047744397e-05,
+      "loss": 0.3022,
+      "step": 20633
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.170349030523239e-05,
+      "loss": 0.574,
+      "step": 20634
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1693656617196427e-05,
+      "loss": 0.7257,
+      "step": 20635
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1683824983794056e-05,
+      "loss": 0.232,
+      "step": 20636
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.167399540518261e-05,
+      "loss": 0.6338,
+      "step": 20637
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.166416788151954e-05,
+      "loss": 0.9524,
+      "step": 20638
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.165434241296222e-05,
+      "loss": 1.6703,
+      "step": 20639
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.164451899966798e-05,
+      "loss": 0.7902,
+      "step": 20640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.163469764179414e-05,
+      "loss": 0.484,
+      "step": 20641
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1624878339497958e-05,
+      "loss": 0.3006,
+      "step": 20642
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1615061092936708e-05,
+      "loss": 0.4598,
+      "step": 20643
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.160524590226755e-05,
+      "loss": 0.5281,
+      "step": 20644
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1595432767647702e-05,
+      "loss": 0.6745,
+      "step": 20645
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1585621689234276e-05,
+      "loss": 0.8631,
+      "step": 20646
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1575812667184417e-05,
+      "loss": 1.2102,
+      "step": 20647
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1566005701655158e-05,
+      "loss": 0.1984,
+      "step": 20648
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.155620079280358e-05,
+      "loss": 0.3749,
+      "step": 20649
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.154639794078667e-05,
+      "loss": 0.5131,
+      "step": 20650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.153659714576142e-05,
+      "loss": 0.5438,
+      "step": 20651
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1526798407884806e-05,
+      "loss": 0.5151,
+      "step": 20652
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1517001727313633e-05,
+      "loss": 0.4088,
+      "step": 20653
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1507207104204875e-05,
+      "loss": 0.4315,
+      "step": 20654
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1497414538715397e-05,
+      "loss": 0.5946,
+      "step": 20655
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1487624031001948e-05,
+      "loss": 1.6939,
+      "step": 20656
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1477835581221274e-05,
+      "loss": 2.4219,
+      "step": 20657
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.146804918953027e-05,
+      "loss": 0.2569,
+      "step": 20658
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1458264856085516e-05,
+      "loss": 0.3038,
+      "step": 20659
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1448482581043698e-05,
+      "loss": 0.0895,
+      "step": 20660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1438702364561565e-05,
+      "loss": 0.6516,
+      "step": 20661
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.142892420679565e-05,
+      "loss": 0.9926,
+      "step": 20662
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.141914810790254e-05,
+      "loss": 0.2586,
+      "step": 20663
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.140937406803881e-05,
+      "loss": 1.0356,
+      "step": 20664
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.139960208736097e-05,
+      "loss": 0.3716,
+      "step": 20665
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.138983216602549e-05,
+      "loss": 0.0063,
+      "step": 20666
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.138006430418883e-05,
+      "loss": 0.2235,
+      "step": 20667
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.137029850200742e-05,
+      "loss": 0.0103,
+      "step": 20668
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1360534759637632e-05,
+      "loss": 0.0463,
+      "step": 20669
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1350773077235824e-05,
+      "loss": 0.4424,
+      "step": 20670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.13410134549583e-05,
+      "loss": 0.4025,
+      "step": 20671
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1331255892961358e-05,
+      "loss": 2.0823,
+      "step": 20672
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1321500391401235e-05,
+      "loss": 1.0266,
+      "step": 20673
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.131174695043421e-05,
+      "loss": 0.001,
+      "step": 20674
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1301995570216356e-05,
+      "loss": 0.0016,
+      "step": 20675
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1292246250903934e-05,
+      "loss": 0.4166,
+      "step": 20676
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1282498992653062e-05,
+      "loss": 0.3848,
+      "step": 20677
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1272753795619706e-05,
+      "loss": 0.4191,
+      "step": 20678
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1263010659960056e-05,
+      "loss": 0.5282,
+      "step": 20679
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.12532695858301e-05,
+      "loss": 0.0102,
+      "step": 20680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1243530573385753e-05,
+      "loss": 0.0015,
+      "step": 20681
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1233793622783086e-05,
+      "loss": 0.3503,
+      "step": 20682
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.122405873417794e-05,
+      "loss": 0.1995,
+      "step": 20683
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.121432590772619e-05,
+      "loss": 0.0029,
+      "step": 20684
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1204595143583804e-05,
+      "loss": 1.4416,
+      "step": 20685
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1194866441906477e-05,
+      "loss": 0.5938,
+      "step": 20686
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1185139802850043e-05,
+      "loss": 0.0088,
+      "step": 20687
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1175415226570324e-05,
+      "loss": 0.5871,
+      "step": 20688
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1165692713222966e-05,
+      "loss": 0.4008,
+      "step": 20689
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1155972262963688e-05,
+      "loss": 0.5043,
+      "step": 20690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1146253875948137e-05,
+      "loss": 1.3087,
+      "step": 20691
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1136537552331955e-05,
+      "loss": 0.0294,
+      "step": 20692
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1126823292270733e-05,
+      "loss": 0.5977,
+      "step": 20693
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1117111095920015e-05,
+      "loss": 0.2457,
+      "step": 20694
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.110740096343534e-05,
+      "loss": 0.4884,
+      "step": 20695
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.10976928949722e-05,
+      "loss": 0.5309,
+      "step": 20696
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1087986890686044e-05,
+      "loss": 0.5952,
+      "step": 20697
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1078282950732323e-05,
+      "loss": 0.0012,
+      "step": 20698
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.10685810752664e-05,
+      "loss": 0.2692,
+      "step": 20699
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1058881264443666e-05,
+      "loss": 0.5628,
+      "step": 20700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1049183518419467e-05,
+      "loss": 0.7336,
+      "step": 20701
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.103948783734899e-05,
+      "loss": 1.0242,
+      "step": 20702
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1029794221387595e-05,
+      "loss": 3.3343,
+      "step": 20703
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1020102670690548e-05,
+      "loss": 0.7386,
+      "step": 20704
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1010413185412934e-05,
+      "loss": 2.4727,
+      "step": 20705
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.1000725765709946e-05,
+      "loss": 1.0959,
+      "step": 20706
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.09910404117368e-05,
+      "loss": 0.4456,
+      "step": 20707
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0981357123648478e-05,
+      "loss": 0.4657,
+      "step": 20708
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0971675901600066e-05,
+      "loss": 0.2602,
+      "step": 20709
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.09619967457467e-05,
+      "loss": 0.7573,
+      "step": 20710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0952319656243243e-05,
+      "loss": 0.5243,
+      "step": 20711
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.094264463324472e-05,
+      "loss": 0.4348,
+      "step": 20712
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0932971676906056e-05,
+      "loss": 0.2947,
+      "step": 20713
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0923300787382163e-05,
+      "loss": 0.4089,
+      "step": 20714
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0913631964827865e-05,
+      "loss": 0.2139,
+      "step": 20715
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0903965209398036e-05,
+      "loss": 0.7594,
+      "step": 20716
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0894300521247458e-05,
+      "loss": 0.6491,
+      "step": 20717
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.088463790053089e-05,
+      "loss": 0.3529,
+      "step": 20718
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0874977347403082e-05,
+      "loss": 0.4582,
+      "step": 20719
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0865318862018702e-05,
+      "loss": 0.2586,
+      "step": 20720
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0855662444532462e-05,
+      "loss": 0.196,
+      "step": 20721
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.084600809509895e-05,
+      "loss": 0.5895,
+      "step": 20722
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0836355813872823e-05,
+      "loss": 0.3526,
+      "step": 20723
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0826705601008547e-05,
+      "loss": 0.4082,
+      "step": 20724
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0817057456660747e-05,
+      "loss": 0.7642,
+      "step": 20725
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0807411380983948e-05,
+      "loss": 0.1996,
+      "step": 20726
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0797767374132473e-05,
+      "loss": 0.104,
+      "step": 20727
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0788125436260887e-05,
+      "loss": 0.0208,
+      "step": 20728
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.07784855675236e-05,
+      "loss": 0.7208,
+      "step": 20729
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0768847768074852e-05,
+      "loss": 0.8184,
+      "step": 20730
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0759212038069122e-05,
+      "loss": 0.2133,
+      "step": 20731
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.074957837766062e-05,
+      "loss": 0.943,
+      "step": 20732
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.073994678700362e-05,
+      "loss": 0.3614,
+      "step": 20733
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0730317266252434e-05,
+      "loss": 0.2211,
+      "step": 20734
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.072068981556117e-05,
+      "loss": 0.2721,
+      "step": 20735
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0711064435084024e-05,
+      "loss": 0.3569,
+      "step": 20736
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0701441124975205e-05,
+      "loss": 0.3299,
+      "step": 20737
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0691819885388706e-05,
+      "loss": 0.3486,
+      "step": 20738
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0682200716478638e-05,
+      "loss": 0.1529,
+      "step": 20739
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0672583618399106e-05,
+      "loss": 0.2917,
+      "step": 20740
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0662968591304015e-05,
+      "loss": 0.4808,
+      "step": 20741
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.065335563534738e-05,
+      "loss": 0.5922,
+      "step": 20742
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0643744750683116e-05,
+      "loss": 0.006,
+      "step": 20743
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0634135937465168e-05,
+      "loss": 0.3285,
+      "step": 20744
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.062452919584736e-05,
+      "loss": 0.6385,
+      "step": 20745
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0614924525983574e-05,
+      "loss": 0.3896,
+      "step": 20746
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.060532192802758e-05,
+      "loss": 0.0687,
+      "step": 20747
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0595721402133158e-05,
+      "loss": 0.1445,
+      "step": 20748
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0586122948454053e-05,
+      "loss": 0.7647,
+      "step": 20749
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.057652656714399e-05,
+      "loss": 0.3911,
+      "step": 20750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0566932258356567e-05,
+      "loss": 0.217,
+      "step": 20751
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0557340022245504e-05,
+      "loss": 0.3502,
+      "step": 20752
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0547749858964403e-05,
+      "loss": 1.0367,
+      "step": 20753
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0538161768666732e-05,
+      "loss": 0.183,
+      "step": 20754
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0528575751506144e-05,
+      "loss": 0.5351,
+      "step": 20755
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0518991807636144e-05,
+      "loss": 0.2831,
+      "step": 20756
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0509409937210126e-05,
+      "loss": 0.5309,
+      "step": 20757
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0499830140381534e-05,
+      "loss": 0.4436,
+      "step": 20758
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.049025241730388e-05,
+      "loss": 0.0019,
+      "step": 20759
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0480676768130445e-05,
+      "loss": 0.4579,
+      "step": 20760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0471103193014583e-05,
+      "loss": 0.3266,
+      "step": 20761
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0461531692109607e-05,
+      "loss": 0.542,
+      "step": 20762
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.045196226556878e-05,
+      "loss": 0.4181,
+      "step": 20763
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0442394913545372e-05,
+      "loss": 0.2126,
+      "step": 20764
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0432829636192553e-05,
+      "loss": 0.2407,
+      "step": 20765
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0423266433663534e-05,
+      "loss": 0.5594,
+      "step": 20766
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0413705306111416e-05,
+      "loss": 0.0235,
+      "step": 20767
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0404146253689335e-05,
+      "loss": 0.3794,
+      "step": 20768
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0394589276550348e-05,
+      "loss": 0.2275,
+      "step": 20769
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.038503437484752e-05,
+      "loss": 0.221,
+      "step": 20770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0375481548733825e-05,
+      "loss": 0.2496,
+      "step": 20771
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.036593079836229e-05,
+      "loss": 0.2181,
+      "step": 20772
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0356382123885763e-05,
+      "loss": 0.2482,
+      "step": 20773
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0346835525457233e-05,
+      "loss": 0.002,
+      "step": 20774
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0337291003229582e-05,
+      "loss": 0.9951,
+      "step": 20775
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0327748557355557e-05,
+      "loss": 0.006,
+      "step": 20776
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.031820818798806e-05,
+      "loss": 0.3284,
+      "step": 20777
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.030866989527988e-05,
+      "loss": 0.2663,
+      "step": 20778
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0299133679383616e-05,
+      "loss": 1.0159,
+      "step": 20779
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0289599540452174e-05,
+      "loss": 0.0039,
+      "step": 20780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0280067478638073e-05,
+      "loss": 0.3876,
+      "step": 20781
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.027053749409398e-05,
+      "loss": 0.2988,
+      "step": 20782
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0261009586972588e-05,
+      "loss": 0.3834,
+      "step": 20783
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.025148375742639e-05,
+      "loss": 0.3751,
+      "step": 20784
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.024196000560793e-05,
+      "loss": 0.2159,
+      "step": 20785
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0232438331669808e-05,
+      "loss": 0.7761,
+      "step": 20786
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.022291873576438e-05,
+      "loss": 1.295,
+      "step": 20787
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0213401218044127e-05,
+      "loss": 0.3677,
+      "step": 20788
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0203885778661526e-05,
+      "loss": 0.3797,
+      "step": 20789
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.019437241776887e-05,
+      "loss": 0.8829,
+      "step": 20790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0184861135518526e-05,
+      "loss": 0.2563,
+      "step": 20791
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0175351932062816e-05,
+      "loss": 1.5991,
+      "step": 20792
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0165844807553988e-05,
+      "loss": 0.5511,
+      "step": 20793
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.015633976214432e-05,
+      "loss": 0.421,
+      "step": 20794
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.014683679598599e-05,
+      "loss": 0.3112,
+      "step": 20795
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.013733590923119e-05,
+      "loss": 0.4684,
+      "step": 20796
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0127837102032052e-05,
+      "loss": 0.388,
+      "step": 20797
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0118340374540703e-05,
+      "loss": 1.2461,
+      "step": 20798
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0108845726909224e-05,
+      "loss": 0.007,
+      "step": 20799
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0099353159289588e-05,
+      "loss": 0.0245,
+      "step": 20800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0089862671833894e-05,
+      "loss": 0.1247,
+      "step": 20801
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0080374264694104e-05,
+      "loss": 0.0012,
+      "step": 20802
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.007088793802208e-05,
+      "loss": 1.2815,
+      "step": 20803
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.006140369196983e-05,
+      "loss": 0.2271,
+      "step": 20804
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.00519215266892e-05,
+      "loss": 0.0185,
+      "step": 20805
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0042441442331974e-05,
+      "loss": 0.9544,
+      "step": 20806
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0032963439050043e-05,
+      "loss": 0.5618,
+      "step": 20807
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.002348751699517e-05,
+      "loss": 0.4569,
+      "step": 20808
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.0014013676319048e-05,
+      "loss": 0.2716,
+      "step": 20809
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.000454191717342e-05,
+      "loss": 0.5368,
+      "step": 20810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.999507223970996e-05,
+      "loss": 0.5448,
+      "step": 20811
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.99856046440803e-05,
+      "loss": 0.1177,
+      "step": 20812
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.9976139130436057e-05,
+      "loss": 0.6183,
+      "step": 20813
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9966675698928814e-05,
+      "loss": 0.6621,
+      "step": 20814
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.995721434971011e-05,
+      "loss": 0.0085,
+      "step": 20815
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9947755082931443e-05,
+      "loss": 1.0281,
+      "step": 20816
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9938297898744292e-05,
+      "loss": 0.0055,
+      "step": 20817
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9928842797300122e-05,
+      "loss": 1.5879,
+      "step": 20818
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9919389778750305e-05,
+      "loss": 0.1933,
+      "step": 20819
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.990993884324626e-05,
+      "loss": 0.01,
+      "step": 20820
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9900489990939295e-05,
+      "loss": 1.2267,
+      "step": 20821
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9891043221980722e-05,
+      "loss": 0.0009,
+      "step": 20822
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9881598536521844e-05,
+      "loss": 0.2633,
+      "step": 20823
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9872155934713894e-05,
+      "loss": 0.0018,
+      "step": 20824
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.986271541670803e-05,
+      "loss": 0.9497,
+      "step": 20825
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.985327698265549e-05,
+      "loss": 0.2774,
+      "step": 20826
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9843840632707425e-05,
+      "loss": 2.7852,
+      "step": 20827
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.983440636701486e-05,
+      "loss": 0.0025,
+      "step": 20828
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9824974185728992e-05,
+      "loss": 0.3599,
+      "step": 20829
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.981554408900075e-05,
+      "loss": 0.4871,
+      "step": 20830
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9806116076981167e-05,
+      "loss": 0.5161,
+      "step": 20831
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9796690149821316e-05,
+      "loss": 0.238,
+      "step": 20832
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9787266307672012e-05,
+      "loss": 0.3879,
+      "step": 20833
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9777844550684203e-05,
+      "loss": 0.6897,
+      "step": 20834
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.976842487900883e-05,
+      "loss": 0.4758,
+      "step": 20835
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9759007292796648e-05,
+      "loss": 3.0107,
+      "step": 20836
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9749591792198477e-05,
+      "loss": 0.7611,
+      "step": 20837
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9740178377365183e-05,
+      "loss": 0.1537,
+      "step": 20838
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.97307670484474e-05,
+      "loss": 0.6847,
+      "step": 20839
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9721357805595868e-05,
+      "loss": 0.0014,
+      "step": 20840
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9711950648961282e-05,
+      "loss": 0.0051,
+      "step": 20841
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.970254557869426e-05,
+      "loss": 1.9338,
+      "step": 20842
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9693142594945433e-05,
+      "loss": 0.3084,
+      "step": 20843
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9683741697865358e-05,
+      "loss": 0.8309,
+      "step": 20844
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9674342887604583e-05,
+      "loss": 0.3007,
+      "step": 20845
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9664946164313622e-05,
+      "loss": 0.5612,
+      "step": 20846
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9655551528142937e-05,
+      "loss": 0.0053,
+      "step": 20847
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.964615897924301e-05,
+      "loss": 0.7337,
+      "step": 20848
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9636768517764133e-05,
+      "loss": 0.2111,
+      "step": 20849
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.962738014385681e-05,
+      "loss": 0.0054,
+      "step": 20850
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.961799385767136e-05,
+      "loss": 0.6716,
+      "step": 20851
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9608609659358004e-05,
+      "loss": 0.3014,
+      "step": 20852
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9599227549067103e-05,
+      "loss": 0.3683,
+      "step": 20853
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.958984752694889e-05,
+      "loss": 2.0217,
+      "step": 20854
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9580469593153502e-05,
+      "loss": 0.6426,
+      "step": 20855
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.957109374783118e-05,
+      "loss": 0.2952,
+      "step": 20856
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.956171999113208e-05,
+      "loss": 0.464,
+      "step": 20857
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9552348323206252e-05,
+      "loss": 0.0068,
+      "step": 20858
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9542978744203773e-05,
+      "loss": 0.7961,
+      "step": 20859
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9533611254274712e-05,
+      "loss": 0.3888,
+      "step": 20860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9524245853569075e-05,
+      "loss": 0.9233,
+      "step": 20861
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9514882542236797e-05,
+      "loss": 0.8752,
+      "step": 20862
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9505521320427874e-05,
+      "loss": 1.2009,
+      "step": 20863
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.949616218829217e-05,
+      "loss": 1.7132,
+      "step": 20864
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.948680514597956e-05,
+      "loss": 0.3874,
+      "step": 20865
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9477450193639898e-05,
+      "loss": 0.0011,
+      "step": 20866
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9468097331422995e-05,
+      "loss": 0.6788,
+      "step": 20867
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9458746559478594e-05,
+      "loss": 0.5672,
+      "step": 20868
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9449397877956462e-05,
+      "loss": 0.0027,
+      "step": 20869
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.94400512870063e-05,
+      "loss": 0.0198,
+      "step": 20870
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.943070678677776e-05,
+      "loss": 0.2935,
+      "step": 20871
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9421364377420502e-05,
+      "loss": 4.6262,
+      "step": 20872
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9412024059084162e-05,
+      "loss": 0.0034,
+      "step": 20873
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9402685831918195e-05,
+      "loss": 0.3555,
+      "step": 20874
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9393349696072245e-05,
+      "loss": 0.0296,
+      "step": 20875
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9384015651695827e-05,
+      "loss": 0.0045,
+      "step": 20876
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.937468369893832e-05,
+      "loss": 0.0031,
+      "step": 20877
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9365353837949255e-05,
+      "loss": 0.6684,
+      "step": 20878
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.935602606887796e-05,
+      "loss": 0.8978,
+      "step": 20879
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9346700391873822e-05,
+      "loss": 0.0032,
+      "step": 20880
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9337376807086247e-05,
+      "loss": 1.7832,
+      "step": 20881
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9328055314664455e-05,
+      "loss": 0.4325,
+      "step": 20882
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.931873591475772e-05,
+      "loss": 0.2425,
+      "step": 20883
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9309418607515365e-05,
+      "loss": 0.5288,
+      "step": 20884
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9300103393086496e-05,
+      "loss": 0.6162,
+      "step": 20885
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9290790271620282e-05,
+      "loss": 0.7808,
+      "step": 20886
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.928147924326597e-05,
+      "loss": 0.2819,
+      "step": 20887
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.927217030817255e-05,
+      "loss": 3.3849,
+      "step": 20888
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9262863466489132e-05,
+      "loss": 0.3689,
+      "step": 20889
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9253558718364736e-05,
+      "loss": 0.2597,
+      "step": 20890
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9244256063948383e-05,
+      "loss": 0.0024,
+      "step": 20891
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9234955503389033e-05,
+      "loss": 1.0621,
+      "step": 20892
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9225657036835617e-05,
+      "loss": 2.1036,
+      "step": 20893
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9216360664437037e-05,
+      "loss": 0.4442,
+      "step": 20894
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9207066386342168e-05,
+      "loss": 0.092,
+      "step": 20895
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.919777420269983e-05,
+      "loss": 0.6643,
+      "step": 20896
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9188484113658862e-05,
+      "loss": 1.7372,
+      "step": 20897
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.917919611936794e-05,
+      "loss": 2.2926,
+      "step": 20898
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9169910219975905e-05,
+      "loss": 0.1333,
+      "step": 20899
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.916062641563142e-05,
+      "loss": 0.844,
+      "step": 20900
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9151344706483084e-05,
+      "loss": 0.3244,
+      "step": 20901
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9142065092679616e-05,
+      "loss": 1.483,
+      "step": 20902
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9132787574369625e-05,
+      "loss": 0.4562,
+      "step": 20903
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9123512151701558e-05,
+      "loss": 0.7315,
+      "step": 20904
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.911423882482407e-05,
+      "loss": 0.2153,
+      "step": 20905
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9104967593885652e-05,
+      "loss": 0.2397,
+      "step": 20906
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9095698459034654e-05,
+      "loss": 0.2041,
+      "step": 20907
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9086431420419656e-05,
+      "loss": 0.6197,
+      "step": 20908
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.907716647818896e-05,
+      "loss": 0.2566,
+      "step": 20909
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9067903632490942e-05,
+      "loss": 0.1894,
+      "step": 20910
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9058642883473957e-05,
+      "loss": 0.4875,
+      "step": 20911
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.904938423128628e-05,
+      "loss": 0.2617,
+      "step": 20912
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9040127676076194e-05,
+      "loss": 0.2882,
+      "step": 20913
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9030873217991915e-05,
+      "loss": 0.0091,
+      "step": 20914
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.902162085718164e-05,
+      "loss": 0.1079,
+      "step": 20915
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.901237059379355e-05,
+      "loss": 0.439,
+      "step": 20916
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.9003122427975748e-05,
+      "loss": 0.001,
+      "step": 20917
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8993876359876343e-05,
+      "loss": 0.6411,
+      "step": 20918
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.89846323896434e-05,
+      "loss": 0.0024,
+      "step": 20919
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8975390517424937e-05,
+      "loss": 0.4665,
+      "step": 20920
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8966150743368967e-05,
+      "loss": 0.1575,
+      "step": 20921
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8956913067623425e-05,
+      "loss": 0.3485,
+      "step": 20922
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8947677490336255e-05,
+      "loss": 0.1523,
+      "step": 20923
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8938444011655358e-05,
+      "loss": 0.8946,
+      "step": 20924
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.892921263172861e-05,
+      "loss": 0.5447,
+      "step": 20925
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8919983350703738e-05,
+      "loss": 0.3003,
+      "step": 20926
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.891075616872868e-05,
+      "loss": 0.154,
+      "step": 20927
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8901531085951106e-05,
+      "loss": 0.4672,
+      "step": 20928
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8892308102518712e-05,
+      "loss": 0.2355,
+      "step": 20929
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8883087218579312e-05,
+      "loss": 1.9261,
+      "step": 20930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.887386843428044e-05,
+      "loss": 0.2,
+      "step": 20931
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8864651749769765e-05,
+      "loss": 1.2109,
+      "step": 20932
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8855437165194932e-05,
+      "loss": 0.9345,
+      "step": 20933
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8846224680703437e-05,
+      "loss": 0.0341,
+      "step": 20934
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8837014296442766e-05,
+      "loss": 0.5416,
+      "step": 20935
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.882780601256055e-05,
+      "loss": 2.5479,
+      "step": 20936
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8818599829204117e-05,
+      "loss": 0.7574,
+      "step": 20937
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8809395746520924e-05,
+      "loss": 0.5265,
+      "step": 20938
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8800193764658377e-05,
+      "loss": 0.413,
+      "step": 20939
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8790993883763817e-05,
+      "loss": 0.0025,
+      "step": 20940
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8781796103984586e-05,
+      "loss": 0.3809,
+      "step": 20941
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8772600425467936e-05,
+      "loss": 0.5379,
+      "step": 20942
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8763406848361166e-05,
+      "loss": 0.693,
+      "step": 20943
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8754215372811477e-05,
+      "loss": 0.4754,
+      "step": 20944
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8745025998966046e-05,
+      "loss": 0.6695,
+      "step": 20945
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8735838726972073e-05,
+      "loss": 0.2443,
+      "step": 20946
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.872665355697658e-05,
+      "loss": 0.1184,
+      "step": 20947
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8717470489126758e-05,
+      "loss": 0.0026,
+      "step": 20948
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.870828952356964e-05,
+      "loss": 0.3985,
+      "step": 20949
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8699110660452147e-05,
+      "loss": 0.0679,
+      "step": 20950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8689933899921395e-05,
+      "loss": 0.7805,
+      "step": 20951
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.868075924212431e-05,
+      "loss": 0.2449,
+      "step": 20952
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.867158668720771e-05,
+      "loss": 0.0111,
+      "step": 20953
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.86624162353186e-05,
+      "loss": 0.0068,
+      "step": 20954
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8653247886603807e-05,
+      "loss": 0.2861,
+      "step": 20955
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.864408164121006e-05,
+      "loss": 0.0011,
+      "step": 20956
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.863491749928427e-05,
+      "loss": 1.2947,
+      "step": 20957
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8625755460973108e-05,
+      "loss": 0.9483,
+      "step": 20958
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8616595526423262e-05,
+      "loss": 0.4627,
+      "step": 20959
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.860743769578151e-05,
+      "loss": 0.4298,
+      "step": 20960
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8598281969194444e-05,
+      "loss": 0.3576,
+      "step": 20961
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8589128346808673e-05,
+      "loss": 0.1577,
+      "step": 20962
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8579976828770798e-05,
+      "loss": 0.5748,
+      "step": 20963
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.857082741522735e-05,
+      "loss": 0.6746,
+      "step": 20964
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8561680106324865e-05,
+      "loss": 0.6615,
+      "step": 20965
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.855253490220981e-05,
+      "loss": 0.3363,
+      "step": 20966
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8543391803028628e-05,
+      "loss": 0.5991,
+      "step": 20967
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8534250808927743e-05,
+      "loss": 0.0021,
+      "step": 20968
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.852511192005353e-05,
+      "loss": 0.4243,
+      "step": 20969
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.851597513655233e-05,
+      "loss": 0.0051,
+      "step": 20970
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8506840458570454e-05,
+      "loss": 0.003,
+      "step": 20971
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8497707886254188e-05,
+      "loss": 0.8537,
+      "step": 20972
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8488577419749777e-05,
+      "loss": 0.0031,
+      "step": 20973
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.847944905920341e-05,
+      "loss": 0.234,
+      "step": 20974
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.84703228047613e-05,
+      "loss": 0.4626,
+      "step": 20975
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8461198656569596e-05,
+      "loss": 0.1871,
+      "step": 20976
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.845207661477435e-05,
+      "loss": 0.0011,
+      "step": 20977
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8442956679521647e-05,
+      "loss": 0.9005,
+      "step": 20978
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8433838850957617e-05,
+      "loss": 0.0018,
+      "step": 20979
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8424723129228174e-05,
+      "loss": 0.4386,
+      "step": 20980
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.841560951447929e-05,
+      "loss": 0.4679,
+      "step": 20981
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8406498006856995e-05,
+      "loss": 0.2709,
+      "step": 20982
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8397388606507108e-05,
+      "loss": 0.0005,
+      "step": 20983
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.838828131357551e-05,
+      "loss": 0.8364,
+      "step": 20984
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8379176128208133e-05,
+      "loss": 0.3961,
+      "step": 20985
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.837007305055068e-05,
+      "loss": 0.0038,
+      "step": 20986
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8360972080748955e-05,
+      "loss": 0.4663,
+      "step": 20987
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.83518732189487e-05,
+      "loss": 0.2312,
+      "step": 20988
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8342776465295616e-05,
+      "loss": 0.5812,
+      "step": 20989
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8333681819935376e-05,
+      "loss": 0.5675,
+      "step": 20990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8324589283013624e-05,
+      "loss": 1.0513,
+      "step": 20991
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8315498854675965e-05,
+      "loss": 0.2906,
+      "step": 20992
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8306410535067943e-05,
+      "loss": 0.2094,
+      "step": 20993
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8297324324335117e-05,
+      "loss": 0.0013,
+      "step": 20994
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.828824022262303e-05,
+      "loss": 0.0034,
+      "step": 20995
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8279158230077033e-05,
+      "loss": 0.0179,
+      "step": 20996
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8270078346842648e-05,
+      "loss": 0.0041,
+      "step": 20997
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8261000573065304e-05,
+      "loss": 0.4943,
+      "step": 20998
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.825192490889027e-05,
+      "loss": 0.2958,
+      "step": 20999
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8242851354462946e-05,
+      "loss": 0.6382,
+      "step": 21000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8233779909928665e-05,
+      "loss": 0.29,
+      "step": 21001
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8224710575432577e-05,
+      "loss": 0.3642,
+      "step": 21002
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.821564335112e-05,
+      "loss": 0.3241,
+      "step": 21003
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8206578237136153e-05,
+      "loss": 1.4385,
+      "step": 21004
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8197515233626104e-05,
+      "loss": 0.0025,
+      "step": 21005
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8188454340735115e-05,
+      "loss": 0.1818,
+      "step": 21006
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8179395558608156e-05,
+      "loss": 0.2282,
+      "step": 21007
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.817033888739031e-05,
+      "loss": 0.4581,
+      "step": 21008
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8161284327226715e-05,
+      "loss": 0.001,
+      "step": 21009
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8152231878262253e-05,
+      "loss": 0.8143,
+      "step": 21010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.814318154064192e-05,
+      "loss": 0.0075,
+      "step": 21011
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.813413331451064e-05,
+      "loss": 1.3061,
+      "step": 21012
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8125087200013327e-05,
+      "loss": 0.4916,
+      "step": 21013
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8116043197294815e-05,
+      "loss": 0.0006,
+      "step": 21014
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.810700130649994e-05,
+      "loss": 0.5945,
+      "step": 21015
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8097961527773502e-05,
+      "loss": 0.0456,
+      "step": 21016
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8088923861260253e-05,
+      "loss": 0.5314,
+      "step": 21017
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8079888307104907e-05,
+      "loss": 0.8495,
+      "step": 21018
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8070854865452167e-05,
+      "loss": 0.3078,
+      "step": 21019
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.80618235364467e-05,
+      "loss": 0.9226,
+      "step": 21020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.805279432023311e-05,
+      "loss": 0.2374,
+      "step": 21021
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8043767216955995e-05,
+      "loss": 0.4066,
+      "step": 21022
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.803474222675991e-05,
+      "loss": 0.3196,
+      "step": 21023
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8025719349789382e-05,
+      "loss": 0.3445,
+      "step": 21024
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8016698586188915e-05,
+      "loss": 0.2275,
+      "step": 21025
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8007679936102913e-05,
+      "loss": 0.3097,
+      "step": 21026
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7998663399675796e-05,
+      "loss": 0.3494,
+      "step": 21027
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7989648977052023e-05,
+      "loss": 0.3006,
+      "step": 21028
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7980636668375888e-05,
+      "loss": 0.6323,
+      "step": 21029
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.79716264737917e-05,
+      "loss": 0.4543,
+      "step": 21030
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7962618393443818e-05,
+      "loss": 0.427,
+      "step": 21031
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7953612427476403e-05,
+      "loss": 0.1992,
+      "step": 21032
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7944608576033697e-05,
+      "loss": 0.4818,
+      "step": 21033
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7935606839259947e-05,
+      "loss": 0.4086,
+      "step": 21034
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7926607217299223e-05,
+      "loss": 0.9405,
+      "step": 21035
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.791760971029567e-05,
+      "loss": 0.2489,
+      "step": 21036
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.790861431839338e-05,
+      "loss": 0.0486,
+      "step": 21037
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7899621041736374e-05,
+      "loss": 0.0282,
+      "step": 21038
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7890629880468692e-05,
+      "loss": 0.509,
+      "step": 21039
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.788164083473431e-05,
+      "loss": 1.365,
+      "step": 21040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7872653904677148e-05,
+      "loss": 0.9276,
+      "step": 21041
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7863669090441152e-05,
+      "loss": 0.386,
+      "step": 21042
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7854686392170192e-05,
+      "loss": 0.2856,
+      "step": 21043
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.784570581000814e-05,
+      "loss": 0.001,
+      "step": 21044
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.783672734409869e-05,
+      "loss": 0.0069,
+      "step": 21045
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.782775099458575e-05,
+      "loss": 3.3904,
+      "step": 21046
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7818776761613053e-05,
+      "loss": 0.2637,
+      "step": 21047
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.780980464532418e-05,
+      "loss": 0.1697,
+      "step": 21048
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.780083464586296e-05,
+      "loss": 0.0011,
+      "step": 21049
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7791866763372986e-05,
+      "loss": 0.0006,
+      "step": 21050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7782900997997778e-05,
+      "loss": 0.0326,
+      "step": 21051
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7773937349881024e-05,
+      "loss": 0.1584,
+      "step": 21052
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.776497581916624e-05,
+      "loss": 0.1745,
+      "step": 21053
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.775601640599686e-05,
+      "loss": 1.7451,
+      "step": 21054
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.774705911051647e-05,
+      "loss": 0.3993,
+      "step": 21055
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7738103932868407e-05,
+      "loss": 0.4844,
+      "step": 21056
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7729150873196073e-05,
+      "loss": 0.2289,
+      "step": 21057
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7720199931642954e-05,
+      "loss": 1.4447,
+      "step": 21058
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7711251108352282e-05,
+      "loss": 0.5505,
+      "step": 21059
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7702304403467342e-05,
+      "loss": 0.2198,
+      "step": 21060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.769335981713151e-05,
+      "loss": 0.5945,
+      "step": 21061
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.7684417349487928e-05,
+      "loss": 0.4655,
+      "step": 21062
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7675477000679823e-05,
+      "loss": 0.0596,
+      "step": 21063
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.766653877085036e-05,
+      "loss": 0.8801,
+      "step": 21064
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7657602660142673e-05,
+      "loss": 0.1871,
+      "step": 21065
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7648668668699866e-05,
+      "loss": 0.2437,
+      "step": 21066
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7639736796664995e-05,
+      "loss": 0.0046,
+      "step": 21067
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7630807044181106e-05,
+      "loss": 0.3583,
+      "step": 21068
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.762187941139117e-05,
+      "loss": 0.2086,
+      "step": 21069
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7612953898438157e-05,
+      "loss": 0.1591,
+      "step": 21070
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7604030505465022e-05,
+      "loss": 0.1506,
+      "step": 21071
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.759510923261463e-05,
+      "loss": 1.3003,
+      "step": 21072
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.758619008002985e-05,
+      "loss": 2.4048,
+      "step": 21073
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.757727304785355e-05,
+      "loss": 0.8341,
+      "step": 21074
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7568358136228405e-05,
+      "loss": 0.9675,
+      "step": 21075
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7559445345297297e-05,
+      "loss": 0.2654,
+      "step": 21076
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7550534675202943e-05,
+      "loss": 0.6651,
+      "step": 21077
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7541626126087948e-05,
+      "loss": 0.6932,
+      "step": 21078
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7532719698094998e-05,
+      "loss": 0.3328,
+      "step": 21079
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7523815391366797e-05,
+      "loss": 0.3649,
+      "step": 21080
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7514913206045843e-05,
+      "loss": 0.3229,
+      "step": 21081
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7506013142274695e-05,
+      "loss": 0.2464,
+      "step": 21082
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7497115200195966e-05,
+      "loss": 0.263,
+      "step": 21083
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.748821937995203e-05,
+      "loss": 0.5488,
+      "step": 21084
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7479325681685403e-05,
+      "loss": 0.6678,
+      "step": 21085
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.747043410553849e-05,
+      "loss": 0.3448,
+      "step": 21086
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7461544651653657e-05,
+      "loss": 0.1091,
+      "step": 21087
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7452657320173296e-05,
+      "loss": 0.0508,
+      "step": 21088
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.744377211123969e-05,
+      "loss": 0.0068,
+      "step": 21089
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.743488902499514e-05,
+      "loss": 0.3916,
+      "step": 21090
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7426008061581888e-05,
+      "loss": 0.0005,
+      "step": 21091
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7417129221142144e-05,
+      "loss": 0.3774,
+      "step": 21092
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7408252503818137e-05,
+      "loss": 0.2514,
+      "step": 21093
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7399377909751917e-05,
+      "loss": 1.3739,
+      "step": 21094
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7390505439085684e-05,
+      "loss": 0.4486,
+      "step": 21095
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.738163509196151e-05,
+      "loss": 0.834,
+      "step": 21096
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7372766868521348e-05,
+      "loss": 0.4734,
+      "step": 21097
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.736390076890732e-05,
+      "loss": 0.0017,
+      "step": 21098
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7355036793261406e-05,
+      "loss": 0.031,
+      "step": 21099
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7346174941725416e-05,
+      "loss": 0.6201,
+      "step": 21100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.73373152144414e-05,
+      "loss": 0.3094,
+      "step": 21101
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7328457611551212e-05,
+      "loss": 0.0027,
+      "step": 21102
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7319602133196603e-05,
+      "loss": 0.0031,
+      "step": 21103
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7310748779519496e-05,
+      "loss": 0.5611,
+      "step": 21104
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7301897550661577e-05,
+      "loss": 0.4035,
+      "step": 21105
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.729304844676458e-05,
+      "loss": 0.0022,
+      "step": 21106
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.728420146797031e-05,
+      "loss": 0.0018,
+      "step": 21107
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7275356614420356e-05,
+      "loss": 0.2692,
+      "step": 21108
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7266513886256333e-05,
+      "loss": 0.8876,
+      "step": 21109
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7257673283619948e-05,
+      "loss": 4.127,
+      "step": 21110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.724883480665269e-05,
+      "loss": 0.2512,
+      "step": 21111
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.723999845549609e-05,
+      "loss": 0.5246,
+      "step": 21112
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7231164230291687e-05,
+      "loss": 0.3568,
+      "step": 21113
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.722233213118092e-05,
+      "loss": 0.3112,
+      "step": 21114
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7213502158305243e-05,
+      "loss": 0.3132,
+      "step": 21115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.720467431180604e-05,
+      "loss": 0.5605,
+      "step": 21116
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.719584859182466e-05,
+      "loss": 0.2805,
+      "step": 21117
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7187024998502485e-05,
+      "loss": 0.3422,
+      "step": 21118
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.717820353198076e-05,
+      "loss": 0.4805,
+      "step": 21119
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7169384192400774e-05,
+      "loss": 0.3478,
+      "step": 21120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.716056697990375e-05,
+      "loss": 1.2061,
+      "step": 21121
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7151751894630873e-05,
+      "loss": 0.6577,
+      "step": 21122
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.714293893672336e-05,
+      "loss": 0.8029,
+      "step": 21123
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.713412810632222e-05,
+      "loss": 0.7843,
+      "step": 21124
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7125319403568636e-05,
+      "loss": 0.2548,
+      "step": 21125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7116512828603686e-05,
+      "loss": 0.2939,
+      "step": 21126
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7107708381568296e-05,
+      "loss": 0.4247,
+      "step": 21127
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.709890606260353e-05,
+      "loss": 1.8621,
+      "step": 21128
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7090105871850367e-05,
+      "loss": 0.5842,
+      "step": 21129
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7081307809449653e-05,
+      "loss": 0.3322,
+      "step": 21130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7072511875542284e-05,
+      "loss": 0.2128,
+      "step": 21131
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.70637180702692e-05,
+      "loss": 0.1564,
+      "step": 21132
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.705492639377114e-05,
+      "loss": 0.4981,
+      "step": 21133
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.704613684618889e-05,
+      "loss": 0.311,
+      "step": 21134
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.703734942766324e-05,
+      "loss": 0.5446,
+      "step": 21135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.702856413833489e-05,
+      "loss": 0.8137,
+      "step": 21136
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7019780978344526e-05,
+      "loss": 1.04,
+      "step": 21137
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.7010999947832776e-05,
+      "loss": 2.0801,
+      "step": 21138
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.700222104694029e-05,
+      "loss": 1.2841,
+      "step": 21139
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6993444275807622e-05,
+      "loss": 0.1512,
+      "step": 21140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6984669634575345e-05,
+      "loss": 1.203,
+      "step": 21141
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.697589712338394e-05,
+      "loss": 0.0042,
+      "step": 21142
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6967126742373905e-05,
+      "loss": 0.7114,
+      "step": 21143
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6958358491685687e-05,
+      "loss": 0.2475,
+      "step": 21144
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.694959237145971e-05,
+      "loss": 0.3298,
+      "step": 21145
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6940828381836275e-05,
+      "loss": 0.0058,
+      "step": 21146
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6932066522955808e-05,
+      "loss": 0.297,
+      "step": 21147
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6923306794958623e-05,
+      "loss": 0.3377,
+      "step": 21148
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.69145491979849e-05,
+      "loss": 0.1356,
+      "step": 21149
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6905793732174955e-05,
+      "loss": 1.5157,
+      "step": 21150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6897040397669027e-05,
+      "loss": 2.093,
+      "step": 21151
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6888289194607167e-05,
+      "loss": 0.0012,
+      "step": 21152
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6879540123129638e-05,
+      "loss": 0.2944,
+      "step": 21153
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.687079318337647e-05,
+      "loss": 0.4556,
+      "step": 21154
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.686204837548772e-05,
+      "loss": 1.2071,
+      "step": 21155
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.685330569960351e-05,
+      "loss": 0.621,
+      "step": 21156
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6844565155863744e-05,
+      "loss": 0.3063,
+      "step": 21157
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.683582674440841e-05,
+      "loss": 0.0407,
+      "step": 21158
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6827090465377512e-05,
+      "loss": 0.786,
+      "step": 21159
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6818356318910874e-05,
+      "loss": 0.1267,
+      "step": 21160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6809624305148334e-05,
+      "loss": 0.2278,
+      "step": 21161
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6800894424229845e-05,
+      "loss": 0.3633,
+      "step": 21162
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6792166676295095e-05,
+      "loss": 0.2552,
+      "step": 21163
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6783441061483887e-05,
+      "loss": 0.7019,
+      "step": 21164
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.677471757993593e-05,
+      "loss": 2.291,
+      "step": 21165
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.676599623179093e-05,
+      "loss": 0.3527,
+      "step": 21166
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.675727701718855e-05,
+      "loss": 1.1982,
+      "step": 21167
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6748559936268392e-05,
+      "loss": 0.4819,
+      "step": 21168
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6739844989170082e-05,
+      "loss": 0.0036,
+      "step": 21169
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6731132176033146e-05,
+      "loss": 0.2294,
+      "step": 21170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.672242149699714e-05,
+      "loss": 0.644,
+      "step": 21171
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6713712952201552e-05,
+      "loss": 0.0049,
+      "step": 21172
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6705006541785753e-05,
+      "loss": 0.0012,
+      "step": 21173
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.669630226588925e-05,
+      "loss": 0.2144,
+      "step": 21174
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6687600124651445e-05,
+      "loss": 0.0061,
+      "step": 21175
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.667890011821158e-05,
+      "loss": 0.6863,
+      "step": 21176
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.667020224670909e-05,
+      "loss": 0.2198,
+      "step": 21177
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6661506510283218e-05,
+      "loss": 0.7914,
+      "step": 21178
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6652812909073198e-05,
+      "loss": 0.3754,
+      "step": 21179
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6644121443218205e-05,
+      "loss": 0.0022,
+      "step": 21180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6635432112857543e-05,
+      "loss": 0.0806,
+      "step": 21181
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6626744918130252e-05,
+      "loss": 0.3802,
+      "step": 21182
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.661805985917547e-05,
+      "loss": 0.6686,
+      "step": 21183
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6609376936132284e-05,
+      "loss": 0.8737,
+      "step": 21184
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6600696149139735e-05,
+      "loss": 0.3012,
+      "step": 21185
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6592017498336828e-05,
+      "loss": 0.3859,
+      "step": 21186
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.658334098386255e-05,
+      "loss": 0.366,
+      "step": 21187
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.657466660585583e-05,
+      "loss": 1.1303,
+      "step": 21188
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.656599436445558e-05,
+      "loss": 0.2096,
+      "step": 21189
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.655732425980068e-05,
+      "loss": 0.0058,
+      "step": 21190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6548656292029966e-05,
+      "loss": 0.3373,
+      "step": 21191
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6539990461282227e-05,
+      "loss": 0.6086,
+      "step": 21192
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.653132676769625e-05,
+      "loss": 0.3445,
+      "step": 21193
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6522665211410778e-05,
+      "loss": 0.0315,
+      "step": 21194
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6514005792564482e-05,
+      "loss": 0.2292,
+      "step": 21195
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6505348511296067e-05,
+      "loss": 0.2896,
+      "step": 21196
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6496693367744157e-05,
+      "loss": 0.388,
+      "step": 21197
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6488040362047304e-05,
+      "loss": 0.7317,
+      "step": 21198
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6479389494344135e-05,
+      "loss": 0.3999,
+      "step": 21199
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.647074076477317e-05,
+      "loss": 0.0009,
+      "step": 21200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6462094173472852e-05,
+      "loss": 0.6437,
+      "step": 21201
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6453449720581745e-05,
+      "loss": 0.86,
+      "step": 21202
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6444807406238165e-05,
+      "loss": 0.8901,
+      "step": 21203
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6436167230580542e-05,
+      "loss": 0.3029,
+      "step": 21204
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.642752919374731e-05,
+      "loss": 0.8499,
+      "step": 21205
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6418893295876696e-05,
+      "loss": 0.459,
+      "step": 21206
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6410259537107013e-05,
+      "loss": 0.0008,
+      "step": 21207
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.640162791757658e-05,
+      "loss": 0.251,
+      "step": 21208
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6392998437423543e-05,
+      "loss": 0.0404,
+      "step": 21209
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6384371096786087e-05,
+      "loss": 0.0167,
+      "step": 21210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6375745895802468e-05,
+      "loss": 0.4723,
+      "step": 21211
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.636712283461071e-05,
+      "loss": 0.1481,
+      "step": 21212
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6358501913348914e-05,
+      "loss": 1.5447,
+      "step": 21213
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6349883132155145e-05,
+      "loss": 0.2665,
+      "step": 21214
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6341266491167422e-05,
+      "loss": 0.3254,
+      "step": 21215
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6332651990523716e-05,
+      "loss": 0.2372,
+      "step": 21216
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6324039630361985e-05,
+      "loss": 0.2956,
+      "step": 21217
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6315429410820137e-05,
+      "loss": 0.0426,
+      "step": 21218
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6306821332036057e-05,
+      "loss": 0.3137,
+      "step": 21219
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.629821539414759e-05,
+      "loss": 0.5566,
+      "step": 21220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6289611597292575e-05,
+      "loss": 0.7861,
+      "step": 21221
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6281009941608696e-05,
+      "loss": 0.003,
+      "step": 21222
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6272410427233784e-05,
+      "loss": 0.448,
+      "step": 21223
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6263813054305556e-05,
+      "loss": 0.2731,
+      "step": 21224
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.625521782296161e-05,
+      "loss": 0.2201,
+      "step": 21225
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6246624733339642e-05,
+      "loss": 0.7328,
+      "step": 21226
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6238033785577264e-05,
+      "loss": 0.1039,
+      "step": 21227
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6229444979811983e-05,
+      "loss": 0.2239,
+      "step": 21228
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.622085831618139e-05,
+      "loss": 0.7772,
+      "step": 21229
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6212273794823022e-05,
+      "loss": 0.3292,
+      "step": 21230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.620369141587427e-05,
+      "loss": 0.4833,
+      "step": 21231
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.619511117947259e-05,
+      "loss": 0.0022,
+      "step": 21232
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.618653308575541e-05,
+      "loss": 0.0049,
+      "step": 21233
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6177957134860065e-05,
+      "loss": 0.7476,
+      "step": 21234
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6169383326923897e-05,
+      "loss": 0.2567,
+      "step": 21235
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.616081166208421e-05,
+      "loss": 0.2882,
+      "step": 21236
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.615224214047825e-05,
+      "loss": 0.1378,
+      "step": 21237
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6143674762243253e-05,
+      "loss": 0.3365,
+      "step": 21238
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6135109527516428e-05,
+      "loss": 0.2506,
+      "step": 21239
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6126546436434913e-05,
+      "loss": 0.243,
+      "step": 21240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6117985489135848e-05,
+      "loss": 0.2734,
+      "step": 21241
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6109426685756312e-05,
+      "loss": 0.3357,
+      "step": 21242
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6100870026433355e-05,
+      "loss": 0.6444,
+      "step": 21243
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.609231551130402e-05,
+      "loss": 0.2267,
+      "step": 21244
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6083763140505285e-05,
+      "loss": 0.6337,
+      "step": 21245
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6075212914174136e-05,
+      "loss": 0.0214,
+      "step": 21246
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6066664832447397e-05,
+      "loss": 0.4801,
+      "step": 21247
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.605811889546204e-05,
+      "loss": 0.9293,
+      "step": 21248
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6049575103354928e-05,
+      "loss": 0.1963,
+      "step": 21249
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.6041033456262765e-05,
+      "loss": 1.5985,
+      "step": 21250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.603249395432249e-05,
+      "loss": 0.0008,
+      "step": 21251
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.602395659767071e-05,
+      "loss": 0.2575,
+      "step": 21252
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.601542138644419e-05,
+      "loss": 0.0014,
+      "step": 21253
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.600688832077966e-05,
+      "loss": 0.7556,
+      "step": 21254
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.59983574008137e-05,
+      "loss": 0.6768,
+      "step": 21255
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.59898286266829e-05,
+      "loss": 0.6032,
+      "step": 21256
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.598130199852393e-05,
+      "loss": 0.6908,
+      "step": 21257
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5972777516473238e-05,
+      "loss": 1.4238,
+      "step": 21258
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5964255180667346e-05,
+      "loss": 0.2053,
+      "step": 21259
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5955734991242807e-05,
+      "loss": 0.3507,
+      "step": 21260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5947216948335983e-05,
+      "loss": 0.4514,
+      "step": 21261
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.593870105208328e-05,
+      "loss": 0.4249,
+      "step": 21262
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.593018730262109e-05,
+      "loss": 0.2402,
+      "step": 21263
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5921675700085733e-05,
+      "loss": 0.0008,
+      "step": 21264
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5913166244613505e-05,
+      "loss": 0.4805,
+      "step": 21265
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.59046589363407e-05,
+      "loss": 0.2628,
+      "step": 21266
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.589615377540353e-05,
+      "loss": 0.9028,
+      "step": 21267
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5887650761938198e-05,
+      "loss": 0.1663,
+      "step": 21268
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5879149896080857e-05,
+      "loss": 0.0029,
+      "step": 21269
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.587065117796768e-05,
+      "loss": 0.4518,
+      "step": 21270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.586215460773466e-05,
+      "loss": 0.1918,
+      "step": 21271
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5853660185517957e-05,
+      "loss": 0.5499,
+      "step": 21272
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5845167911453593e-05,
+      "loss": 0.3244,
+      "step": 21273
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.583667778567746e-05,
+      "loss": 0.2253,
+      "step": 21274
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5828189808325614e-05,
+      "loss": 0.5676,
+      "step": 21275
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5819703979533964e-05,
+      "loss": 1.6963,
+      "step": 21276
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5811220299438337e-05,
+      "loss": 0.0048,
+      "step": 21277
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5802738768174638e-05,
+      "loss": 0.609,
+      "step": 21278
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.579425938587871e-05,
+      "loss": 0.2688,
+      "step": 21279
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5785782152686262e-05,
+      "loss": 0.492,
+      "step": 21280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.577730706873313e-05,
+      "loss": 0.3275,
+      "step": 21281
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5768834134154962e-05,
+      "loss": 0.2209,
+      "step": 21282
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5760363349087446e-05,
+      "loss": 0.2618,
+      "step": 21283
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.575189471366626e-05,
+      "loss": 0.4836,
+      "step": 21284
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5743428228027004e-05,
+      "loss": 0.0995,
+      "step": 21285
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5734963892305245e-05,
+      "loss": 1.305,
+      "step": 21286
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.572650170663654e-05,
+      "loss": 0.973,
+      "step": 21287
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5718041671156395e-05,
+      "loss": 0.2902,
+      "step": 21288
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5709583786000286e-05,
+      "loss": 0.223,
+      "step": 21289
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.570112805130365e-05,
+      "loss": 0.2217,
+      "step": 21290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5692674467201888e-05,
+      "loss": 0.0036,
+      "step": 21291
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5684223033830385e-05,
+      "loss": 0.2014,
+      "step": 21292
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5675773751324468e-05,
+      "loss": 0.4582,
+      "step": 21293
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5667326619819442e-05,
+      "loss": 0.6129,
+      "step": 21294
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.565888163945057e-05,
+      "loss": 0.1309,
+      "step": 21295
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5650438810353093e-05,
+      "loss": 2.3246,
+      "step": 21296
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5641998132662203e-05,
+      "loss": 2.3269,
+      "step": 21297
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5633559606513107e-05,
+      "loss": 0.4913,
+      "step": 21298
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.562512323204083e-05,
+      "loss": 0.23,
+      "step": 21299
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5616689009380618e-05,
+      "loss": 0.59,
+      "step": 21300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5608256938667408e-05,
+      "loss": 0.0115,
+      "step": 21301
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5599827020036225e-05,
+      "loss": 0.3533,
+      "step": 21302
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5591399253622185e-05,
+      "loss": 0.3973,
+      "step": 21303
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.558297363956012e-05,
+      "loss": 0.8605,
+      "step": 21304
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5574550177984978e-05,
+      "loss": 0.4588,
+      "step": 21305
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5566128869031734e-05,
+      "loss": 0.8418,
+      "step": 21306
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.555770971283515e-05,
+      "loss": 0.2857,
+      "step": 21307
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5549292709530042e-05,
+      "loss": 0.6867,
+      "step": 21308
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5540877859251295e-05,
+      "loss": 0.1377,
+      "step": 21309
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.5532465162133563e-05,
+      "loss": 0.6204,
+      "step": 21310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.552405461831159e-05,
+      "loss": 0.3524,
+      "step": 21311
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5515646227920066e-05,
+      "loss": 0.5941,
+      "step": 21312
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5507239991093617e-05,
+      "loss": 1.9795,
+      "step": 21313
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.549883590796689e-05,
+      "loss": 0.0356,
+      "step": 21314
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5490433978674454e-05,
+      "loss": 0.4942,
+      "step": 21315
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.548203420335084e-05,
+      "loss": 0.5301,
+      "step": 21316
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5473636582130554e-05,
+      "loss": 0.0075,
+      "step": 21317
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5465241115148093e-05,
+      "loss": 0.0013,
+      "step": 21318
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5456847802537926e-05,
+      "loss": 0.2645,
+      "step": 21319
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5448456644434348e-05,
+      "loss": 0.1612,
+      "step": 21320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5440067640971838e-05,
+      "loss": 0.0051,
+      "step": 21321
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5431680792284734e-05,
+      "loss": 0.0044,
+      "step": 21322
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5423296098507232e-05,
+      "loss": 0.3835,
+      "step": 21323
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5414913559773706e-05,
+      "loss": 0.2333,
+      "step": 21324
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5406533176218384e-05,
+      "loss": 0.0197,
+      "step": 21325
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5398154947975373e-05,
+      "loss": 0.8705,
+      "step": 21326
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.538977887517895e-05,
+      "loss": 0.0007,
+      "step": 21327
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5381404957963207e-05,
+      "loss": 0.228,
+      "step": 21328
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.537303319646217e-05,
+      "loss": 0.2385,
+      "step": 21329
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5364663590810032e-05,
+      "loss": 0.2832,
+      "step": 21330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5356296141140707e-05,
+      "loss": 0.2288,
+      "step": 21331
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.534793084758823e-05,
+      "loss": 0.3416,
+      "step": 21332
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5339567710286567e-05,
+      "loss": 0.3619,
+      "step": 21333
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5331206729369605e-05,
+      "loss": 1.0225,
+      "step": 21334
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5322847904971275e-05,
+      "loss": 0.1918,
+      "step": 21335
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5314491237225417e-05,
+      "loss": 0.5692,
+      "step": 21336
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5306136726265843e-05,
+      "loss": 0.1996,
+      "step": 21337
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5297784372226346e-05,
+      "loss": 0.0041,
+      "step": 21338
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5289434175240667e-05,
+      "loss": 0.45,
+      "step": 21339
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5281086135442517e-05,
+      "loss": 0.0015,
+      "step": 21340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5272740252965603e-05,
+      "loss": 0.2518,
+      "step": 21341
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5264396527943555e-05,
+      "loss": 1.3347,
+      "step": 21342
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5256054960509984e-05,
+      "loss": 0.5718,
+      "step": 21343
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5247715550798462e-05,
+      "loss": 0.268,
+      "step": 21344
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.523937829894255e-05,
+      "loss": 0.5766,
+      "step": 21345
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5231043205075743e-05,
+      "loss": 0.2427,
+      "step": 21346
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5222710269331534e-05,
+      "loss": 0.0046,
+      "step": 21347
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5214379491843332e-05,
+      "loss": 0.8449,
+      "step": 21348
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5206050872744597e-05,
+      "loss": 0.0012,
+      "step": 21349
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5197724412168622e-05,
+      "loss": 0.3568,
+      "step": 21350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.518940011024875e-05,
+      "loss": 0.9026,
+      "step": 21351
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5181077967118377e-05,
+      "loss": 0.4545,
+      "step": 21352
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5172757982910677e-05,
+      "loss": 0.065,
+      "step": 21353
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5164440157758878e-05,
+      "loss": 0.1029,
+      "step": 21354
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5156124491796273e-05,
+      "loss": 0.1167,
+      "step": 21355
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5147810985155923e-05,
+      "loss": 0.2232,
+      "step": 21356
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.513949963797097e-05,
+      "loss": 0.0775,
+      "step": 21357
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5131190450374592e-05,
+      "loss": 0.0045,
+      "step": 21358
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5122883422499765e-05,
+      "loss": 0.1026,
+      "step": 21359
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5114578554479533e-05,
+      "loss": 0.3487,
+      "step": 21360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5106275846446892e-05,
+      "loss": 0.6487,
+      "step": 21361
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5097975298534797e-05,
+      "loss": 0.0047,
+      "step": 21362
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5089676910876163e-05,
+      "loss": 0.0057,
+      "step": 21363
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5081380683603882e-05,
+      "loss": 0.9884,
+      "step": 21364
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5073086616850815e-05,
+      "loss": 0.1636,
+      "step": 21365
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5064794710749756e-05,
+      "loss": 0.0017,
+      "step": 21366
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5056504965433513e-05,
+      "loss": 0.3361,
+      "step": 21367
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.504821738103485e-05,
+      "loss": 0.9082,
+      "step": 21368
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5039931957686375e-05,
+      "loss": 0.1932,
+      "step": 21369
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.503164869552088e-05,
+      "loss": 0.0015,
+      "step": 21370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5023367594671014e-05,
+      "loss": 0.6421,
+      "step": 21371
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5015088655269269e-05,
+      "loss": 0.43,
+      "step": 21372
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.5006811877448339e-05,
+      "loss": 0.4723,
+      "step": 21373
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4998537261340749e-05,
+      "loss": 2.1952,
+      "step": 21374
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.499026480707891e-05,
+      "loss": 2.0976,
+      "step": 21375
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4981994514795387e-05,
+      "loss": 0.3467,
+      "step": 21376
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.497372638462262e-05,
+      "loss": 0.5825,
+      "step": 21377
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.496546041669292e-05,
+      "loss": 0.4782,
+      "step": 21378
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4957196611138766e-05,
+      "loss": 0.0532,
+      "step": 21379
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4948934968092419e-05,
+      "loss": 0.2297,
+      "step": 21380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4940675487686171e-05,
+      "loss": 0.8936,
+      "step": 21381
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4932418170052351e-05,
+      "loss": 0.2862,
+      "step": 21382
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4924163015323121e-05,
+      "loss": 0.3245,
+      "step": 21383
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4915910023630706e-05,
+      "loss": 0.0016,
+      "step": 21384
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4907659195107252e-05,
+      "loss": 1.2621,
+      "step": 21385
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4899410529884903e-05,
+      "loss": 0.4942,
+      "step": 21386
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4891164028095721e-05,
+      "loss": 0.8397,
+      "step": 21387
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.488291968987178e-05,
+      "loss": 0.2191,
+      "step": 21388
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4874677515345113e-05,
+      "loss": 0.6964,
+      "step": 21389
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4866437504647676e-05,
+      "loss": 0.4817,
+      "step": 21390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4858199657911436e-05,
+      "loss": 0.4176,
+      "step": 21391
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4849963975268313e-05,
+      "loss": 0.1386,
+      "step": 21392
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4841730456850192e-05,
+      "loss": 0.3682,
+      "step": 21393
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4833499102788898e-05,
+      "loss": 0.4685,
+      "step": 21394
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4825269913216275e-05,
+      "loss": 0.0235,
+      "step": 21395
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4817042888264069e-05,
+      "loss": 0.3116,
+      "step": 21396
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4808818028064057e-05,
+      "loss": 1.1758,
+      "step": 21397
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.480059533274795e-05,
+      "loss": 0.2285,
+      "step": 21398
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4792374802447377e-05,
+      "loss": 0.0021,
+      "step": 21399
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4784156437293965e-05,
+      "loss": 0.5054,
+      "step": 21400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4775940237419427e-05,
+      "loss": 1.0148,
+      "step": 21401
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4767726202955237e-05,
+      "loss": 0.1635,
+      "step": 21402
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4759514334032913e-05,
+      "loss": 0.0401,
+      "step": 21403
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4751304630784077e-05,
+      "loss": 0.0041,
+      "step": 21404
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4743097093340062e-05,
+      "loss": 0.258,
+      "step": 21405
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4734891721832342e-05,
+      "loss": 0.6473,
+      "step": 21406
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4726688516392382e-05,
+      "loss": 0.0017,
+      "step": 21407
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4718487477151475e-05,
+      "loss": 0.0015,
+      "step": 21408
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4710288604240934e-05,
+      "loss": 1.0335,
+      "step": 21409
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4702091897792101e-05,
+      "loss": 0.3691,
+      "step": 21410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4693897357936207e-05,
+      "loss": 0.2738,
+      "step": 21411
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.468570498480448e-05,
+      "loss": 0.6179,
+      "step": 21412
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4677514778528098e-05,
+      "loss": 0.3631,
+      "step": 21413
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4669326739238242e-05,
+      "loss": 0.0012,
+      "step": 21414
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4661140867066002e-05,
+      "loss": 0.9344,
+      "step": 21415
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4652957162142492e-05,
+      "loss": 0.3499,
+      "step": 21416
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4644775624598759e-05,
+      "loss": 0.4112,
+      "step": 21417
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4636596254565763e-05,
+      "loss": 1.4165,
+      "step": 21418
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4628419052174533e-05,
+      "loss": 0.0043,
+      "step": 21419
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4620244017556065e-05,
+      "loss": 0.6481,
+      "step": 21420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4612071150841138e-05,
+      "loss": 0.2242,
+      "step": 21421
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4603900452160727e-05,
+      "loss": 0.195,
+      "step": 21422
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4595731921645682e-05,
+      "loss": 0.6635,
+      "step": 21423
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4587565559426728e-05,
+      "loss": 0.6139,
+      "step": 21424
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4579401365634696e-05,
+      "loss": 0.2497,
+      "step": 21425
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4571239340400365e-05,
+      "loss": 0.4836,
+      "step": 21426
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4563079483854312e-05,
+      "loss": 0.4307,
+      "step": 21427
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4554921796127334e-05,
+      "loss": 2.5947,
+      "step": 21428
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4546766277349991e-05,
+      "loss": 0.0073,
+      "step": 21429
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4538612927652865e-05,
+      "loss": 0.1832,
+      "step": 21430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4530461747166617e-05,
+      "loss": 0.049,
+      "step": 21431
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4522312736021674e-05,
+      "loss": 0.2081,
+      "step": 21432
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.451416589434855e-05,
+      "loss": 0.0865,
+      "step": 21433
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4506021222277775e-05,
+      "loss": 0.2625,
+      "step": 21434
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4497878719939693e-05,
+      "loss": 0.2395,
+      "step": 21435
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4489738387464734e-05,
+      "loss": 0.002,
+      "step": 21436
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4481600224983225e-05,
+      "loss": 0.3783,
+      "step": 21437
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4473464232625515e-05,
+      "loss": 0.323,
+      "step": 21438
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4465330410521881e-05,
+      "loss": 0.349,
+      "step": 21439
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4457198758802569e-05,
+      "loss": 0.7704,
+      "step": 21440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4449069277597792e-05,
+      "loss": 0.9639,
+      "step": 21441
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4440941967037744e-05,
+      "loss": 0.3443,
+      "step": 21442
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4432816827252558e-05,
+      "loss": 0.3635,
+      "step": 21443
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.442469385837236e-05,
+      "loss": 0.5328,
+      "step": 21444
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4416573060527214e-05,
+      "loss": 0.3512,
+      "step": 21445
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4408454433847167e-05,
+      "loss": 0.2152,
+      "step": 21446
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4400337978462262e-05,
+      "loss": 0.6313,
+      "step": 21447
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4392223694502363e-05,
+      "loss": 0.7069,
+      "step": 21448
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4384111582097535e-05,
+      "loss": 0.0023,
+      "step": 21449
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4376001641377654e-05,
+      "loss": 0.003,
+      "step": 21450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4367893872472519e-05,
+      "loss": 0.8994,
+      "step": 21451
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.435978827551199e-05,
+      "loss": 0.6103,
+      "step": 21452
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4351684850625949e-05,
+      "loss": 0.4323,
+      "step": 21453
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4343583597944059e-05,
+      "loss": 0.0412,
+      "step": 21454
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4335484517596064e-05,
+      "loss": 0.1023,
+      "step": 21455
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4327387609711728e-05,
+      "loss": 0.4565,
+      "step": 21456
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.431929287442063e-05,
+      "loss": 0.657,
+      "step": 21457
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.431120031185245e-05,
+      "loss": 0.0675,
+      "step": 21458
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4303109922136735e-05,
+      "loss": 0.7835,
+      "step": 21459
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4295021705403065e-05,
+      "loss": 0.393,
+      "step": 21460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4286935661780968e-05,
+      "loss": 0.0566,
+      "step": 21461
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4278851791399909e-05,
+      "loss": 0.0019,
+      "step": 21462
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4270770094389366e-05,
+      "loss": 0.2169,
+      "step": 21463
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4262690570878721e-05,
+      "loss": 0.0016,
+      "step": 21464
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4254613220997385e-05,
+      "loss": 0.7555,
+      "step": 21465
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4246538044874722e-05,
+      "loss": 0.0011,
+      "step": 21466
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4238465042639945e-05,
+      "loss": 0.8363,
+      "step": 21467
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4230394214422435e-05,
+      "loss": 0.7203,
+      "step": 21468
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4222325560351439e-05,
+      "loss": 0.3039,
+      "step": 21469
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.421425908055605e-05,
+      "loss": 2.0577,
+      "step": 21470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.420619477516557e-05,
+      "loss": 0.3817,
+      "step": 21471
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.419813264430909e-05,
+      "loss": 0.3162,
+      "step": 21472
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4190072688115661e-05,
+      "loss": 0.3292,
+      "step": 21473
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4182014906714412e-05,
+      "loss": 0.1123,
+      "step": 21474
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4173959300234389e-05,
+      "loss": 0.5738,
+      "step": 21475
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4165905868804506e-05,
+      "loss": 0.3252,
+      "step": 21476
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4157854612553842e-05,
+      "loss": 0.4948,
+      "step": 21477
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4149805531611247e-05,
+      "loss": 0.7212,
+      "step": 21478
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4141758626105598e-05,
+      "loss": 0.6662,
+      "step": 21479
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4133713896165844e-05,
+      "loss": 0.8482,
+      "step": 21480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.412567134192073e-05,
+      "loss": 0.023,
+      "step": 21481
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4117630963499038e-05,
+      "loss": 1.62,
+      "step": 21482
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4109592761029614e-05,
+      "loss": 0.7118,
+      "step": 21483
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.410155673464109e-05,
+      "loss": 0.4168,
+      "step": 21484
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4093522884462178e-05,
+      "loss": 0.0016,
+      "step": 21485
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4085491210621542e-05,
+      "loss": 0.3747,
+      "step": 21486
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4077461713247778e-05,
+      "loss": 1.0777,
+      "step": 21487
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.406943439246947e-05,
+      "loss": 0.2758,
+      "step": 21488
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4061409248415162e-05,
+      "loss": 0.7507,
+      "step": 21489
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4053386281213385e-05,
+      "loss": 0.4953,
+      "step": 21490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4045365490992589e-05,
+      "loss": 5.3291,
+      "step": 21491
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4037346877881217e-05,
+      "loss": 0.007,
+      "step": 21492
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4029330442007703e-05,
+      "loss": 0.7668,
+      "step": 21493
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4021316183500392e-05,
+      "loss": 0.0007,
+      "step": 21494
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4013304102487617e-05,
+      "loss": 0.1974,
+      "step": 21495
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4005294199097722e-05,
+      "loss": 0.5224,
+      "step": 21496
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.399728647345889e-05,
+      "loss": 0.4501,
+      "step": 21497
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.398928092569942e-05,
+      "loss": 0.3245,
+      "step": 21498
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.398127755594754e-05,
+      "loss": 0.2268,
+      "step": 21499
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3973276364331298e-05,
+      "loss": 0.3535,
+      "step": 21500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.396527735097891e-05,
+      "loss": 0.463,
+      "step": 21501
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3957280516018487e-05,
+      "loss": 0.689,
+      "step": 21502
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3949285859577996e-05,
+      "loss": 0.4683,
+      "step": 21503
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.39412933817855e-05,
+      "loss": 0.3761,
+      "step": 21504
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3933303082769048e-05,
+      "loss": 0.033,
+      "step": 21505
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3925314962656502e-05,
+      "loss": 0.001,
+      "step": 21506
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3917329021575829e-05,
+      "loss": 0.2507,
+      "step": 21507
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3909345259654891e-05,
+      "loss": 0.1884,
+      "step": 21508
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3901363677021553e-05,
+      "loss": 0.1711,
+      "step": 21509
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3893384273803615e-05,
+      "loss": 1.1166,
+      "step": 21510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3885407050128855e-05,
+      "loss": 0.1311,
+      "step": 21511
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3877432006125022e-05,
+      "loss": 0.2532,
+      "step": 21512
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3869459141919831e-05,
+      "loss": 0.1901,
+      "step": 21513
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3861488457640946e-05,
+      "loss": 0.5007,
+      "step": 21514
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3853519953416015e-05,
+      "loss": 0.3324,
+      "step": 21515
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3845553629372635e-05,
+      "loss": 0.0198,
+      "step": 21516
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3837589485638372e-05,
+      "loss": 0.444,
+      "step": 21517
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3829627522340774e-05,
+      "loss": 1.0271,
+      "step": 21518
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3821667739607289e-05,
+      "loss": 1.6609,
+      "step": 21519
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3813710137565447e-05,
+      "loss": 0.0016,
+      "step": 21520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3805754716342682e-05,
+      "loss": 0.0044,
+      "step": 21521
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3797801476066289e-05,
+      "loss": 2.4258,
+      "step": 21522
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3789850416863717e-05,
+      "loss": 0.4687,
+      "step": 21523
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3781901538862317e-05,
+      "loss": 0.0029,
+      "step": 21524
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.377395484218925e-05,
+      "loss": 0.1155,
+      "step": 21525
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3766010326971932e-05,
+      "loss": 0.03,
+      "step": 21526
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3758067993337446e-05,
+      "loss": 0.4591,
+      "step": 21527
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3750127841413022e-05,
+      "loss": 0.3191,
+      "step": 21528
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.374218987132586e-05,
+      "loss": 0.1593,
+      "step": 21529
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3734254083203022e-05,
+      "loss": 0.4734,
+      "step": 21530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.372632047717156e-05,
+      "loss": 0.7994,
+      "step": 21531
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3718389053358602e-05,
+      "loss": 0.2901,
+      "step": 21532
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.37104598118911e-05,
+      "loss": 0.6152,
+      "step": 21533
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3702532752896034e-05,
+      "loss": 0.4529,
+      "step": 21534
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3694607876500368e-05,
+      "loss": 0.5074,
+      "step": 21535
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.368668518283097e-05,
+      "loss": 0.0012,
+      "step": 21536
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3678764672014736e-05,
+      "loss": 0.7832,
+      "step": 21537
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3670846344178515e-05,
+      "loss": 0.251,
+      "step": 21538
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3662930199449074e-05,
+      "loss": 0.6188,
+      "step": 21539
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3655016237953192e-05,
+      "loss": 0.5668,
+      "step": 21540
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3647104459817603e-05,
+      "loss": 0.3497,
+      "step": 21541
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.363919486516899e-05,
+      "loss": 1.0895,
+      "step": 21542
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3631287454134032e-05,
+      "loss": 0.0004,
+      "step": 21543
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3623382226839347e-05,
+      "loss": 0.4556,
+      "step": 21544
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3615479183411532e-05,
+      "loss": 0.2448,
+      "step": 21545
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3607578323977104e-05,
+      "loss": 0.0214,
+      "step": 21546
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3599679648662626e-05,
+      "loss": 0.4447,
+      "step": 21547
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.35917831575946e-05,
+      "loss": 0.5649,
+      "step": 21548
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3583888850899388e-05,
+      "loss": 0.1824,
+      "step": 21549
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3575996728703492e-05,
+      "loss": 0.0005,
+      "step": 21550
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3568106791133294e-05,
+      "loss": 0.2592,
+      "step": 21551
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.356021903831509e-05,
+      "loss": 0.5973,
+      "step": 21552
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3552333470375183e-05,
+      "loss": 0.2847,
+      "step": 21553
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3544450087439918e-05,
+      "loss": 1.8066,
+      "step": 21554
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.353656888963548e-05,
+      "loss": 0.001,
+      "step": 21555
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3528689877088066e-05,
+      "loss": 0.1303,
+      "step": 21556
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3520813049923895e-05,
+      "loss": 0.5196,
+      "step": 21557
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3512938408269064e-05,
+      "loss": 0.0278,
+      "step": 21558
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3505065952249689e-05,
+      "loss": 0.3275,
+      "step": 21559
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.3497195681991835e-05,
+      "loss": 1.0687,
+      "step": 21560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.348932759762152e-05,
+      "loss": 0.6637,
+      "step": 21561
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3481461699264756e-05,
+      "loss": 1.2072,
+      "step": 21562
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3473597987047496e-05,
+      "loss": 0.4003,
+      "step": 21563
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3465736461095672e-05,
+      "loss": 0.1359,
+      "step": 21564
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3457877121535166e-05,
+      "loss": 0.4117,
+      "step": 21565
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3450019968491843e-05,
+      "loss": 0.5946,
+      "step": 21566
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3442165002091505e-05,
+      "loss": 0.9078,
+      "step": 21567
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3434312222459948e-05,
+      "loss": 0.204,
+      "step": 21568
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.342646162972294e-05,
+      "loss": 0.0315,
+      "step": 21569
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3418613224006197e-05,
+      "loss": 0.0005,
+      "step": 21570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3410767005435336e-05,
+      "loss": 0.2733,
+      "step": 21571
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.340292297413607e-05,
+      "loss": 0.3231,
+      "step": 21572
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.339508113023402e-05,
+      "loss": 0.1254,
+      "step": 21573
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3387241473854681e-05,
+      "loss": 0.125,
+      "step": 21574
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3379404005123689e-05,
+      "loss": 0.0019,
+      "step": 21575
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3371568724166476e-05,
+      "loss": 0.103,
+      "step": 21576
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3363735631108507e-05,
+      "loss": 0.0687,
+      "step": 21577
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3355904726075317e-05,
+      "loss": 0.1841,
+      "step": 21578
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3348076009192204e-05,
+      "loss": 0.1822,
+      "step": 21579
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3340249480584553e-05,
+      "loss": 1.6797,
+      "step": 21580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3332425140377745e-05,
+      "loss": 0.0013,
+      "step": 21581
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.332460298869703e-05,
+      "loss": 0.319,
+      "step": 21582
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.331678302566766e-05,
+      "loss": 0.2254,
+      "step": 21583
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3308965251414883e-05,
+      "loss": 0.3906,
+      "step": 21584
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3301149666063881e-05,
+      "loss": 0.317,
+      "step": 21585
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3293336269739824e-05,
+      "loss": 0.5156,
+      "step": 21586
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3285525062567808e-05,
+      "loss": 0.0007,
+      "step": 21587
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.327771604467292e-05,
+      "loss": 0.2998,
+      "step": 21588
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3269909216180224e-05,
+      "loss": 0.0066,
+      "step": 21589
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3262104577214738e-05,
+      "loss": 0.0017,
+      "step": 21590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3254302127901428e-05,
+      "loss": 0.245,
+      "step": 21591
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3246501868365245e-05,
+      "loss": 0.4366,
+      "step": 21592
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3238703798731087e-05,
+      "loss": 0.3856,
+      "step": 21593
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3230907919123874e-05,
+      "loss": 0.2556,
+      "step": 21594
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3223114229668353e-05,
+      "loss": 0.2173,
+      "step": 21595
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3215322730489408e-05,
+      "loss": 0.4554,
+      "step": 21596
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3207533421711808e-05,
+      "loss": 1.1349,
+      "step": 21597
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3199746303460218e-05,
+      "loss": 0.0015,
+      "step": 21598
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3191961375859405e-05,
+      "loss": 0.0966,
+      "step": 21599
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3184178639034037e-05,
+      "loss": 0.6127,
+      "step": 21600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3176398093108665e-05,
+      "loss": 0.3055,
+      "step": 21601
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3168619738207953e-05,
+      "loss": 0.2794,
+      "step": 21602
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3160843574456487e-05,
+      "loss": 0.0016,
+      "step": 21603
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.31530696019787e-05,
+      "loss": 0.3315,
+      "step": 21604
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3145297820899126e-05,
+      "loss": 0.0018,
+      "step": 21605
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3137528231342215e-05,
+      "loss": 0.93,
+      "step": 21606
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3129760833432402e-05,
+      "loss": 0.4427,
+      "step": 21607
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3121995627294051e-05,
+      "loss": 0.5648,
+      "step": 21608
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3114232613051501e-05,
+      "loss": 0.0431,
+      "step": 21609
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3106471790829099e-05,
+      "loss": 0.2496,
+      "step": 21610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3098713160751095e-05,
+      "loss": 0.7348,
+      "step": 21611
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3090956722941726e-05,
+      "loss": 0.0088,
+      "step": 21612
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3083202477525223e-05,
+      "loss": 0.5336,
+      "step": 21613
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3075450424625755e-05,
+      "loss": 0.5103,
+      "step": 21614
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3067700564367457e-05,
+      "loss": 0.3494,
+      "step": 21615
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.305995289687441e-05,
+      "loss": 0.0021,
+      "step": 21616
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3052207422270716e-05,
+      "loss": 0.0071,
+      "step": 21617
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3044464140680377e-05,
+      "loss": 0.3235,
+      "step": 21618
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3036723052227443e-05,
+      "loss": 0.4798,
+      "step": 21619
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3028984157035783e-05,
+      "loss": 0.0075,
+      "step": 21620
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.3021247455229394e-05,
+      "loss": 0.8455,
+      "step": 21621
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.301351294693218e-05,
+      "loss": 0.2202,
+      "step": 21622
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.300578063226791e-05,
+      "loss": 0.266,
+      "step": 21623
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2998050511360531e-05,
+      "loss": 0.0022,
+      "step": 21624
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2990322584333712e-05,
+      "loss": 0.0023,
+      "step": 21625
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2982596851311256e-05,
+      "loss": 0.6502,
+      "step": 21626
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2974873312416911e-05,
+      "loss": 0.0088,
+      "step": 21627
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2967151967774297e-05,
+      "loss": 0.0009,
+      "step": 21628
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2959432817507065e-05,
+      "loss": 0.3463,
+      "step": 21629
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2951715861738914e-05,
+      "loss": 0.3749,
+      "step": 21630
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2944001100593299e-05,
+      "loss": 0.4428,
+      "step": 21631
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2936288534193834e-05,
+      "loss": 2.0985,
+      "step": 21632
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2928578162664006e-05,
+      "loss": 0.6928,
+      "step": 21633
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.292086998612728e-05,
+      "loss": 0.5274,
+      "step": 21634
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2913164004707077e-05,
+      "loss": 0.3148,
+      "step": 21635
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2913164004707077e-05,
+      "loss": 1.3826,
+      "step": 21636
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.290546021852683e-05,
+      "loss": 0.9693,
+      "step": 21637
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2897758627709892e-05,
+      "loss": 0.007,
+      "step": 21638
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2890059232379579e-05,
+      "loss": 0.7194,
+      "step": 21639
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2882362032659193e-05,
+      "loss": 0.2391,
+      "step": 21640
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2874667028672003e-05,
+      "loss": 0.3469,
+      "step": 21641
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2866974220541209e-05,
+      "loss": 0.3521,
+      "step": 21642
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2859283608390031e-05,
+      "loss": 0.1943,
+      "step": 21643
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2851595192341635e-05,
+      "loss": 0.0169,
+      "step": 21644
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.284390897251904e-05,
+      "loss": 0.6669,
+      "step": 21645
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2836224949045431e-05,
+      "loss": 0.0006,
+      "step": 21646
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.282854312204386e-05,
+      "loss": 0.6904,
+      "step": 21647
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2820863491637262e-05,
+      "loss": 0.6315,
+      "step": 21648
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2813186057948671e-05,
+      "loss": 0.4904,
+      "step": 21649
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2805510821101039e-05,
+      "loss": 0.0063,
+      "step": 21650
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2797837781217202e-05,
+      "loss": 0.148,
+      "step": 21651
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.279016693842011e-05,
+      "loss": 0.3136,
+      "step": 21652
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.27824982928326e-05,
+      "loss": 0.0135,
+      "step": 21653
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2774831844577388e-05,
+      "loss": 1.5793,
+      "step": 21654
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.276716759377736e-05,
+      "loss": 0.8045,
+      "step": 21655
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2759505540555154e-05,
+      "loss": 0.155,
+      "step": 21656
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2751845685033485e-05,
+      "loss": 0.6562,
+      "step": 21657
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.274418802733504e-05,
+      "loss": 0.2428,
+      "step": 21658
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2736532567582436e-05,
+      "loss": 0.7828,
+      "step": 21659
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2728879305898243e-05,
+      "loss": 0.7169,
+      "step": 21660
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2721228242405045e-05,
+      "loss": 0.929,
+      "step": 21661
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2713579377225346e-05,
+      "loss": 0.0016,
+      "step": 21662
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2705932710481631e-05,
+      "loss": 1.0567,
+      "step": 21663
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2698288242296367e-05,
+      "loss": 0.2408,
+      "step": 21664
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.269064597279194e-05,
+      "loss": 0.2063,
+      "step": 21665
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2683005902090754e-05,
+      "loss": 0.5899,
+      "step": 21666
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2675368030315141e-05,
+      "loss": 0.219,
+      "step": 21667
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2667732357587424e-05,
+      "loss": 0.4046,
+      "step": 21668
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2660098884029851e-05,
+      "loss": 0.1492,
+      "step": 21669
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2652467609764693e-05,
+      "loss": 0.2821,
+      "step": 21670
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.264483853491412e-05,
+      "loss": 0.7441,
+      "step": 21671
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.263721165960035e-05,
+      "loss": 0.3815,
+      "step": 21672
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2629586983945434e-05,
+      "loss": 0.0044,
+      "step": 21673
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2621964508071575e-05,
+      "loss": 0.1612,
+      "step": 21674
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2614344232100742e-05,
+      "loss": 0.2828,
+      "step": 21675
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2606726156154973e-05,
+      "loss": 1.2764,
+      "step": 21676
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2599110280356333e-05,
+      "loss": 0.0055,
+      "step": 21677
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2591496604826712e-05,
+      "loss": 0.4667,
+      "step": 21678
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2583885129688026e-05,
+      "loss": 0.1909,
+      "step": 21679
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2576275855062245e-05,
+      "loss": 0.0026,
+      "step": 21680
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2568668781071139e-05,
+      "loss": 0.7731,
+      "step": 21681
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2561063907836545e-05,
+      "loss": 0.7728,
+      "step": 21682
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2553461235480227e-05,
+      "loss": 0.0066,
+      "step": 21683
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.254586076412396e-05,
+      "loss": 1.6651,
+      "step": 21684
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2538262493889446e-05,
+      "loss": 0.3207,
+      "step": 21685
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2530666424898366e-05,
+      "loss": 0.4207,
+      "step": 21686
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.252307255727233e-05,
+      "loss": 0.3912,
+      "step": 21687
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2515480891132983e-05,
+      "loss": 0.0093,
+      "step": 21688
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2507891426601869e-05,
+      "loss": 0.782,
+      "step": 21689
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2500304163800518e-05,
+      "loss": 0.247,
+      "step": 21690
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2492719102850435e-05,
+      "loss": 0.4118,
+      "step": 21691
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.248513624387309e-05,
+      "loss": 0.4838,
+      "step": 21692
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.247755558698995e-05,
+      "loss": 0.0071,
+      "step": 21693
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2469977132322289e-05,
+      "loss": 0.2993,
+      "step": 21694
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2462400879991575e-05,
+      "loss": 0.3699,
+      "step": 21695
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2454826830119125e-05,
+      "loss": 0.0043,
+      "step": 21696
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2447254982826143e-05,
+      "loss": 0.5946,
+      "step": 21697
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2439685338233967e-05,
+      "loss": 3.1171,
+      "step": 21698
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.24321178964638e-05,
+      "loss": 0.003,
+      "step": 21699
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.242455265763676e-05,
+      "loss": 0.005,
+      "step": 21700
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2416989621874051e-05,
+      "loss": 0.5737,
+      "step": 21701
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.240942878929681e-05,
+      "loss": 0.0012,
+      "step": 21702
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2401870160026006e-05,
+      "loss": 0.185,
+      "step": 21703
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2394313734182808e-05,
+      "loss": 0.5572,
+      "step": 21704
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2386759511888123e-05,
+      "loss": 0.8336,
+      "step": 21705
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2379207493262966e-05,
+      "loss": 0.6632,
+      "step": 21706
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2371657678428259e-05,
+      "loss": 0.2652,
+      "step": 21707
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2364110067504907e-05,
+      "loss": 0.2566,
+      "step": 21708
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.235656466061376e-05,
+      "loss": 2.0748,
+      "step": 21709
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2349021457875658e-05,
+      "loss": 0.4379,
+      "step": 21710
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2341480459411384e-05,
+      "loss": 0.3567,
+      "step": 21711
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2333941665341728e-05,
+      "loss": 0.4395,
+      "step": 21712
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2326405075787377e-05,
+      "loss": 0.1858,
+      "step": 21713
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2318870690869014e-05,
+      "loss": 0.0016,
+      "step": 21714
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.231133851070733e-05,
+      "loss": 0.4439,
+      "step": 21715
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2303808535422893e-05,
+      "loss": 0.0108,
+      "step": 21716
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2296280765136323e-05,
+      "loss": 0.0139,
+      "step": 21717
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2288755199968158e-05,
+      "loss": 0.0009,
+      "step": 21718
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2281231840038901e-05,
+      "loss": 0.0025,
+      "step": 21719
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2273710685469024e-05,
+      "loss": 0.9036,
+      "step": 21720
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2266191736378977e-05,
+      "loss": 0.3024,
+      "step": 21721
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.225867499288915e-05,
+      "loss": 0.0015,
+      "step": 21722
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2251160455119964e-05,
+      "loss": 0.3543,
+      "step": 21723
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2243648123191668e-05,
+      "loss": 0.3896,
+      "step": 21724
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2236137997224588e-05,
+      "loss": 0.2694,
+      "step": 21725
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2228630077339057e-05,
+      "loss": 0.5592,
+      "step": 21726
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2221124363655217e-05,
+      "loss": 0.3151,
+      "step": 21727
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2213620856293266e-05,
+      "loss": 0.0026,
+      "step": 21728
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2206119555373446e-05,
+      "loss": 0.0012,
+      "step": 21729
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2198620461015806e-05,
+      "loss": 0.8925,
+      "step": 21730
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2191123573340422e-05,
+      "loss": 0.4548,
+      "step": 21731
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2183628892467396e-05,
+      "loss": 0.2436,
+      "step": 21732
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2176136418516713e-05,
+      "loss": 0.301,
+      "step": 21733
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2168646151608347e-05,
+      "loss": 0.2631,
+      "step": 21734
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2161158091862283e-05,
+      "loss": 0.0009,
+      "step": 21735
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2153672239398377e-05,
+      "loss": 1.6175,
+      "step": 21736
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2146188594336547e-05,
+      "loss": 0.8563,
+      "step": 21737
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2138707156796617e-05,
+      "loss": 0.7368,
+      "step": 21738
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.213122792689839e-05,
+      "loss": 0.6823,
+      "step": 21739
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2123750904761636e-05,
+      "loss": 0.438,
+      "step": 21740
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2116276090506077e-05,
+      "loss": 0.2675,
+      "step": 21741
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2108803484251467e-05,
+      "loss": 0.2752,
+      "step": 21742
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.210133308611736e-05,
+      "loss": 0.2237,
+      "step": 21743
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2093864896223477e-05,
+      "loss": 1.1387,
+      "step": 21744
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2086398914689405e-05,
+      "loss": 0.0024,
+      "step": 21745
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2078935141634617e-05,
+      "loss": 0.2193,
+      "step": 21746
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.207147357717873e-05,
+      "loss": 0.8985,
+      "step": 21747
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2064014221441204e-05,
+      "loss": 0.0027,
+      "step": 21748
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.205655707454144e-05,
+      "loss": 0.0249,
+      "step": 21749
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2049102136598893e-05,
+      "loss": 0.2576,
+      "step": 21750
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2041649407732984e-05,
+      "loss": 0.1152,
+      "step": 21751
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2034198888062968e-05,
+      "loss": 0.3767,
+      "step": 21752
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2026750577708233e-05,
+      "loss": 0.3536,
+      "step": 21753
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2019304476788e-05,
+      "loss": 0.2209,
+      "step": 21754
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2011860585421507e-05,
+      "loss": 0.2512,
+      "step": 21755
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.2004418903728024e-05,
+      "loss": 0.3422,
+      "step": 21756
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1996979431826642e-05,
+      "loss": 0.8057,
+      "step": 21757
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1989542169836514e-05,
+      "loss": 2.1783,
+      "step": 21758
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1982107117876743e-05,
+      "loss": 0.2114,
+      "step": 21759
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1974674276066404e-05,
+      "loss": 0.1595,
+      "step": 21760
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1967243644524499e-05,
+      "loss": 0.3096,
+      "step": 21761
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1959815223370033e-05,
+      "loss": 0.5311,
+      "step": 21762
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1952389012721963e-05,
+      "loss": 0.7538,
+      "step": 21763
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1944965012699193e-05,
+      "loss": 0.2462,
+      "step": 21764
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1937543223420627e-05,
+      "loss": 0.5106,
+      "step": 21765
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1930123645005085e-05,
+      "loss": 0.0114,
+      "step": 21766
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.192270627757141e-05,
+      "loss": 0.0025,
+      "step": 21767
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1915291121238368e-05,
+      "loss": 0.0022,
+      "step": 21768
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1907878176124702e-05,
+      "loss": 4.4749,
+      "step": 21769
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1900467442349132e-05,
+      "loss": 0.4405,
+      "step": 21770
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1893058920030296e-05,
+      "loss": 0.0991,
+      "step": 21771
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.18856526092869e-05,
+      "loss": 0.2516,
+      "step": 21772
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1878248510237449e-05,
+      "loss": 1.9424,
+      "step": 21773
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1870846623000546e-05,
+      "loss": 0.4079,
+      "step": 21774
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1863446947694783e-05,
+      "loss": 0.305,
+      "step": 21775
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1856049484438578e-05,
+      "loss": 0.3815,
+      "step": 21776
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1848654233350374e-05,
+      "loss": 0.5751,
+      "step": 21777
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1841261194548707e-05,
+      "loss": 0.39,
+      "step": 21778
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1833870368151849e-05,
+      "loss": 0.7999,
+      "step": 21779
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1826481754278222e-05,
+      "loss": 0.4395,
+      "step": 21780
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1819095353046099e-05,
+      "loss": 0.7234,
+      "step": 21781
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.18117111645738e-05,
+      "loss": 0.3754,
+      "step": 21782
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.180432918897955e-05,
+      "loss": 0.1542,
+      "step": 21783
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1796949426381552e-05,
+      "loss": 0.5379,
+      "step": 21784
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1789571876898013e-05,
+      "loss": 0.7102,
+      "step": 21785
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1782196540647054e-05,
+      "loss": 0.6958,
+      "step": 21786
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1774823417746765e-05,
+      "loss": 0.8654,
+      "step": 21787
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.176745250831525e-05,
+      "loss": 0.9913,
+      "step": 21788
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1760083812470516e-05,
+      "loss": 0.4425,
+      "step": 21789
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1752717330330569e-05,
+      "loss": 0.4396,
+      "step": 21790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1745353062013412e-05,
+      "loss": 0.769,
+      "step": 21791
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1737991007636872e-05,
+      "loss": 1.1802,
+      "step": 21792
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1730631167318932e-05,
+      "loss": 0.5538,
+      "step": 21793
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1723273541177452e-05,
+      "loss": 0.5263,
+      "step": 21794
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.171591812933017e-05,
+      "loss": 0.5509,
+      "step": 21795
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1708564931894942e-05,
+      "loss": 0.1049,
+      "step": 21796
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1701213948989523e-05,
+      "loss": 0.2981,
+      "step": 21797
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1693865180731555e-05,
+      "loss": 0.4731,
+      "step": 21798
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1686518627238805e-05,
+      "loss": 0.7844,
+      "step": 21799
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1679174288628901e-05,
+      "loss": 0.6542,
+      "step": 21800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1671832165019379e-05,
+      "loss": 0.6505,
+      "step": 21801
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1664492256527914e-05,
+      "loss": 0.2799,
+      "step": 21802
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1657154563271958e-05,
+      "loss": 0.3723,
+      "step": 21803
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1649819085369022e-05,
+      "loss": 0.4375,
+      "step": 21804
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1642485822936659e-05,
+      "loss": 0.1676,
+      "step": 21805
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1635154776092226e-05,
+      "loss": 0.4895,
+      "step": 21806
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1627825944953096e-05,
+      "loss": 0.0036,
+      "step": 21807
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1620499329636707e-05,
+      "loss": 0.6084,
+      "step": 21808
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1613174930260333e-05,
+      "loss": 1.2718,
+      "step": 21809
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1605852746941263e-05,
+      "loss": 0.2404,
+      "step": 21810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1598532779796754e-05,
+      "loss": 0.4166,
+      "step": 21811
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1591215028944045e-05,
+      "loss": 0.4831,
+      "step": 21812
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1583899494500293e-05,
+      "loss": 0.3645,
+      "step": 21813
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.157658617658267e-05,
+      "loss": 2.1633,
+      "step": 21814
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1569275075308266e-05,
+      "loss": 2.1881,
+      "step": 21815
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.156196619079417e-05,
+      "loss": 0.23,
+      "step": 21816
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1554659523157422e-05,
+      "loss": 1.0332,
+      "step": 21817
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.154735507251503e-05,
+      "loss": 0.0125,
+      "step": 21818
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1540052838983949e-05,
+      "loss": 0.5015,
+      "step": 21819
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1532752822681119e-05,
+      "loss": 0.0552,
+      "step": 21820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1525455023723462e-05,
+      "loss": 0.4761,
+      "step": 21821
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1518159442227786e-05,
+      "loss": 2.5978,
+      "step": 21822
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1510866078310965e-05,
+      "loss": 0.2739,
+      "step": 21823
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.150357493208982e-05,
+      "loss": 0.156,
+      "step": 21824
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1496286003681043e-05,
+      "loss": 0.2589,
+      "step": 21825
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.148899929320134e-05,
+      "loss": 0.6192,
+      "step": 21826
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.14817148007675e-05,
+      "loss": 0.3021,
+      "step": 21827
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1474432526496097e-05,
+      "loss": 0.0007,
+      "step": 21828
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1467152470503737e-05,
+      "loss": 0.2754,
+      "step": 21829
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1459874632907029e-05,
+      "loss": 0.201,
+      "step": 21830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.145259901382251e-05,
+      "loss": 0.2597,
+      "step": 21831
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.144532561336669e-05,
+      "loss": 0.2056,
+      "step": 21832
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.143805443165604e-05,
+      "loss": 0.7612,
+      "step": 21833
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1430785468806986e-05,
+      "loss": 0.2334,
+      "step": 21834
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1423518724935949e-05,
+      "loss": 0.0007,
+      "step": 21835
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1416254200159286e-05,
+      "loss": 0.4955,
+      "step": 21836
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1408991894593323e-05,
+      "loss": 0.0013,
+      "step": 21837
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1401731808354365e-05,
+      "loss": 0.443,
+      "step": 21838
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1394473941558651e-05,
+      "loss": 0.6335,
+      "step": 21839
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.138721829432246e-05,
+      "loss": 0.2993,
+      "step": 21840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1379964866761875e-05,
+      "loss": 0.4242,
+      "step": 21841
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1372713658993143e-05,
+      "loss": 0.4753,
+      "step": 21842
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1365464671132385e-05,
+      "loss": 1.2628,
+      "step": 21843
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1358217903295592e-05,
+      "loss": 0.6885,
+      "step": 21844
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1350973355598886e-05,
+      "loss": 0.502,
+      "step": 21845
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1343731028158293e-05,
+      "loss": 0.1792,
+      "step": 21846
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1336490921089702e-05,
+      "loss": 0.2419,
+      "step": 21847
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1329253034509156e-05,
+      "loss": 0.6778,
+      "step": 21848
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1322017368532476e-05,
+      "loss": 0.6395,
+      "step": 21849
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1314783923275539e-05,
+      "loss": 0.8848,
+      "step": 21850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1307552698854249e-05,
+      "loss": 2.2838,
+      "step": 21851
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1300323695384333e-05,
+      "loss": 0.3169,
+      "step": 21852
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1293096912981548e-05,
+      "loss": 0.2785,
+      "step": 21853
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1285872351761699e-05,
+      "loss": 0.0016,
+      "step": 21854
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1278650011840395e-05,
+      "loss": 0.2359,
+      "step": 21855
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1271429893333279e-05,
+      "loss": 0.8362,
+      "step": 21856
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1264211996356087e-05,
+      "loss": 0.5336,
+      "step": 21857
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.125699632102428e-05,
+      "loss": 0.7024,
+      "step": 21858
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1249782867453466e-05,
+      "loss": 0.5367,
+      "step": 21859
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1242571635759151e-05,
+      "loss": 0.9254,
+      "step": 21860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1235362626056793e-05,
+      "loss": 0.4156,
+      "step": 21861
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.122815583846185e-05,
+      "loss": 0.0006,
+      "step": 21862
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1220951273089713e-05,
+      "loss": 0.3194,
+      "step": 21863
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1213748930055771e-05,
+      "loss": 1.2473,
+      "step": 21864
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1206548809475352e-05,
+      "loss": 0.7381,
+      "step": 21865
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1199350911463762e-05,
+      "loss": 0.0056,
+      "step": 21866
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1192155236136258e-05,
+      "loss": 0.3786,
+      "step": 21867
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1184961783608049e-05,
+      "loss": 0.3953,
+      "step": 21868
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1177770553994358e-05,
+      "loss": 1.0327,
+      "step": 21869
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1170581547410345e-05,
+      "loss": 0.2071,
+      "step": 21870
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1163394763971068e-05,
+      "loss": 0.0018,
+      "step": 21871
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1156210203791683e-05,
+      "loss": 0.1107,
+      "step": 21872
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1149027866987248e-05,
+      "loss": 0.3344,
+      "step": 21873
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1141847753672723e-05,
+      "loss": 0.5478,
+      "step": 21874
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1134669863963065e-05,
+      "loss": 2.031,
+      "step": 21875
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.112749419797333e-05,
+      "loss": 0.0006,
+      "step": 21876
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.112032075581833e-05,
+      "loss": 0.0031,
+      "step": 21877
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1113149537612953e-05,
+      "loss": 0.3726,
+      "step": 21878
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.110598054347206e-05,
+      "loss": 0.0046,
+      "step": 21879
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1098813773510423e-05,
+      "loss": 1.4476,
+      "step": 21880
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1091649227842837e-05,
+      "loss": 0.6407,
+      "step": 21881
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1084486906584006e-05,
+      "loss": 3.379,
+      "step": 21882
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.107732680984864e-05,
+      "loss": 0.6949,
+      "step": 21883
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1070168937751384e-05,
+      "loss": 0.5067,
+      "step": 21884
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1063013290406875e-05,
+      "loss": 0.656,
+      "step": 21885
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1055859867929689e-05,
+      "loss": 0.431,
+      "step": 21886
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1048708670434386e-05,
+      "loss": 0.3293,
+      "step": 21887
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1041559698035473e-05,
+      "loss": 2.3376,
+      "step": 21888
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1034412950847426e-05,
+      "loss": 0.0017,
+      "step": 21889
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.102726842898472e-05,
+      "loss": 0.8569,
+      "step": 21890
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1020126132561713e-05,
+      "loss": 0.811,
+      "step": 21891
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1012986061692846e-05,
+      "loss": 0.4597,
+      "step": 21892
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.100584821649238e-05,
+      "loss": 0.2635,
+      "step": 21893
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0998712597074654e-05,
+      "loss": 0.251,
+      "step": 21894
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.099157920355398e-05,
+      "loss": 0.0019,
+      "step": 21895
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.098444803604448e-05,
+      "loss": 0.2973,
+      "step": 21896
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0977319094660465e-05,
+      "loss": 0.825,
+      "step": 21897
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0970192379516007e-05,
+      "loss": 0.1361,
+      "step": 21898
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0963067890725253e-05,
+      "loss": 0.8682,
+      "step": 21899
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.095594562840234e-05,
+      "loss": 0.0004,
+      "step": 21900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0948825592661265e-05,
+      "loss": 0.4939,
+      "step": 21901
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0941707783616032e-05,
+      "loss": 0.0016,
+      "step": 21902
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0934592201380721e-05,
+      "loss": 0.2209,
+      "step": 21903
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0927478846069154e-05,
+      "loss": 0.259,
+      "step": 21904
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0920367717795292e-05,
+      "loss": 0.3174,
+      "step": 21905
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0913258816673043e-05,
+      "loss": 0.0057,
+      "step": 21906
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0906152142816199e-05,
+      "loss": 0.41,
+      "step": 21907
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.089904769633857e-05,
+      "loss": 0.9986,
+      "step": 21908
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0891945477353947e-05,
+      "loss": 0.0022,
+      "step": 21909
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0884845485976023e-05,
+      "loss": 0.8604,
+      "step": 21910
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0877747722318541e-05,
+      "loss": 0.192,
+      "step": 21911
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.087065218649511e-05,
+      "loss": 0.0009,
+      "step": 21912
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0863558878619404e-05,
+      "loss": 0.3314,
+      "step": 21913
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0856467798804968e-05,
+      "loss": 0.7475,
+      "step": 21914
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0849378947165394e-05,
+      "loss": 0.0238,
+      "step": 21915
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0842292323814173e-05,
+      "loss": 0.0007,
+      "step": 21916
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0835207928864782e-05,
+      "loss": 0.0011,
+      "step": 21917
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0828125762430696e-05,
+      "loss": 0.2108,
+      "step": 21918
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0821045824625324e-05,
+      "loss": 0.0006,
+      "step": 21919
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0813968115561976e-05,
+      "loss": 0.4994,
+      "step": 21920
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0806892635354063e-05,
+      "loss": 0.0033,
+      "step": 21921
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0799819384114894e-05,
+      "loss": 0.6673,
+      "step": 21922
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.079274836195766e-05,
+      "loss": 0.5732,
+      "step": 21923
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0785679568995653e-05,
+      "loss": 1.0332,
+      "step": 21924
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0778613005342085e-05,
+      "loss": 0.8055,
+      "step": 21925
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0771548671110064e-05,
+      "loss": 0.2442,
+      "step": 21926
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.076448656641275e-05,
+      "loss": 0.5745,
+      "step": 21927
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0757426691363202e-05,
+      "loss": 0.2374,
+      "step": 21928
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0750369046074515e-05,
+      "loss": 0.0573,
+      "step": 21929
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0743313630659661e-05,
+      "loss": 0.7231,
+      "step": 21930
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0736260445231654e-05,
+      "loss": 0.0672,
+      "step": 21931
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0729209489903418e-05,
+      "loss": 0.0006,
+      "step": 21932
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0722160764787896e-05,
+      "loss": 0.6586,
+      "step": 21933
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0715114269997932e-05,
+      "loss": 0.2359,
+      "step": 21934
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0708070005646369e-05,
+      "loss": 0.8507,
+      "step": 21935
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0701027971846032e-05,
+      "loss": 0.0807,
+      "step": 21936
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0693988168709666e-05,
+      "loss": 0.3629,
+      "step": 21937
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0686950596350014e-05,
+      "loss": 0.0006,
+      "step": 21938
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0679915254879768e-05,
+      "loss": 0.512,
+      "step": 21939
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0672882144411587e-05,
+      "loss": 0.5321,
+      "step": 21940
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0665851265058117e-05,
+      "loss": 0.2221,
+      "step": 21941
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0658822616931884e-05,
+      "loss": 0.238,
+      "step": 21942
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0651796200145513e-05,
+      "loss": 0.9662,
+      "step": 21943
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0644772014811498e-05,
+      "loss": 0.0009,
+      "step": 21944
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0637750061042283e-05,
+      "loss": 0.4991,
+      "step": 21945
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.063073033895041e-05,
+      "loss": 0.0019,
+      "step": 21946
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0623712848648175e-05,
+      "loss": 0.689,
+      "step": 21947
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0616697590248002e-05,
+      "loss": 0.3442,
+      "step": 21948
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.060968456386227e-05,
+      "loss": 0.0296,
+      "step": 21949
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.060267376960322e-05,
+      "loss": 0.0165,
+      "step": 21950
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0595665207583115e-05,
+      "loss": 0.3446,
+      "step": 21951
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.058865887791428e-05,
+      "loss": 0.6835,
+      "step": 21952
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0581654780708809e-05,
+      "loss": 0.397,
+      "step": 21953
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.057465291607888e-05,
+      "loss": 0.393,
+      "step": 21954
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0567653284136684e-05,
+      "loss": 0.5563,
+      "step": 21955
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0560655884994235e-05,
+      "loss": 0.0267,
+      "step": 21956
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0553660718763606e-05,
+      "loss": 0.7532,
+      "step": 21957
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0546667785556829e-05,
+      "loss": 0.1923,
+      "step": 21958
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0539677085485876e-05,
+      "loss": 0.3818,
+      "step": 21959
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0532688618662694e-05,
+      "loss": 0.3343,
+      "step": 21960
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0525702385199192e-05,
+      "loss": 2.7363,
+      "step": 21961
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0518718385207248e-05,
+      "loss": 1.0056,
+      "step": 21962
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0511736618798705e-05,
+      "loss": 0.0169,
+      "step": 21963
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0504757086085341e-05,
+      "loss": 0.0006,
+      "step": 21964
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.049777978717895e-05,
+      "loss": 0.0192,
+      "step": 21965
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0490804722191243e-05,
+      "loss": 0.4699,
+      "step": 21966
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.048383189123393e-05,
+      "loss": 0.293,
+      "step": 21967
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0476861294418688e-05,
+      "loss": 0.0005,
+      "step": 21968
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0469892931857077e-05,
+      "loss": 0.355,
+      "step": 21969
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0462926803660726e-05,
+      "loss": 0.4881,
+      "step": 21970
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.045596290994123e-05,
+      "loss": 0.0006,
+      "step": 21971
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0449001250810014e-05,
+      "loss": 0.7891,
+      "step": 21972
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0442041826378623e-05,
+      "loss": 0.3147,
+      "step": 21973
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0435084636758518e-05,
+      "loss": 0.2742,
+      "step": 21974
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0428129682061026e-05,
+      "loss": 0.3136,
+      "step": 21975
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0421176962397609e-05,
+      "loss": 1.6953,
+      "step": 21976
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0414226477879545e-05,
+      "loss": 0.1344,
+      "step": 21977
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0407278228618144e-05,
+      "loss": 0.0009,
+      "step": 21978
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0400332214724699e-05,
+      "loss": 0.1104,
+      "step": 21979
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0393388436310423e-05,
+      "loss": 0.598,
+      "step": 21980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0386446893486494e-05,
+      "loss": 0.0049,
+      "step": 21981
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0379507586364106e-05,
+      "loss": 0.4336,
+      "step": 21982
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0372570515054351e-05,
+      "loss": 0.0029,
+      "step": 21983
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0365635679668327e-05,
+      "loss": 0.7178,
+      "step": 21984
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0358703080317093e-05,
+      "loss": 0.8399,
+      "step": 21985
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0351772717111662e-05,
+      "loss": 3.616,
+      "step": 21986
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0344844590162993e-05,
+      "loss": 0.1708,
+      "step": 21987
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0337918699582066e-05,
+      "loss": 0.387,
+      "step": 21988
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0330995045479756e-05,
+      "loss": 0.132,
+      "step": 21989
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0324073627966945e-05,
+      "loss": 0.2607,
+      "step": 21990
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0317154447154457e-05,
+      "loss": 0.3436,
+      "step": 21991
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0310237503153123e-05,
+      "loss": 0.0034,
+      "step": 21992
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0303322796073721e-05,
+      "loss": 0.9812,
+      "step": 21993
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0296410326026894e-05,
+      "loss": 0.2085,
+      "step": 21994
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0289500093123421e-05,
+      "loss": 0.0007,
+      "step": 21995
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.028259209747393e-05,
+      "loss": 0.0026,
+      "step": 21996
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0275686339189e-05,
+      "loss": 0.5876,
+      "step": 21997
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0268782818379306e-05,
+      "loss": 0.3746,
+      "step": 21998
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.026188153515533e-05,
+      "loss": 0.3046,
+      "step": 21999
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0254982489627583e-05,
+      "loss": 0.2503,
+      "step": 22000
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0248085681906593e-05,
+      "loss": 0.011,
+      "step": 22001
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.024119111210277e-05,
+      "loss": 0.3688,
+      "step": 22002
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0234298780326478e-05,
+      "loss": 0.4279,
+      "step": 22003
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0227408686688194e-05,
+      "loss": 0.3218,
+      "step": 22004
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0220520831298163e-05,
+      "loss": 0.629,
+      "step": 22005
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0213635214266714e-05,
+      "loss": 0.2665,
+      "step": 22006
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0206751835704108e-05,
+      "loss": 0.557,
+      "step": 22007
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0199870695720574e-05,
+      "loss": 1.2526,
+      "step": 22008
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0192991794426292e-05,
+      "loss": 1.1025,
+      "step": 22009
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.018611513193142e-05,
+      "loss": 0.5424,
+      "step": 22010
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0179240708346093e-05,
+      "loss": 1.112,
+      "step": 22011
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0172368523780384e-05,
+      "loss": 0.4749,
+      "step": 22012
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.016549857834434e-05,
+      "loss": 0.1253,
+      "step": 22013
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0158630872147972e-05,
+      "loss": 0.8293,
+      "step": 22014
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.015176540530126e-05,
+      "loss": 0.0029,
+      "step": 22015
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.014490217791415e-05,
+      "loss": 0.0011,
+      "step": 22016
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0138041190096552e-05,
+      "loss": 0.2467,
+      "step": 22017
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.013118244195828e-05,
+      "loss": 0.8996,
+      "step": 22018
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0124325933609229e-05,
+      "loss": 1.5305,
+      "step": 22019
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0117471665159193e-05,
+      "loss": 0.3844,
+      "step": 22020
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.011061963671787e-05,
+      "loss": 0.8096,
+      "step": 22021
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0103769848395054e-05,
+      "loss": 1.8246,
+      "step": 22022
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0096922300300425e-05,
+      "loss": 0.6719,
+      "step": 22023
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0090076992543578e-05,
+      "loss": 0.1533,
+      "step": 22024
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0083233925234225e-05,
+      "loss": 0.005,
+      "step": 22025
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0076393098481861e-05,
+      "loss": 0.3919,
+      "step": 22026
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0069554512396066e-05,
+      "loss": 0.2097,
+      "step": 22027
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.006271816708637e-05,
+      "loss": 1.1452,
+      "step": 22028
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.00558840626622e-05,
+      "loss": 0.0034,
+      "step": 22029
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0049052199233037e-05,
+      "loss": 0.0013,
+      "step": 22030
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.004222257690826e-05,
+      "loss": 0.1554,
+      "step": 22031
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0035395195797246e-05,
+      "loss": 0.1367,
+      "step": 22032
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.002857005600931e-05,
+      "loss": 0.6101,
+      "step": 22033
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0021747157653765e-05,
+      "loss": 0.3212,
+      "step": 22034
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0014926500839854e-05,
+      "loss": 0.0281,
+      "step": 22035
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0008108085676808e-05,
+      "loss": 0.3794,
+      "step": 22036
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.0001291912273823e-05,
+      "loss": 0.8637,
+      "step": 22037
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.994477980740028e-06,
+      "loss": 0.0084,
+      "step": 22038
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.987666291184553e-06,
+      "loss": 1.3714,
+      "step": 22039
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.980856843716479e-06,
+      "loss": 0.3662,
+      "step": 22040
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.974049638444832e-06,
+      "loss": 0.8353,
+      "step": 22041
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.967244675478642e-06,
+      "loss": 0.394,
+      "step": 22042
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.960441954926873e-06,
+      "loss": 0.6152,
+      "step": 22043
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.953641476898472e-06,
+      "loss": 0.1718,
+      "step": 22044
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.946843241502284e-06,
+      "loss": 0.1023,
+      "step": 22045
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.940047248847206e-06,
+      "loss": 0.012,
+      "step": 22046
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.933253499042116e-06,
+      "loss": 0.3676,
+      "step": 22047
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.926461992195728e-06,
+      "loss": 0.002,
+      "step": 22048
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.919672728416823e-06,
+      "loss": 0.7416,
+      "step": 22049
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.912885707814162e-06,
+      "loss": 0.5505,
+      "step": 22050
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.906100930496375e-06,
+      "loss": 2.7527,
+      "step": 22051
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.89931839657211e-06,
+      "loss": 0.3994,
+      "step": 22052
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.892538106150028e-06,
+      "loss": 0.4518,
+      "step": 22053
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.885760059338676e-06,
+      "loss": 0.7313,
+      "step": 22054
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.878984256246587e-06,
+      "loss": 2.5488,
+      "step": 22055
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.87221069698227e-06,
+      "loss": 1.7956,
+      "step": 22056
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.865439381654205e-06,
+      "loss": 0.0143,
+      "step": 22057
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.85867031037081e-06,
+      "loss": 0.2587,
+      "step": 22058
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.851903483240475e-06,
+      "loss": 0.0027,
+      "step": 22059
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.845138900371585e-06,
+      "loss": 0.0043,
+      "step": 22060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.838376561872451e-06,
+      "loss": 0.4209,
+      "step": 22061
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.831616467851372e-06,
+      "loss": 0.3481,
+      "step": 22062
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.824858618416576e-06,
+      "loss": 0.3257,
+      "step": 22063
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.818103013676309e-06,
+      "loss": 0.0498,
+      "step": 22064
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.811349653738737e-06,
+      "loss": 0.031,
+      "step": 22065
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.804598538712022e-06,
+      "loss": 0.2445,
+      "step": 22066
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.797849668704211e-06,
+      "loss": 0.4684,
+      "step": 22067
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.79110304382345e-06,
+      "loss": 0.4652,
+      "step": 22068
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.784358664177771e-06,
+      "loss": 0.003,
+      "step": 22069
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.777616529875121e-06,
+      "loss": 0.087,
+      "step": 22070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.770876641023512e-06,
+      "loss": 0.4126,
+      "step": 22071
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.764138997730875e-06,
+      "loss": 0.7718,
+      "step": 22072
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.757403600105041e-06,
+      "loss": 0.6533,
+      "step": 22073
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.750670448253972e-06,
+      "loss": 0.8442,
+      "step": 22074
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.7439395422854e-06,
+      "loss": 1.4307,
+      "step": 22075
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.737210882307122e-06,
+      "loss": 0.9089,
+      "step": 22076
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.730484468426952e-06,
+      "loss": 0.266,
+      "step": 22077
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.723760300752537e-06,
+      "loss": 2.6692,
+      "step": 22078
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.717038379391572e-06,
+      "loss": 0.3902,
+      "step": 22079
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.710318704451691e-06,
+      "loss": 0.0037,
+      "step": 22080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.703601276040523e-06,
+      "loss": 0.404,
+      "step": 22081
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.696886094265633e-06,
+      "loss": 0.3105,
+      "step": 22082
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.690173159234532e-06,
+      "loss": 0.9781,
+      "step": 22083
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.683462471054737e-06,
+      "loss": 0.0015,
+      "step": 22084
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.676754029833711e-06,
+      "loss": 0.2318,
+      "step": 22085
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.670047835678868e-06,
+      "loss": 0.0014,
+      "step": 22086
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.663343888697606e-06,
+      "loss": 0.4417,
+      "step": 22087
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.65664218899727e-06,
+      "loss": 0.6299,
+      "step": 22088
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.649942736685195e-06,
+      "loss": 0.5008,
+      "step": 22089
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.643245531868643e-06,
+      "loss": 0.5962,
+      "step": 22090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.636550574654861e-06,
+      "loss": 0.3224,
+      "step": 22091
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.62985786515108e-06,
+      "loss": 0.0038,
+      "step": 22092
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.623167403464465e-06,
+      "loss": 0.2803,
+      "step": 22093
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.616479189702147e-06,
+      "loss": 0.3034,
+      "step": 22094
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.60979322397119e-06,
+      "loss": 1.0479,
+      "step": 22095
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.603109506378741e-06,
+      "loss": 0.417,
+      "step": 22096
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.596428037031767e-06,
+      "loss": 0.7819,
+      "step": 22097
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.589748816037263e-06,
+      "loss": 0.3173,
+      "step": 22098
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.58307184350226e-06,
+      "loss": 0.3772,
+      "step": 22099
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.576397119533592e-06,
+      "loss": 0.3041,
+      "step": 22100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.569724644238152e-06,
+      "loss": 0.5582,
+      "step": 22101
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.563054417722859e-06,
+      "loss": 0.352,
+      "step": 22102
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.556386440094476e-06,
+      "loss": 1.2403,
+      "step": 22103
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.549720711459785e-06,
+      "loss": 0.4481,
+      "step": 22104
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.543057231925516e-06,
+      "loss": 0.1374,
+      "step": 22105
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.5363960015984e-06,
+      "loss": 0.8526,
+      "step": 22106
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.529737020585087e-06,
+      "loss": 0.5579,
+      "step": 22107
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.523080288992225e-06,
+      "loss": 0.3893,
+      "step": 22108
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.516425806926409e-06,
+      "loss": 0.2114,
+      "step": 22109
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.509773574494189e-06,
+      "loss": 0.224,
+      "step": 22110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.503123591802115e-06,
+      "loss": 0.6238,
+      "step": 22111
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.496475858956648e-06,
+      "loss": 0.4507,
+      "step": 22112
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.489830376064255e-06,
+      "loss": 0.6371,
+      "step": 22113
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.483187143231352e-06,
+      "loss": 0.3003,
+      "step": 22114
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.476546160564335e-06,
+      "loss": 0.3218,
+      "step": 22115
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.469907428169505e-06,
+      "loss": 0.2478,
+      "step": 22116
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.463270946153223e-06,
+      "loss": 0.2678,
+      "step": 22117
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.456636714621757e-06,
+      "loss": 0.748,
+      "step": 22118
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.45000473368127e-06,
+      "loss": 0.7781,
+      "step": 22119
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.443375003438064e-06,
+      "loss": 0.3583,
+      "step": 22120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.436747523998284e-06,
+      "loss": 0.7478,
+      "step": 22121
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.430122295467979e-06,
+      "loss": 0.5087,
+      "step": 22122
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.423499317953349e-06,
+      "loss": 0.0004,
+      "step": 22123
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.41687859156039e-06,
+      "loss": 0.3041,
+      "step": 22124
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.410260116395101e-06,
+      "loss": 0.6524,
+      "step": 22125
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.403643892563533e-06,
+      "loss": 0.0089,
+      "step": 22126
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.3970299201716e-06,
+      "loss": 0.3483,
+      "step": 22127
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.390418199325184e-06,
+      "loss": 3.4547,
+      "step": 22128
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.383808730130249e-06,
+      "loss": 0.4665,
+      "step": 22129
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.377201512692561e-06,
+      "loss": 0.001,
+      "step": 22130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.370596547117937e-06,
+      "loss": 0.0015,
+      "step": 22131
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.363993833512156e-06,
+      "loss": 0.0803,
+      "step": 22132
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.357393371980953e-06,
+      "loss": 0.6465,
+      "step": 22133
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.350795162630009e-06,
+      "loss": 0.4753,
+      "step": 22134
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.344199205565006e-06,
+      "loss": 0.4269,
+      "step": 22135
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.33760550089156e-06,
+      "loss": 0.2983,
+      "step": 22136
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.331014048715253e-06,
+      "loss": 0.0006,
+      "step": 22137
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.324424849141648e-06,
+      "loss": 2.3242,
+      "step": 22138
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.317837902276248e-06,
+      "loss": 2.0293,
+      "step": 22139
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.311253208224534e-06,
+      "loss": 0.3771,
+      "step": 22140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.304670767091953e-06,
+      "loss": 0.6047,
+      "step": 22141
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.298090578983958e-06,
+      "loss": 0.367,
+      "step": 22142
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.291512644005811e-06,
+      "loss": 0.6377,
+      "step": 22143
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.284936962262945e-06,
+      "loss": 0.2294,
+      "step": 22144
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.278363533860661e-06,
+      "loss": 0.9866,
+      "step": 22145
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.271792358904173e-06,
+      "loss": 0.7957,
+      "step": 22146
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.26522343749868e-06,
+      "loss": 1.7845,
+      "step": 22147
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.258656769749485e-06,
+      "loss": 0.9254,
+      "step": 22148
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.252092355761648e-06,
+      "loss": 1.8389,
+      "step": 22149
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.245530195640305e-06,
+      "loss": 1.4867,
+      "step": 22150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.238970289490588e-06,
+      "loss": 0.2097,
+      "step": 22151
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.232412637417497e-06,
+      "loss": 0.581,
+      "step": 22152
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.225857239526047e-06,
+      "loss": 0.271,
+      "step": 22153
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.21930409592122e-06,
+      "loss": 0.5181,
+      "step": 22154
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.212753206707968e-06,
+      "loss": 0.0379,
+      "step": 22155
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.20620457199117e-06,
+      "loss": 0.056,
+      "step": 22156
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.199658191875698e-06,
+      "loss": 0.4096,
+      "step": 22157
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.193114066466394e-06,
+      "loss": 0.9331,
+      "step": 22158
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.186572195868047e-06,
+      "loss": 0.5034,
+      "step": 22159
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.18003258018542e-06,
+      "loss": 0.7827,
+      "step": 22160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.173495219523231e-06,
+      "loss": 1.4247,
+      "step": 22161
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.16696011398616e-06,
+      "loss": 0.1966,
+      "step": 22162
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.16042726367886e-06,
+      "loss": 4.198,
+      "step": 22163
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.153896668705979e-06,
+      "loss": 0.4152,
+      "step": 22164
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.147368329172e-06,
+      "loss": 0.4337,
+      "step": 22165
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.140842245181573e-06,
+      "loss": 0.744,
+      "step": 22166
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.134318416839165e-06,
+      "loss": 0.4027,
+      "step": 22167
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.12779684424919e-06,
+      "loss": 0.4712,
+      "step": 22168
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.121277527516152e-06,
+      "loss": 0.3404,
+      "step": 22169
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.114760466744465e-06,
+      "loss": 0.4357,
+      "step": 22170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.108245662038394e-06,
+      "loss": 0.4586,
+      "step": 22171
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.101733113502357e-06,
+      "loss": 0.3596,
+      "step": 22172
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.095222821240588e-06,
+      "loss": 0.5117,
+      "step": 22173
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.088714785357337e-06,
+      "loss": 0.3576,
+      "step": 22174
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.082209005956887e-06,
+      "loss": 0.5222,
+      "step": 22175
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.075705483143336e-06,
+      "loss": 0.9033,
+      "step": 22176
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.069204217020853e-06,
+      "loss": 0.3487,
+      "step": 22177
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.062705207693604e-06,
+      "loss": 0.0011,
+      "step": 22178
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.05620845526559e-06,
+      "loss": 0.8136,
+      "step": 22179
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.04971395984086e-06,
+      "loss": 0.3146,
+      "step": 22180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.043221721523431e-06,
+      "loss": 0.2124,
+      "step": 22181
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.036731740417269e-06,
+      "loss": 0.5148,
+      "step": 22182
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.030244016626275e-06,
+      "loss": 0.1857,
+      "step": 22183
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.023758550254368e-06,
+      "loss": 0.4472,
+      "step": 22184
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.017275341405394e-06,
+      "loss": 0.963,
+      "step": 22185
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.010794390183173e-06,
+      "loss": 1.3153,
+      "step": 22186
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.004315696691472e-06,
+      "loss": 0.2361,
+      "step": 22187
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.997839261034057e-06,
+      "loss": 2.4746,
+      "step": 22188
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.991365083314628e-06,
+      "loss": 0.7764,
+      "step": 22189
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.984893163636869e-06,
+      "loss": 0.0313,
+      "step": 22190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.978423502104431e-06,
+      "loss": 0.0018,
+      "step": 22191
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.971956098820848e-06,
+      "loss": 0.3969,
+      "step": 22192
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.96549095388977e-06,
+      "loss": 0.5161,
+      "step": 22193
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.959028067414697e-06,
+      "loss": 0.3414,
+      "step": 22194
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.952567439499082e-06,
+      "loss": 0.4899,
+      "step": 22195
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.946109070246438e-06,
+      "loss": 0.0047,
+      "step": 22196
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.939652959760185e-06,
+      "loss": 0.5442,
+      "step": 22197
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.933199108143675e-06,
+      "loss": 0.7031,
+      "step": 22198
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.926747515500237e-06,
+      "loss": 0.0037,
+      "step": 22199
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.920298181933261e-06,
+      "loss": 0.6308,
+      "step": 22200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.913851107545961e-06,
+      "loss": 1.2657,
+      "step": 22201
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.907406292441588e-06,
+      "loss": 0.0026,
+      "step": 22202
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.900963736723359e-06,
+      "loss": 0.0003,
+      "step": 22203
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.894523440494428e-06,
+      "loss": 0.1379,
+      "step": 22204
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.888085403857941e-06,
+      "loss": 0.7821,
+      "step": 22205
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.881649626916986e-06,
+      "loss": 0.8433,
+      "step": 22206
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.87521610977463e-06,
+      "loss": 0.2631,
+      "step": 22207
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.868784852533872e-06,
+      "loss": 0.7149,
+      "step": 22208
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.862355855297731e-06,
+      "loss": 0.2723,
+      "step": 22209
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.855929118169142e-06,
+      "loss": 0.7347,
+      "step": 22210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.84950464125102e-06,
+      "loss": 0.7174,
+      "step": 22211
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.843082424646237e-06,
+      "loss": 0.0046,
+      "step": 22212
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.836662468457672e-06,
+      "loss": 0.3075,
+      "step": 22213
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.830244772788047e-06,
+      "loss": 0.2381,
+      "step": 22214
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.823829337740213e-06,
+      "loss": 0.1931,
+      "step": 22215
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.817416163416901e-06,
+      "loss": 0.3531,
+      "step": 22216
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.81100524992075e-06,
+      "loss": 0.0014,
+      "step": 22217
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.804596597354458e-06,
+      "loss": 0.5582,
+      "step": 22218
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.798190205820677e-06,
+      "loss": 0.2472,
+      "step": 22219
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.791786075421926e-06,
+      "loss": 0.574,
+      "step": 22220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.785384206260837e-06,
+      "loss": 1.5693,
+      "step": 22221
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.778984598439865e-06,
+      "loss": 0.3145,
+      "step": 22222
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.772587252061475e-06,
+      "loss": 0.7671,
+      "step": 22223
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.76619216722822e-06,
+      "loss": 0.0776,
+      "step": 22224
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.759799344042384e-06,
+      "loss": 0.074,
+      "step": 22225
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.753408782606386e-06,
+      "loss": 0.001,
+      "step": 22226
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.747020483022609e-06,
+      "loss": 1.197,
+      "step": 22227
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.740634445393275e-06,
+      "loss": 0.5566,
+      "step": 22228
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.734250669820663e-06,
+      "loss": 0.4069,
+      "step": 22229
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.727869156407063e-06,
+      "loss": 0.2728,
+      "step": 22230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.721489905254591e-06,
+      "loss": 0.2828,
+      "step": 22231
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.715112916465432e-06,
+      "loss": 0.0019,
+      "step": 22232
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.708738190141706e-06,
+      "loss": 0.0015,
+      "step": 22233
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.702365726385497e-06,
+      "loss": 0.2003,
+      "step": 22234
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.695995525298821e-06,
+      "loss": 0.0014,
+      "step": 22235
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.689627586983717e-06,
+      "loss": 0.0025,
+      "step": 22236
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.68326191154215e-06,
+      "loss": 0.8053,
+      "step": 22237
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.676898499076057e-06,
+      "loss": 0.3318,
+      "step": 22238
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.67053734968734e-06,
+      "loss": 0.3511,
+      "step": 22239
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.6641784634779e-06,
+      "loss": 0.8429,
+      "step": 22240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.657821840549457e-06,
+      "loss": 0.7446,
+      "step": 22241
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.65146748100391e-06,
+      "loss": 1.4999,
+      "step": 22242
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.64511538494298e-06,
+      "loss": 0.3719,
+      "step": 22243
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.638765552468351e-06,
+      "loss": 0.8487,
+      "step": 22244
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.632417983681761e-06,
+      "loss": 0.0026,
+      "step": 22245
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.62607267868486e-06,
+      "loss": 0.0014,
+      "step": 22246
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.619729637579215e-06,
+      "loss": 0.2501,
+      "step": 22247
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.61338886046638e-06,
+      "loss": 0.0059,
+      "step": 22248
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.607050347447974e-06,
+      "loss": 0.001,
+      "step": 22249
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.600714098625449e-06,
+      "loss": 1.9213,
+      "step": 22250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.594380114100275e-06,
+      "loss": 0.0466,
+      "step": 22251
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.588048393973868e-06,
+      "loss": 0.2207,
+      "step": 22252
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.581718938347648e-06,
+      "loss": 0.2661,
+      "step": 22253
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.575391747322952e-06,
+      "loss": 0.0029,
+      "step": 22254
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.56906682100113e-06,
+      "loss": 0.1483,
+      "step": 22255
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.56274415948342e-06,
+      "loss": 1.4668,
+      "step": 22256
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.556423762871106e-06,
+      "loss": 0.5131,
+      "step": 22257
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.550105631265392e-06,
+      "loss": 0.7974,
+      "step": 22258
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.543789764767445e-06,
+      "loss": 0.6285,
+      "step": 22259
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.537476163478401e-06,
+      "loss": 0.0155,
+      "step": 22260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.531164827499382e-06,
+      "loss": 0.3747,
+      "step": 22261
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.524855756931438e-06,
+      "loss": 0.6357,
+      "step": 22262
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.518548951875603e-06,
+      "loss": 0.3324,
+      "step": 22263
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.512244412432883e-06,
+      "loss": 0.2003,
+      "step": 22264
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.505942138704231e-06,
+      "loss": 0.0061,
+      "step": 22265
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.499642130790529e-06,
+      "loss": 0.0055,
+      "step": 22266
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.493344388792716e-06,
+      "loss": 0.2998,
+      "step": 22267
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.487048912811645e-06,
+      "loss": 0.2057,
+      "step": 22268
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.48075570294805e-06,
+      "loss": 0.3984,
+      "step": 22269
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.474464759302817e-06,
+      "loss": 0.2379,
+      "step": 22270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.468176081976618e-06,
+      "loss": 0.4668,
+      "step": 22271
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.46188967107012e-06,
+      "loss": 0.4621,
+      "step": 22272
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.455605526684112e-06,
+      "loss": 0.0955,
+      "step": 22273
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.44932364891911e-06,
+      "loss": 0.0012,
+      "step": 22274
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.443044037875751e-06,
+      "loss": 0.5892,
+      "step": 22275
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.436766693654622e-06,
+      "loss": 0.0021,
+      "step": 22276
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.430491616356194e-06,
+      "loss": 0.1886,
+      "step": 22277
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.424218806080968e-06,
+      "loss": 0.142,
+      "step": 22278
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.417948262929431e-06,
+      "loss": 0.6344,
+      "step": 22279
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.41167998700195e-06,
+      "loss": 1.0195,
+      "step": 22280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.405413978398934e-06,
+      "loss": 0.1571,
+      "step": 22281
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.399150237220697e-06,
+      "loss": 1.3892,
+      "step": 22282
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.392888763567562e-06,
+      "loss": 0.0226,
+      "step": 22283
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.386629557539781e-06,
+      "loss": 0.0012,
+      "step": 22284
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.380372619237607e-06,
+      "loss": 0.8261,
+      "step": 22285
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.374117948761211e-06,
+      "loss": 0.7045,
+      "step": 22286
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.367865546210778e-06,
+      "loss": 0.0019,
+      "step": 22287
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.361615411686412e-06,
+      "loss": 0.3795,
+      "step": 22288
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.355367545288233e-06,
+      "loss": 0.6119,
+      "step": 22289
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.34912194711621e-06,
+      "loss": 0.8276,
+      "step": 22290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.342878617270431e-06,
+      "loss": 2.3762,
+      "step": 22291
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.336637555850884e-06,
+      "loss": 0.4941,
+      "step": 22292
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.330398762957435e-06,
+      "loss": 0.0016,
+      "step": 22293
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.324162238690058e-06,
+      "loss": 0.5035,
+      "step": 22294
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.31792798314862e-06,
+      "loss": 1.0671,
+      "step": 22295
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.311695996432876e-06,
+      "loss": 0.238,
+      "step": 22296
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.305466278642714e-06,
+      "loss": 0.3774,
+      "step": 22297
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.29923882987787e-06,
+      "loss": 0.3432,
+      "step": 22298
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.29301365023803e-06,
+      "loss": 0.4213,
+      "step": 22299
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.286790739822895e-06,
+      "loss": 0.0009,
+      "step": 22300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.280570098732142e-06,
+      "loss": 0.454,
+      "step": 22301
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.274351727065354e-06,
+      "loss": 1.9287,
+      "step": 22302
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.268135624922134e-06,
+      "loss": 0.2519,
+      "step": 22303
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.261921792402004e-06,
+      "loss": 1.1281,
+      "step": 22304
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.255710229604467e-06,
+      "loss": 0.0015,
+      "step": 22305
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.249500936629027e-06,
+      "loss": 0.2344,
+      "step": 22306
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.243293913575072e-06,
+      "loss": 0.3047,
+      "step": 22307
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 8.237089160542021e-06,
+      "loss": 0.4374,
+      "step": 22308
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.230886677629246e-06,
+      "loss": 0.7226,
+      "step": 22309
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.224686464936047e-06,
+      "loss": 0.7839,
+      "step": 22310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.218488522561717e-06,
+      "loss": 0.5092,
+      "step": 22311
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.212292850605506e-06,
+      "loss": 0.4861,
+      "step": 22312
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.206099449166637e-06,
+      "loss": 0.0046,
+      "step": 22313
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.199908318344295e-06,
+      "loss": 0.1382,
+      "step": 22314
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.193719458237568e-06,
+      "loss": 0.0022,
+      "step": 22315
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.187532868945629e-06,
+      "loss": 0.4241,
+      "step": 22316
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.181348550567545e-06,
+      "loss": 0.0859,
+      "step": 22317
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.175166503202273e-06,
+      "loss": 0.9021,
+      "step": 22318
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.168986726948918e-06,
+      "loss": 0.8207,
+      "step": 22319
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.162809221906348e-06,
+      "loss": 0.5575,
+      "step": 22320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.1566339881735e-06,
+      "loss": 0.1415,
+      "step": 22321
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.150461025849314e-06,
+      "loss": 0.0008,
+      "step": 22322
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.14429033503261e-06,
+      "loss": 0.0011,
+      "step": 22323
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.13812191582216e-06,
+      "loss": 0.6123,
+      "step": 22324
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.131955768316833e-06,
+      "loss": 0.1659,
+      "step": 22325
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.125791892615302e-06,
+      "loss": 0.8507,
+      "step": 22326
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.11963028881627e-06,
+      "loss": 0.5316,
+      "step": 22327
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.113470957018458e-06,
+      "loss": 0.4009,
+      "step": 22328
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.107313897320455e-06,
+      "loss": 0.2167,
+      "step": 22329
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.101159109820865e-06,
+      "loss": 0.7952,
+      "step": 22330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.095006594618258e-06,
+      "loss": 0.1123,
+      "step": 22331
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.088856351811141e-06,
+      "loss": 0.3418,
+      "step": 22332
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.082708381498015e-06,
+      "loss": 0.6879,
+      "step": 22333
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.07656268377732e-06,
+      "loss": 0.5921,
+      "step": 22334
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.070419258747496e-06,
+      "loss": 0.0047,
+      "step": 22335
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.064278106506878e-06,
+      "loss": 0.0033,
+      "step": 22336
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.058139227153837e-06,
+      "loss": 0.5462,
+      "step": 22337
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.052002620786695e-06,
+      "loss": 0.0038,
+      "step": 22338
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.04586828750366e-06,
+      "loss": 0.0025,
+      "step": 22339
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.039736227403015e-06,
+      "loss": 0.4741,
+      "step": 22340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.033606440582968e-06,
+      "loss": 0.8484,
+      "step": 22341
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.027478927141605e-06,
+      "loss": 1.1364,
+      "step": 22342
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.021353687177117e-06,
+      "loss": 0.266,
+      "step": 22343
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.015230720787607e-06,
+      "loss": 0.7612,
+      "step": 22344
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.009110028071031e-06,
+      "loss": 0.7505,
+      "step": 22345
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 8.002991609125475e-06,
+      "loss": 0.9315,
+      "step": 22346
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.99687546404893e-06,
+      "loss": 0.4018,
+      "step": 22347
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.990761592939265e-06,
+      "loss": 0.2907,
+      "step": 22348
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.98464999589447e-06,
+      "loss": 0.2646,
+      "step": 22349
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.978540673012352e-06,
+      "loss": 0.981,
+      "step": 22350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.972433624390746e-06,
+      "loss": 0.7644,
+      "step": 22351
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.966328850127474e-06,
+      "loss": 0.1892,
+      "step": 22352
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.960226350320276e-06,
+      "loss": 0.5164,
+      "step": 22353
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.95412612506689e-06,
+      "loss": 0.2292,
+      "step": 22354
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.948028174464988e-06,
+      "loss": 0.6039,
+      "step": 22355
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.941932498612224e-06,
+      "loss": 0.9537,
+      "step": 22356
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.935839097606223e-06,
+      "loss": 0.9655,
+      "step": 22357
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.929747971544537e-06,
+      "loss": 0.3348,
+      "step": 22358
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.923659120524722e-06,
+      "loss": 0.262,
+      "step": 22359
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.917572544644269e-06,
+      "loss": 0.3057,
+      "step": 22360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.91148824400068e-06,
+      "loss": 0.2353,
+      "step": 22361
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.905406218691346e-06,
+      "loss": 0.6813,
+      "step": 22362
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.899326468813671e-06,
+      "loss": 0.4227,
+      "step": 22363
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.89324899446503e-06,
+      "loss": 0.3959,
+      "step": 22364
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.88717379574274e-06,
+      "loss": 0.8969,
+      "step": 22365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.881100872744095e-06,
+      "loss": 0.0013,
+      "step": 22366
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.87503022556628e-06,
+      "loss": 0.3059,
+      "step": 22367
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.868961854306604e-06,
+      "loss": 0.6667,
+      "step": 22368
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.86289575906217e-06,
+      "loss": 0.5074,
+      "step": 22369
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.856831939930136e-06,
+      "loss": 0.4384,
+      "step": 22370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.850770397007638e-06,
+      "loss": 0.3832,
+      "step": 22371
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.844711130391701e-06,
+      "loss": 0.1684,
+      "step": 22372
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.838654140179346e-06,
+      "loss": 0.3102,
+      "step": 22373
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.832599426467646e-06,
+      "loss": 0.3091,
+      "step": 22374
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.826546989353455e-06,
+      "loss": 0.0345,
+      "step": 22375
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.820496828933747e-06,
+      "loss": 0.3224,
+      "step": 22376
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.814448945305429e-06,
+      "loss": 0.4839,
+      "step": 22377
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.808403338565306e-06,
+      "loss": 0.7434,
+      "step": 22378
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.802360008810199e-06,
+      "loss": 0.3053,
+      "step": 22379
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.796318956136882e-06,
+      "loss": 0.377,
+      "step": 22380
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.790280180642111e-06,
+      "loss": 0.402,
+      "step": 22381
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.784243682422575e-06,
+      "loss": 0.5262,
+      "step": 22382
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.77820946157493e-06,
+      "loss": 0.5291,
+      "step": 22383
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.772177518195816e-06,
+      "loss": 0.528,
+      "step": 22384
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.766147852381837e-06,
+      "loss": 0.6412,
+      "step": 22385
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.760120464229536e-06,
+      "loss": 0.6222,
+      "step": 22386
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.75409535383545e-06,
+      "loss": 0.2283,
+      "step": 22387
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.748072521296e-06,
+      "loss": 0.368,
+      "step": 22388
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.742051966707696e-06,
+      "loss": 0.2548,
+      "step": 22389
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.736033690166959e-06,
+      "loss": 0.0016,
+      "step": 22390
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.730017691770096e-06,
+      "loss": 0.2313,
+      "step": 22391
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.724003971613497e-06,
+      "loss": 1.6906,
+      "step": 22392
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.717992529793465e-06,
+      "loss": 0.388,
+      "step": 22393
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.71198336640621e-06,
+      "loss": 0.0014,
+      "step": 22394
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.70597648154802e-06,
+      "loss": 0.2405,
+      "step": 22395
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.699971875315086e-06,
+      "loss": 0.3551,
+      "step": 22396
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.693969547803496e-06,
+      "loss": 0.8508,
+      "step": 22397
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.687969499109458e-06,
+      "loss": 0.1739,
+      "step": 22398
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.681971729328979e-06,
+      "loss": 0.314,
+      "step": 22399
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.675976238558129e-06,
+      "loss": 0.6773,
+      "step": 22400
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.669983026892935e-06,
+      "loss": 0.5795,
+      "step": 22401
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.66399209442935e-06,
+      "loss": 1.3876,
+      "step": 22402
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.658003441263317e-06,
+      "loss": 0.2635,
+      "step": 22403
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.652017067490723e-06,
+      "loss": 0.365,
+      "step": 22404
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.646032973207445e-06,
+      "loss": 0.305,
+      "step": 22405
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.64005115850932e-06,
+      "loss": 0.0236,
+      "step": 22406
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.634071623492105e-06,
+      "loss": 0.0499,
+      "step": 22407
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.628094368251575e-06,
+      "loss": 0.8699,
+      "step": 22408
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.622119392883452e-06,
+      "loss": 0.5734,
+      "step": 22409
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.61614669748341e-06,
+      "loss": 0.0804,
+      "step": 22410
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.6101762821470734e-06,
+      "loss": 0.0834,
+      "step": 22411
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.604208146970081e-06,
+      "loss": 0.303,
+      "step": 22412
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.59824229204799e-06,
+      "loss": 0.1578,
+      "step": 22413
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.592278717476325e-06,
+      "loss": 0.2162,
+      "step": 22414
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.586317423350591e-06,
+      "loss": 0.0021,
+      "step": 22415
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.580358409766263e-06,
+      "loss": 0.7011,
+      "step": 22416
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.5744016768187645e-06,
+      "loss": 0.0018,
+      "step": 22417
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.568447224603452e-06,
+      "loss": 0.3739,
+      "step": 22418
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.562495053215667e-06,
+      "loss": 0.0006,
+      "step": 22419
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.556545162750816e-06,
+      "loss": 0.6056,
+      "step": 22420
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.550597553304088e-06,
+      "loss": 0.005,
+      "step": 22421
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.544652224970726e-06,
+      "loss": 0.5502,
+      "step": 22422
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.53870917784602e-06,
+      "loss": 0.4793,
+      "step": 22423
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.532768412025042e-06,
+      "loss": 0.2419,
+      "step": 22424
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.5268299276029506e-06,
+      "loss": 0.2411,
+      "step": 22425
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.520893724674903e-06,
+      "loss": 1.5759,
+      "step": 22426
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.51495980333589e-06,
+      "loss": 0.1599,
+      "step": 22427
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.509028163680952e-06,
+      "loss": 0.3919,
+      "step": 22428
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.50309880580508e-06,
+      "loss": 0.1952,
+      "step": 22429
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.4971717298032145e-06,
+      "loss": 0.9854,
+      "step": 22430
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.491246935770295e-06,
+      "loss": 0.3278,
+      "step": 22431
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.485324423801165e-06,
+      "loss": 0.3043,
+      "step": 22432
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.479404193990679e-06,
+      "loss": 0.6323,
+      "step": 22433
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.473486246433663e-06,
+      "loss": 0.4992,
+      "step": 22434
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.467570581224841e-06,
+      "loss": 0.3022,
+      "step": 22435
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.461657198458987e-06,
+      "loss": 0.2638,
+      "step": 22436
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.455746098230742e-06,
+      "loss": 0.0008,
+      "step": 22437
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.449837280634796e-06,
+      "loss": 0.3002,
+      "step": 22438
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.4439307457658085e-06,
+      "loss": 0.0046,
+      "step": 22439
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.438026493718269e-06,
+      "loss": 0.168,
+      "step": 22440
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.4321245245868185e-06,
+      "loss": 0.3208,
+      "step": 22441
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.426224838465933e-06,
+      "loss": 0.4491,
+      "step": 22442
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.420327435450052e-06,
+      "loss": 0.9179,
+      "step": 22443
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.41443231563365e-06,
+      "loss": 0.6504,
+      "step": 22444
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.408539479111153e-06,
+      "loss": 0.0288,
+      "step": 22445
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.402648925976867e-06,
+      "loss": 0.2573,
+      "step": 22446
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.396760656325185e-06,
+      "loss": 0.4946,
+      "step": 22447
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.39087467025033e-06,
+      "loss": 0.001,
+      "step": 22448
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.384990967846577e-06,
+      "loss": 0.3846,
+      "step": 22449
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.379109549208201e-06,
+      "loss": 0.2442,
+      "step": 22450
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.373230414429327e-06,
+      "loss": 0.8456,
+      "step": 22451
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.367353563604112e-06,
+      "loss": 0.0037,
+      "step": 22452
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.361478996826664e-06,
+      "loss": 0.0013,
+      "step": 22453
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.355606714191059e-06,
+      "loss": 1.0743,
+      "step": 22454
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.3497367157913376e-06,
+      "loss": 0.0012,
+      "step": 22455
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.3438690017214914e-06,
+      "loss": 0.6093,
+      "step": 22456
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.338003572075496e-06,
+      "loss": 0.334,
+      "step": 22457
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.3321404269472585e-06,
+      "loss": 0.5937,
+      "step": 22458
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.326279566430671e-06,
+      "loss": 0.0072,
+      "step": 22459
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.3204209906196075e-06,
+      "loss": 0.3252,
+      "step": 22460
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.314564699607861e-06,
+      "loss": 1.9585,
+      "step": 22461
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.308710693489222e-06,
+      "loss": 0.0078,
+      "step": 22462
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.302858972357417e-06,
+      "loss": 0.0117,
+      "step": 22463
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.297009536306187e-06,
+      "loss": 0.8878,
+      "step": 22464
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.291162385429172e-06,
+      "loss": 0.9728,
+      "step": 22465
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.28531751982005e-06,
+      "loss": 0.0213,
+      "step": 22466
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.279474939572344e-06,
+      "loss": 0.315,
+      "step": 22467
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.273634644779646e-06,
+      "loss": 0.0749,
+      "step": 22468
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.267796635535517e-06,
+      "loss": 0.5653,
+      "step": 22469
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.261960911933412e-06,
+      "loss": 0.3338,
+      "step": 22470
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.256127474066742e-06,
+      "loss": 0.2641,
+      "step": 22471
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.250296322029014e-06,
+      "loss": 0.0295,
+      "step": 22472
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.244467455913539e-06,
+      "loss": 0.6992,
+      "step": 22473
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.238640875813639e-06,
+      "loss": 0.2503,
+      "step": 22474
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.232816581822692e-06,
+      "loss": 0.0016,
+      "step": 22475
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.226994574033906e-06,
+      "loss": 0.0134,
+      "step": 22476
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.221174852540523e-06,
+      "loss": 0.2294,
+      "step": 22477
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.215357417435752e-06,
+      "loss": 0.0005,
+      "step": 22478
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.209542268812752e-06,
+      "loss": 0.4499,
+      "step": 22479
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.2037294067646136e-06,
+      "loss": 0.8199,
+      "step": 22480
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.1979188313844475e-06,
+      "loss": 1.0325,
+      "step": 22481
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.192110542765312e-06,
+      "loss": 0.2564,
+      "step": 22482
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.186304541000182e-06,
+      "loss": 0.0745,
+      "step": 22483
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.180500826182051e-06,
+      "loss": 0.0005,
+      "step": 22484
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.174699398403894e-06,
+      "loss": 0.6749,
+      "step": 22485
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.168900257758519e-06,
+      "loss": 0.8967,
+      "step": 22486
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.163103404338871e-06,
+      "loss": 0.7123,
+      "step": 22487
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.157308838237774e-06,
+      "loss": 0.574,
+      "step": 22488
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.151516559547971e-06,
+      "loss": 0.0236,
+      "step": 22489
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.145726568362253e-06,
+      "loss": 0.4535,
+      "step": 22490
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.139938864773348e-06,
+      "loss": 0.0055,
+      "step": 22491
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.1341534488738804e-06,
+      "loss": 0.4838,
+      "step": 22492
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.12837032075656e-06,
+      "loss": 3.2766,
+      "step": 22493
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.12258948051398e-06,
+      "loss": 0.0016,
+      "step": 22494
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.116810928238664e-06,
+      "loss": 0.3562,
+      "step": 22495
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.111034664023241e-06,
+      "loss": 0.3374,
+      "step": 22496
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.1052606879601015e-06,
+      "loss": 0.9626,
+      "step": 22497
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.099489000141756e-06,
+      "loss": 0.9238,
+      "step": 22498
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.093719600660663e-06,
+      "loss": 0.2017,
+      "step": 22499
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.087952489609166e-06,
+      "loss": 0.5274,
+      "step": 22500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.082187667079608e-06,
+      "loss": 0.3566,
+      "step": 22501
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.076425133164348e-06,
+      "loss": 0.4361,
+      "step": 22502
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.070664887955646e-06,
+      "loss": 0.0036,
+      "step": 22503
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.064906931545728e-06,
+      "loss": 0.3324,
+      "step": 22504
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.059151264026802e-06,
+      "loss": 1.2095,
+      "step": 22505
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.053397885491047e-06,
+      "loss": 0.0011,
+      "step": 22506
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.047646796030604e-06,
+      "loss": 0.1925,
+      "step": 22507
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.04189799573755e-06,
+      "loss": 0.671,
+      "step": 22508
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.036151484703945e-06,
+      "loss": 0.7077,
+      "step": 22509
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.030407263021815e-06,
+      "loss": 0.5785,
+      "step": 22510
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.024665330783152e-06,
+      "loss": 0.7832,
+      "step": 22511
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.018925688079885e-06,
+      "loss": 0.9105,
+      "step": 22512
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.013188335003955e-06,
+      "loss": 0.9157,
+      "step": 22513
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.007453271647207e-06,
+      "loss": 0.1083,
+      "step": 22514
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 7.001720498101515e-06,
+      "loss": 0.0531,
+      "step": 22515
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.995990014458624e-06,
+      "loss": 0.0009,
+      "step": 22516
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.990261820810361e-06,
+      "loss": 0.1061,
+      "step": 22517
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.984535917248435e-06,
+      "loss": 0.0161,
+      "step": 22518
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.978812303864523e-06,
+      "loss": 1.5547,
+      "step": 22519
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.973090980750251e-06,
+      "loss": 0.3214,
+      "step": 22520
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.96737194799733e-06,
+      "loss": 0.1249,
+      "step": 22521
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.961655205697253e-06,
+      "loss": 0.2066,
+      "step": 22522
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.95594075394158e-06,
+      "loss": 0.1019,
+      "step": 22523
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.950228592821871e-06,
+      "loss": 1.0029,
+      "step": 22524
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.944518722429554e-06,
+      "loss": 0.7795,
+      "step": 22525
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.938811142856071e-06,
+      "loss": 0.1855,
+      "step": 22526
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.933105854192816e-06,
+      "loss": 0.0049,
+      "step": 22527
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.927402856531167e-06,
+      "loss": 0.371,
+      "step": 22528
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.921702149962433e-06,
+      "loss": 0.3611,
+      "step": 22529
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.916003734577924e-06,
+      "loss": 1.4619,
+      "step": 22530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.910307610468851e-06,
+      "loss": 0.0008,
+      "step": 22531
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.904613777726475e-06,
+      "loss": 0.2532,
+      "step": 22532
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.898922236441939e-06,
+      "loss": 0.094,
+      "step": 22533
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.893232986706421e-06,
+      "loss": 0.599,
+      "step": 22534
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.8875460286109795e-06,
+      "loss": 0.5584,
+      "step": 22535
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.88186136224671e-06,
+      "loss": 0.0679,
+      "step": 22536
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.876178987704656e-06,
+      "loss": 0.6441,
+      "step": 22537
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.8704989050757775e-06,
+      "loss": 0.2417,
+      "step": 22538
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.864821114451053e-06,
+      "loss": 0.0268,
+      "step": 22539
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.8591456159214246e-06,
+      "loss": 0.3085,
+      "step": 22540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.85347240957772e-06,
+      "loss": 0.001,
+      "step": 22541
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.847801495510835e-06,
+      "loss": 0.4156,
+      "step": 22542
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.842132873811596e-06,
+      "loss": 0.0069,
+      "step": 22543
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.8364665445707125e-06,
+      "loss": 0.1181,
+      "step": 22544
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.830802507878996e-06,
+      "loss": 0.4483,
+      "step": 22545
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.825140763827092e-06,
+      "loss": 0.3903,
+      "step": 22546
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.819481312505659e-06,
+      "loss": 0.45,
+      "step": 22547
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.813824154005376e-06,
+      "loss": 2.2053,
+      "step": 22548
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.808169288416804e-06,
+      "loss": 0.2618,
+      "step": 22549
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.8025167158304864e-06,
+      "loss": 0.4861,
+      "step": 22550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.796866436336967e-06,
+      "loss": 0.0009,
+      "step": 22551
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.791218450026726e-06,
+      "loss": 0.2602,
+      "step": 22552
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.785572756990171e-06,
+      "loss": 0.5989,
+      "step": 22553
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.779929357317748e-06,
+      "loss": 0.5127,
+      "step": 22554
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.774288251099802e-06,
+      "loss": 0.6923,
+      "step": 22555
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.768649438426677e-06,
+      "loss": 0.5031,
+      "step": 22556
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.7630129193886825e-06,
+      "loss": 0.8853,
+      "step": 22557
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.757378694076065e-06,
+      "loss": 2.1163,
+      "step": 22558
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.751746762579052e-06,
+      "loss": 0.377,
+      "step": 22559
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.746117124987821e-06,
+      "loss": 0.002,
+      "step": 22560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.740489781392533e-06,
+      "loss": 0.0015,
+      "step": 22561
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.734864731883299e-06,
+      "loss": 0.6091,
+      "step": 22562
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.729241976550198e-06,
+      "loss": 0.371,
+      "step": 22563
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.7236215154832904e-06,
+      "loss": 2.5986,
+      "step": 22564
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.718003348772505e-06,
+      "loss": 0.0365,
+      "step": 22565
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.712387476507885e-06,
+      "loss": 0.4047,
+      "step": 22566
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.706773898779361e-06,
+      "loss": 0.7075,
+      "step": 22567
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.701162615676758e-06,
+      "loss": 0.3021,
+      "step": 22568
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.69555362728999e-06,
+      "loss": 1.6573,
+      "step": 22569
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.689946933708884e-06,
+      "loss": 0.8347,
+      "step": 22570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.684342535023185e-06,
+      "loss": 0.1229,
+      "step": 22571
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.678740431322621e-06,
+      "loss": 0.2116,
+      "step": 22572
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.673140622696988e-06,
+      "loss": 0.2861,
+      "step": 22573
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.667543109235862e-06,
+      "loss": 0.3995,
+      "step": 22574
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.6619478910289405e-06,
+      "loss": 0.001,
+      "step": 22575
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.6563549681658e-06,
+      "loss": 0.4263,
+      "step": 22576
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.650764340736004e-06,
+      "loss": 0.7835,
+      "step": 22577
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.645176008829078e-06,
+      "loss": 0.4034,
+      "step": 22578
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.6395899725345205e-06,
+      "loss": 0.1061,
+      "step": 22579
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.634006231941774e-06,
+      "loss": 0.3992,
+      "step": 22580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.628424787140252e-06,
+      "loss": 0.4,
+      "step": 22581
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.622845638219332e-06,
+      "loss": 2.9618,
+      "step": 22582
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.617268785268359e-06,
+      "loss": 0.7069,
+      "step": 22583
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.611694228376646e-06,
+      "loss": 0.9621,
+      "step": 22584
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.606121967633454e-06,
+      "loss": 0.2319,
+      "step": 22585
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.600552003128029e-06,
+      "loss": 0.0027,
+      "step": 22586
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.594984334949516e-06,
+      "loss": 0.101,
+      "step": 22587
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.589418963187127e-06,
+      "loss": 0.251,
+      "step": 22588
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.58385588792999e-06,
+      "loss": 0.3632,
+      "step": 22589
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.578295109267118e-06,
+      "loss": 1.0484,
+      "step": 22590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.5727366272876225e-06,
+      "loss": 0.6232,
+      "step": 22591
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.567180442080516e-06,
+      "loss": 0.0487,
+      "step": 22592
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.561626553734711e-06,
+      "loss": 0.6717,
+      "step": 22593
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.556074962339236e-06,
+      "loss": 0.0022,
+      "step": 22594
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.55052566798292e-06,
+      "loss": 0.179,
+      "step": 22595
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.544978670754608e-06,
+      "loss": 0.4996,
+      "step": 22596
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.53943397074323e-06,
+      "loss": 0.0008,
+      "step": 22597
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.5338915680374805e-06,
+      "loss": 0.5524,
+      "step": 22598
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.528351462726123e-06,
+      "loss": 0.9432,
+      "step": 22599
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.522813654897952e-06,
+      "loss": 0.2666,
+      "step": 22600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.517278144641547e-06,
+      "loss": 0.2562,
+      "step": 22601
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.511744932045587e-06,
+      "loss": 0.876,
+      "step": 22602
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.506214017198735e-06,
+      "loss": 0.8604,
+      "step": 22603
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.500685400189487e-06,
+      "loss": 0.0014,
+      "step": 22604
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.495159081106388e-06,
+      "loss": 1.5686,
+      "step": 22605
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.489635060037968e-06,
+      "loss": 0.5413,
+      "step": 22606
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.484113337072638e-06,
+      "loss": 0.4087,
+      "step": 22607
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.478593912298863e-06,
+      "loss": 0.7249,
+      "step": 22608
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.473076785805004e-06,
+      "loss": 0.0011,
+      "step": 22609
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.4675619576794234e-06,
+      "loss": 0.5904,
+      "step": 22610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.462049428010419e-06,
+      "loss": 0.2327,
+      "step": 22611
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.456539196886268e-06,
+      "loss": 0.3316,
+      "step": 22612
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.4510312643952515e-06,
+      "loss": 0.0213,
+      "step": 22613
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.445525630625481e-06,
+      "loss": 0.2909,
+      "step": 22614
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.440022295665187e-06,
+      "loss": 0.8942,
+      "step": 22615
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.434521259602515e-06,
+      "loss": 0.5107,
+      "step": 22616
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.429022522525462e-06,
+      "loss": 0.3066,
+      "step": 22617
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.4235260845221905e-06,
+      "loss": 0.0653,
+      "step": 22618
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.4180319456806794e-06,
+      "loss": 0.7161,
+      "step": 22619
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.412540106088876e-06,
+      "loss": 0.2703,
+      "step": 22620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.407050565834726e-06,
+      "loss": 0.2418,
+      "step": 22621
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.401563325006209e-06,
+      "loss": 0.2663,
+      "step": 22622
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.396078383691123e-06,
+      "loss": 0.3663,
+      "step": 22623
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.390595741977312e-06,
+      "loss": 0.6064,
+      "step": 22624
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.385115399952589e-06,
+      "loss": 0.6389,
+      "step": 22625
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.379637357704703e-06,
+      "loss": 0.2921,
+      "step": 22626
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.374161615321382e-06,
+      "loss": 0.0015,
+      "step": 22627
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.3686881728903235e-06,
+      "loss": 0.9351,
+      "step": 22628
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.36321703049914e-06,
+      "loss": 0.0031,
+      "step": 22629
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.3577481882354955e-06,
+      "loss": 0.8261,
+      "step": 22630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.352281646186919e-06,
+      "loss": 0.0008,
+      "step": 22631
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.346817404440957e-06,
+      "loss": 0.6253,
+      "step": 22632
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.34135546308514e-06,
+      "loss": 0.7195,
+      "step": 22633
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.335895822206916e-06,
+      "loss": 0.5428,
+      "step": 22634
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.330438481893713e-06,
+      "loss": 0.417,
+      "step": 22635
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.3249834422328945e-06,
+      "loss": 1.0336,
+      "step": 22636
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.319530703311859e-06,
+      "loss": 0.9481,
+      "step": 22637
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.314080265217919e-06,
+      "loss": 0.8919,
+      "step": 22638
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.308632128038304e-06,
+      "loss": 0.001,
+      "step": 22639
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.303186291860312e-06,
+      "loss": 0.6362,
+      "step": 22640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.297742756771157e-06,
+      "loss": 1.1514,
+      "step": 22641
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.292301522857934e-06,
+      "loss": 0.0115,
+      "step": 22642
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.286862590207875e-06,
+      "loss": 0.5097,
+      "step": 22643
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.281425958908009e-06,
+      "loss": 0.1132,
+      "step": 22644
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.2759916290454015e-06,
+      "loss": 0.3232,
+      "step": 22645
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.270559600707115e-06,
+      "loss": 0.3974,
+      "step": 22646
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.265129873980096e-06,
+      "loss": 0.5708,
+      "step": 22647
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.2597024489512925e-06,
+      "loss": 0.1906,
+      "step": 22648
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.254277325707668e-06,
+      "loss": 0.3165,
+      "step": 22649
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.24885450433602e-06,
+      "loss": 0.6336,
+      "step": 22650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.243433984923229e-06,
+      "loss": 0.8614,
+      "step": 22651
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.238015767556126e-06,
+      "loss": 0.908,
+      "step": 22652
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.232599852321424e-06,
+      "loss": 0.0005,
+      "step": 22653
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.2271862393058865e-06,
+      "loss": 0.0012,
+      "step": 22654
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.221774928596163e-06,
+      "loss": 0.1928,
+      "step": 22655
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.216365920278949e-06,
+      "loss": 0.9142,
+      "step": 22656
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.210959214440842e-06,
+      "loss": 0.954,
+      "step": 22657
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.205554811168423e-06,
+      "loss": 0.003,
+      "step": 22658
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.20015271054824e-06,
+      "loss": 0.359,
+      "step": 22659
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.194752912666806e-06,
+      "loss": 1.0056,
+      "step": 22660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.18935541761057e-06,
+      "loss": 0.3981,
+      "step": 22661
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.183960225466012e-06,
+      "loss": 0.8631,
+      "step": 22662
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.1785673363194455e-06,
+      "loss": 1.9043,
+      "step": 22663
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.173176750257286e-06,
+      "loss": 0.3886,
+      "step": 22664
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.16778846736588e-06,
+      "loss": 0.3747,
+      "step": 22665
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.162402487731443e-06,
+      "loss": 0.216,
+      "step": 22666
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.157018811440273e-06,
+      "loss": 1.1337,
+      "step": 22667
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.151637438578583e-06,
+      "loss": 0.2706,
+      "step": 22668
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.146258369232504e-06,
+      "loss": 0.7185,
+      "step": 22669
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.140881603488218e-06,
+      "loss": 0.5565,
+      "step": 22670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.135507141431839e-06,
+      "loss": 1.4309,
+      "step": 22671
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.130134983149365e-06,
+      "loss": 0.6057,
+      "step": 22672
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.124765128726877e-06,
+      "loss": 0.0036,
+      "step": 22673
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.11939757825034e-06,
+      "loss": 0.7564,
+      "step": 22674
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.114032331805718e-06,
+      "loss": 0.0012,
+      "step": 22675
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.108669389478943e-06,
+      "loss": 0.2634,
+      "step": 22676
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.103308751355845e-06,
+      "loss": 0.6719,
+      "step": 22677
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.097950417522324e-06,
+      "loss": 0.7989,
+      "step": 22678
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.092594388064159e-06,
+      "loss": 0.2772,
+      "step": 22679
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.0872406630671e-06,
+      "loss": 0.4422,
+      "step": 22680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.081889242616911e-06,
+      "loss": 0.0005,
+      "step": 22681
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.076540126799289e-06,
+      "loss": 0.2161,
+      "step": 22682
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.071193315699852e-06,
+      "loss": 0.2599,
+      "step": 22683
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.065848809404261e-06,
+      "loss": 0.9317,
+      "step": 22684
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.060506607998083e-06,
+      "loss": 0.5,
+      "step": 22685
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.055166711566884e-06,
+      "loss": 0.0269,
+      "step": 22686
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.049829120196159e-06,
+      "loss": 0.3613,
+      "step": 22687
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.044493833971342e-06,
+      "loss": 3.2325,
+      "step": 22688
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.039160852977948e-06,
+      "loss": 2.0563,
+      "step": 22689
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.033830177301341e-06,
+      "loss": 0.7416,
+      "step": 22690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.028501807026853e-06,
+      "loss": 0.9664,
+      "step": 22691
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.023175742239867e-06,
+      "loss": 0.0014,
+      "step": 22692
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.017851983025629e-06,
+      "loss": 0.2867,
+      "step": 22693
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.012530529469373e-06,
+      "loss": 1.3936,
+      "step": 22694
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.007211381656396e-06,
+      "loss": 0.5456,
+      "step": 22695
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.001894539671798e-06,
+      "loss": 0.3957,
+      "step": 22696
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.996580003600743e-06,
+      "loss": 0.2086,
+      "step": 22697
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.9912677735283635e-06,
+      "loss": 0.0006,
+      "step": 22698
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.985957849539691e-06,
+      "loss": 0.7908,
+      "step": 22699
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.980650231719758e-06,
+      "loss": 0.0011,
+      "step": 22700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.975344920153596e-06,
+      "loss": 0.1997,
+      "step": 22701
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.970041914926121e-06,
+      "loss": 0.6336,
+      "step": 22702
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.964741216122265e-06,
+      "loss": 0.1213,
+      "step": 22703
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.95944282382691e-06,
+      "loss": 0.1226,
+      "step": 22704
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.9541467381249046e-06,
+      "loss": 1.9918,
+      "step": 22705
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.948852959101047e-06,
+      "loss": 0.1055,
+      "step": 22706
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.943561486840137e-06,
+      "loss": 0.0089,
+      "step": 22707
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.9382723214268735e-06,
+      "loss": 1.425,
+      "step": 22708
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.932985462945988e-06,
+      "loss": 0.4932,
+      "step": 22709
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.927700911482114e-06,
+      "loss": 0.695,
+      "step": 22710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.922418667119916e-06,
+      "loss": 0.4736,
+      "step": 22711
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.91713872994391e-06,
+      "loss": 0.0004,
+      "step": 22712
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.911861100038712e-06,
+      "loss": 0.3844,
+      "step": 22713
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.90658577748882e-06,
+      "loss": 0.32,
+      "step": 22714
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.901312762378668e-06,
+      "loss": 0.3057,
+      "step": 22715
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.8960420547927545e-06,
+      "loss": 0.1231,
+      "step": 22716
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.890773654815478e-06,
+      "loss": 0.0014,
+      "step": 22717
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.885507562531156e-06,
+      "loss": 0.3276,
+      "step": 22718
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.880243778024152e-06,
+      "loss": 0.2329,
+      "step": 22719
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.874982301378767e-06,
+      "loss": 0.001,
+      "step": 22720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.8697231326792e-06,
+      "loss": 0.1728,
+      "step": 22721
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.864466272009749e-06,
+      "loss": 0.6648,
+      "step": 22722
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.859211719454532e-06,
+      "loss": 0.0632,
+      "step": 22723
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.8539594750977136e-06,
+      "loss": 0.6217,
+      "step": 22724
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.848709539023411e-06,
+      "loss": 0.8434,
+      "step": 22725
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.843461911315672e-06,
+      "loss": 0.9533,
+      "step": 22726
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.838216592058548e-06,
+      "loss": 0.9932,
+      "step": 22727
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.832973581336037e-06,
+      "loss": 0.6779,
+      "step": 22728
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.827732879232072e-06,
+      "loss": 1.6115,
+      "step": 22729
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.822494485830603e-06,
+      "loss": 2.9186,
+      "step": 22730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.8172584012155135e-06,
+      "loss": 0.49,
+      "step": 22731
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.812024625470635e-06,
+      "loss": 0.3077,
+      "step": 22732
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.806793158679785e-06,
+      "loss": 0.5742,
+      "step": 22733
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.801564000926728e-06,
+      "loss": 0.8476,
+      "step": 22734
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.7963371522952325e-06,
+      "loss": 0.875,
+      "step": 22735
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.791112612868964e-06,
+      "loss": 0.707,
+      "step": 22736
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.785890382731606e-06,
+      "loss": 0.3961,
+      "step": 22737
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.7806704619667745e-06,
+      "loss": 0.0084,
+      "step": 22738
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.775452850658085e-06,
+      "loss": 0.2391,
+      "step": 22739
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.7702375488890226e-06,
+      "loss": 0.1818,
+      "step": 22740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.765024556743186e-06,
+      "loss": 0.901,
+      "step": 22741
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.759813874304009e-06,
+      "loss": 0.2788,
+      "step": 22742
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.7546055016549075e-06,
+      "loss": 0.2704,
+      "step": 22743
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.749399438879348e-06,
+      "loss": 0.003,
+      "step": 22744
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.744195686060665e-06,
+      "loss": 0.0016,
+      "step": 22745
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.738994243282158e-06,
+      "loss": 0.2689,
+      "step": 22746
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.733795110627193e-06,
+      "loss": 0.0036,
+      "step": 22747
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.728598288178971e-06,
+      "loss": 0.6193,
+      "step": 22748
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.723403776020708e-06,
+      "loss": 0.0003,
+      "step": 22749
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.7182115742356535e-06,
+      "loss": 0.3478,
+      "step": 22750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.71302168290686e-06,
+      "loss": 0.3486,
+      "step": 22751
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.707834102117492e-06,
+      "loss": 0.2532,
+      "step": 22752
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.702648831950618e-06,
+      "loss": 1.9499,
+      "step": 22753
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.697465872489254e-06,
+      "loss": 0.2977,
+      "step": 22754
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.692285223816401e-06,
+      "loss": 0.1695,
+      "step": 22755
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.6871068860150415e-06,
+      "loss": 0.4453,
+      "step": 22756
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.68193085916806e-06,
+      "loss": 0.2196,
+      "step": 22757
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.676757143358373e-06,
+      "loss": 0.4595,
+      "step": 22758
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.671585738668832e-06,
+      "loss": 0.3579,
+      "step": 22759
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.666416645182253e-06,
+      "loss": 0.492,
+      "step": 22760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.661249862981354e-06,
+      "loss": 0.6072,
+      "step": 22761
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.656085392148935e-06,
+      "loss": 0.2487,
+      "step": 22762
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.6509232327676956e-06,
+      "loss": 0.0043,
+      "step": 22763
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.645763384920254e-06,
+      "loss": 0.0052,
+      "step": 22764
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.640605848689278e-06,
+      "loss": 0.2799,
+      "step": 22765
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.635450624157367e-06,
+      "loss": 0.5855,
+      "step": 22766
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.630297711407022e-06,
+      "loss": 0.3356,
+      "step": 22767
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.6251471105207945e-06,
+      "loss": 0.3745,
+      "step": 22768
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.619998821581201e-06,
+      "loss": 0.4805,
+      "step": 22769
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.6148528446705924e-06,
+      "loss": 2.0078,
+      "step": 22770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.609709179871452e-06,
+      "loss": 1.9502,
+      "step": 22771
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.604567827266116e-06,
+      "loss": 1.1083,
+      "step": 22772
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.599428786936899e-06,
+      "loss": 0.9616,
+      "step": 22773
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.594292058966137e-06,
+      "loss": 0.397,
+      "step": 22774
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.589157643436049e-06,
+      "loss": 0.2679,
+      "step": 22775
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.5840255404288826e-06,
+      "loss": 0.7681,
+      "step": 22776
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.578895750026807e-06,
+      "loss": 0.4009,
+      "step": 22777
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.573768272311973e-06,
+      "loss": 0.9303,
+      "step": 22778
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.5686431073664825e-06,
+      "loss": 0.974,
+      "step": 22779
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.563520255272402e-06,
+      "loss": 0.165,
+      "step": 22780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.558399716111783e-06,
+      "loss": 0.7666,
+      "step": 22781
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.553281489966626e-06,
+      "loss": 0.6497,
+      "step": 22782
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.5481655769188674e-06,
+      "loss": 0.4848,
+      "step": 22783
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.543051977050455e-06,
+      "loss": 0.001,
+      "step": 22784
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.53794069044326e-06,
+      "loss": 0.3269,
+      "step": 22785
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.532831717179132e-06,
+      "loss": 0.1876,
+      "step": 22786
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.527725057339904e-06,
+      "loss": 0.2258,
+      "step": 22787
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.522620711007347e-06,
+      "loss": 0.3724,
+      "step": 22788
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.517518678263161e-06,
+      "loss": 0.4169,
+      "step": 22789
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.5124189591890965e-06,
+      "loss": 0.0007,
+      "step": 22790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.5073215538667895e-06,
+      "loss": 0.6897,
+      "step": 22791
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.502226462377856e-06,
+      "loss": 0.2595,
+      "step": 22792
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.4971336848039334e-06,
+      "loss": 0.3247,
+      "step": 22793
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.492043221226538e-06,
+      "loss": 0.2542,
+      "step": 22794
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.486955071727173e-06,
+      "loss": 0.0007,
+      "step": 22795
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.481869236387371e-06,
+      "loss": 0.0157,
+      "step": 22796
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.476785715288518e-06,
+      "loss": 2.1689,
+      "step": 22797
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.471704508512048e-06,
+      "loss": 0.4934,
+      "step": 22798
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.466625616139331e-06,
+      "loss": 0.0009,
+      "step": 22799
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.4615490382517004e-06,
+      "loss": 0.0524,
+      "step": 22800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.4564747749304246e-06,
+      "loss": 0.5339,
+      "step": 22801
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.451402826256773e-06,
+      "loss": 0.3808,
+      "step": 22802
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.44633319231198e-06,
+      "loss": 0.0302,
+      "step": 22803
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.441265873177198e-06,
+      "loss": 0.0524,
+      "step": 22804
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.43620086893361e-06,
+      "loss": 0.8638,
+      "step": 22805
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.431138179662287e-06,
+      "loss": 0.0019,
+      "step": 22806
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 5.426077805444329e-06,
+      "loss": 0.4331,
+      "step": 22807
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.421019746360772e-06,
+      "loss": 0.0035,
+      "step": 22808
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.415964002492601e-06,
+      "loss": 1.0222,
+      "step": 22809
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.410910573920751e-06,
+      "loss": 0.0042,
+      "step": 22810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.405859460726175e-06,
+      "loss": 1.9639,
+      "step": 22811
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.400810662989774e-06,
+      "loss": 0.3976,
+      "step": 22812
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.39576418079235e-06,
+      "loss": 0.5467,
+      "step": 22813
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.390720014214756e-06,
+      "loss": 0.5625,
+      "step": 22814
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.385678163337775e-06,
+      "loss": 0.0015,
+      "step": 22815
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.380638628242095e-06,
+      "loss": 0.6742,
+      "step": 22816
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.375601409008451e-06,
+      "loss": 0.7329,
+      "step": 22817
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.37056650571751e-06,
+      "loss": 1.3291,
+      "step": 22818
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.36553391844986e-06,
+      "loss": 0.0065,
+      "step": 22819
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.360503647286151e-06,
+      "loss": 0.3742,
+      "step": 22820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.355475692306888e-06,
+      "loss": 0.1157,
+      "step": 22821
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.350450053592587e-06,
+      "loss": 0.4183,
+      "step": 22822
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.345426731223768e-06,
+      "loss": 0.4924,
+      "step": 22823
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.340405725280833e-06,
+      "loss": 0.9008,
+      "step": 22824
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.3353870358441855e-06,
+      "loss": 0.3507,
+      "step": 22825
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.33037066299421e-06,
+      "loss": 0.6249,
+      "step": 22826
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.325356606811243e-06,
+      "loss": 0.3004,
+      "step": 22827
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.320344867375537e-06,
+      "loss": 0.2825,
+      "step": 22828
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.315335444767393e-06,
+      "loss": 0.2098,
+      "step": 22829
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.310328339067e-06,
+      "loss": 0.328,
+      "step": 22830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.305323550354557e-06,
+      "loss": 0.5464,
+      "step": 22831
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.300321078710201e-06,
+      "loss": 0.209,
+      "step": 22832
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.295320924214036e-06,
+      "loss": 1.0796,
+      "step": 22833
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.290323086946147e-06,
+      "loss": 0.6756,
+      "step": 22834
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.285327566986536e-06,
+      "loss": 0.3157,
+      "step": 22835
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.280334364415223e-06,
+      "loss": 0.5169,
+      "step": 22836
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.275343479312177e-06,
+      "loss": 0.0235,
+      "step": 22837
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2703549117572856e-06,
+      "loss": 0.3617,
+      "step": 22838
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.265368661830482e-06,
+      "loss": 1.0125,
+      "step": 22839
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.260384729611556e-06,
+      "loss": 0.2459,
+      "step": 22840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.255403115180323e-06,
+      "loss": 0.3825,
+      "step": 22841
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2504238186166226e-06,
+      "loss": 0.3437,
+      "step": 22842
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.245446840000122e-06,
+      "loss": 1.0545,
+      "step": 22843
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.240472179410543e-06,
+      "loss": 0.5678,
+      "step": 22844
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2354998369275705e-06,
+      "loss": 0.2155,
+      "step": 22845
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.230529812630791e-06,
+      "loss": 0.0134,
+      "step": 22846
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.22556210659979e-06,
+      "loss": 0.4212,
+      "step": 22847
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.220596718914155e-06,
+      "loss": 0.536,
+      "step": 22848
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.2156336496533894e-06,
+      "loss": 0.3269,
+      "step": 22849
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.210672898896945e-06,
+      "loss": 0.3665,
+      "step": 22850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.205714466724276e-06,
+      "loss": 0.8421,
+      "step": 22851
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.200758353214784e-06,
+      "loss": 0.0477,
+      "step": 22852
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.195804558447842e-06,
+      "loss": 0.1278,
+      "step": 22853
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.190853082502766e-06,
+      "loss": 0.7656,
+      "step": 22854
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.185903925458845e-06,
+      "loss": 0.0007,
+      "step": 22855
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.180957087395349e-06,
+      "loss": 0.2648,
+      "step": 22856
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.176012568391463e-06,
+      "loss": 0.1487,
+      "step": 22857
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.1710703685264085e-06,
+      "loss": 0.4928,
+      "step": 22858
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.1661304878792885e-06,
+      "loss": 1.0822,
+      "step": 22859
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.1611929265292226e-06,
+      "loss": 0.8402,
+      "step": 22860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.156257684555298e-06,
+      "loss": 0.0747,
+      "step": 22861
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.151324762036502e-06,
+      "loss": 0.2071,
+      "step": 22862
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.146394159051853e-06,
+      "loss": 0.3542,
+      "step": 22863
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.141465875680339e-06,
+      "loss": 0.289,
+      "step": 22864
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.136539912000798e-06,
+      "loss": 0.7597,
+      "step": 22865
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.131616268092198e-06,
+      "loss": 0.0037,
+      "step": 22866
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.126694944033344e-06,
+      "loss": 0.4844,
+      "step": 22867
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.1217759399030226e-06,
+      "loss": 0.3127,
+      "step": 22868
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.116859255780054e-06,
+      "loss": 0.3588,
+      "step": 22869
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.111944891743141e-06,
+      "loss": 0.2847,
+      "step": 22870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.107032847870956e-06,
+      "loss": 0.8364,
+      "step": 22871
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.102123124242219e-06,
+      "loss": 0.5063,
+      "step": 22872
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.097215720935516e-06,
+      "loss": 0.0769,
+      "step": 22873
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.092310638029418e-06,
+      "loss": 1.2503,
+      "step": 22874
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.087407875602495e-06,
+      "loss": 0.1618,
+      "step": 22875
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.082507433733251e-06,
+      "loss": 0.3596,
+      "step": 22876
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.077609312500175e-06,
+      "loss": 0.237,
+      "step": 22877
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.072713511981669e-06,
+      "loss": 0.0784,
+      "step": 22878
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.067820032256154e-06,
+      "loss": 0.3152,
+      "step": 22879
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.0629288734019835e-06,
+      "loss": 0.7557,
+      "step": 22880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.058040035497496e-06,
+      "loss": 0.0012,
+      "step": 22881
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.053153518620978e-06,
+      "loss": 0.4734,
+      "step": 22882
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.048269322850651e-06,
+      "loss": 0.3294,
+      "step": 22883
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.043387448264768e-06,
+      "loss": 0.5189,
+      "step": 22884
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.038507894941468e-06,
+      "loss": 0.1809,
+      "step": 22885
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.033630662958921e-06,
+      "loss": 0.7704,
+      "step": 22886
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.028755752395214e-06,
+      "loss": 0.254,
+      "step": 22887
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.023883163328418e-06,
+      "loss": 0.8035,
+      "step": 22888
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.019012895836539e-06,
+      "loss": 0.2114,
+      "step": 22889
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.014144949997578e-06,
+      "loss": 0.0014,
+      "step": 22890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.009279325889526e-06,
+      "loss": 0.2901,
+      "step": 22891
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 5.0044160235902355e-06,
+      "loss": 0.3359,
+      "step": 22892
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.999555043177611e-06,
+      "loss": 0.3697,
+      "step": 22893
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.994696384729524e-06,
+      "loss": 0.4136,
+      "step": 22894
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.989840048323729e-06,
+      "loss": 0.3234,
+      "step": 22895
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.984986034038013e-06,
+      "loss": 0.2308,
+      "step": 22896
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.980134341950132e-06,
+      "loss": 0.0184,
+      "step": 22897
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.975284972137738e-06,
+      "loss": 0.0036,
+      "step": 22898
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.970437924678505e-06,
+      "loss": 0.7296,
+      "step": 22899
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.965593199650053e-06,
+      "loss": 0.3275,
+      "step": 22900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.960750797129953e-06,
+      "loss": 0.2786,
+      "step": 22901
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.95591071719576e-06,
+      "loss": 0.2313,
+      "step": 22902
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.951072959924978e-06,
+      "loss": 0.0042,
+      "step": 22903
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.9462375253950806e-06,
+      "loss": 1.4762,
+      "step": 22904
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.941404413683487e-06,
+      "loss": 0.2397,
+      "step": 22905
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.936573624867585e-06,
+      "loss": 0.0145,
+      "step": 22906
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.9317451590247805e-06,
+      "loss": 0.0042,
+      "step": 22907
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.926919016232328e-06,
+      "loss": 0.2916,
+      "step": 22908
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.922095196567549e-06,
+      "loss": 0.3473,
+      "step": 22909
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.917273700107716e-06,
+      "loss": 0.7258,
+      "step": 22910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.9124545269299655e-06,
+      "loss": 0.3125,
+      "step": 22911
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.90763767711152e-06,
+      "loss": 0.0012,
+      "step": 22912
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.9028231507295175e-06,
+      "loss": 0.3923,
+      "step": 22913
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.8980109478610144e-06,
+      "loss": 0.7112,
+      "step": 22914
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.893201068583131e-06,
+      "loss": 0.0029,
+      "step": 22915
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.8883935129728544e-06,
+      "loss": 0.4093,
+      "step": 22916
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.883588281107142e-06,
+      "loss": 0.5481,
+      "step": 22917
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.878785373063016e-06,
+      "loss": 0.1843,
+      "step": 22918
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.873984788917329e-06,
+      "loss": 2.0936,
+      "step": 22919
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.8691865287469375e-06,
+      "loss": 0.3506,
+      "step": 22920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.864390592628764e-06,
+      "loss": 0.2272,
+      "step": 22921
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.859596980639546e-06,
+      "loss": 0.2949,
+      "step": 22922
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.854805692856023e-06,
+      "loss": 0.662,
+      "step": 22923
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.850016729354999e-06,
+      "loss": 0.4819,
+      "step": 22924
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.845230090213098e-06,
+      "loss": 0.2116,
+      "step": 22925
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.840445775506991e-06,
+      "loss": 0.0004,
+      "step": 22926
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.8356637853132814e-06,
+      "loss": 0.4863,
+      "step": 22927
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.830884119708561e-06,
+      "loss": 0.0007,
+      "step": 22928
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.826106778769368e-06,
+      "loss": 0.4099,
+      "step": 22929
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.821331762572189e-06,
+      "loss": 0.071,
+      "step": 22930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.816559071193516e-06,
+      "loss": 0.0015,
+      "step": 22931
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.811788704709752e-06,
+      "loss": 0.149,
+      "step": 22932
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.807020663197287e-06,
+      "loss": 0.002,
+      "step": 22933
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.8022549467324925e-06,
+      "loss": 0.539,
+      "step": 22934
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.797491555391675e-06,
+      "loss": 0.3412,
+      "step": 22935
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.792730489251106e-06,
+      "loss": 0.8726,
+      "step": 22936
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.7879717483870425e-06,
+      "loss": 0.4492,
+      "step": 22937
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.783215332875656e-06,
+      "loss": 0.3755,
+      "step": 22938
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.778461242793152e-06,
+      "loss": 0.4627,
+      "step": 22939
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.773709478215654e-06,
+      "loss": 0.4965,
+      "step": 22940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.768960039219233e-06,
+      "loss": 0.7452,
+      "step": 22941
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.7642129258799286e-06,
+      "loss": 0.3801,
+      "step": 22942
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.759468138273831e-06,
+      "loss": 0.0039,
+      "step": 22943
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.754725676476862e-06,
+      "loss": 0.3156,
+      "step": 22944
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.74998554056496e-06,
+      "loss": 0.3709,
+      "step": 22945
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.745247730614066e-06,
+      "loss": 0.002,
+      "step": 22946
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.740512246700034e-06,
+      "loss": 0.6572,
+      "step": 22947
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.735779088898689e-06,
+      "loss": 0.7891,
+      "step": 22948
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.731048257285836e-06,
+      "loss": 0.1946,
+      "step": 22949
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.726319751937229e-06,
+      "loss": 0.1768,
+      "step": 22950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.721593572928578e-06,
+      "loss": 0.0089,
+      "step": 22951
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.71686972033557e-06,
+      "loss": 0.1897,
+      "step": 22952
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.712148194233861e-06,
+      "loss": 0.4056,
+      "step": 22953
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.707428994699042e-06,
+      "loss": 0.0011,
+      "step": 22954
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.702712121806718e-06,
+      "loss": 0.4925,
+      "step": 22955
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.697997575632379e-06,
+      "loss": 2.1777,
+      "step": 22956
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.693285356251547e-06,
+      "loss": 0.5088,
+      "step": 22957
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.688575463739663e-06,
+      "loss": 0.7199,
+      "step": 22958
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.6838678981722e-06,
+      "loss": 0.0682,
+      "step": 22959
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.679162659624447e-06,
+      "loss": 0.1509,
+      "step": 22960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.6744597481718435e-06,
+      "loss": 0.8245,
+      "step": 22961
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.669759163889664e-06,
+      "loss": 0.4655,
+      "step": 22962
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.665060906853163e-06,
+      "loss": 0.7581,
+      "step": 22963
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.660364977137599e-06,
+      "loss": 0.0122,
+      "step": 22964
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.655671374818193e-06,
+      "loss": 0.7758,
+      "step": 22965
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.6509800999700365e-06,
+      "loss": 0.1038,
+      "step": 22966
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.646291152668335e-06,
+      "loss": 0.0031,
+      "step": 22967
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.6416045329881136e-06,
+      "loss": 0.0909,
+      "step": 22968
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.636920241004444e-06,
+      "loss": 0.4192,
+      "step": 22969
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.632238276792349e-06,
+      "loss": 0.1395,
+      "step": 22970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.627558640426787e-06,
+      "loss": 0.5556,
+      "step": 22971
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.622881331982681e-06,
+      "loss": 1.3047,
+      "step": 22972
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.61820635153497e-06,
+      "loss": 0.0044,
+      "step": 22973
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.613533699158495e-06,
+      "loss": 0.5489,
+      "step": 22974
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.608863374928079e-06,
+      "loss": 0.3076,
+      "step": 22975
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.604195378918512e-06,
+      "loss": 0.3301,
+      "step": 22976
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.599529711204553e-06,
+      "loss": 1.5778,
+      "step": 22977
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.5948663718609056e-06,
+      "loss": 3.3433,
+      "step": 22978
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.590205360962246e-06,
+      "loss": 0.712,
+      "step": 22979
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.585546678583229e-06,
+      "loss": 0.2167,
+      "step": 22980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.580890324798448e-06,
+      "loss": 0.0014,
+      "step": 22981
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.576236299682456e-06,
+      "loss": 0.5266,
+      "step": 22982
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.57158460330978e-06,
+      "loss": 0.0202,
+      "step": 22983
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.566935235754943e-06,
+      "loss": 0.0276,
+      "step": 22984
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.562288197092351e-06,
+      "loss": 0.3361,
+      "step": 22985
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.5576434873964615e-06,
+      "loss": 0.181,
+      "step": 22986
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.553001106741616e-06,
+      "loss": 0.0073,
+      "step": 22987
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.548361055202171e-06,
+      "loss": 0.2383,
+      "step": 22988
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.543723332852467e-06,
+      "loss": 0.0003,
+      "step": 22989
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.539087939766694e-06,
+      "loss": 0.8802,
+      "step": 22990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.534454876019128e-06,
+      "loss": 0.4683,
+      "step": 22991
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.5298241416839745e-06,
+      "loss": 0.2284,
+      "step": 22992
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.5251957368353575e-06,
+      "loss": 0.0712,
+      "step": 22993
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.520569661547385e-06,
+      "loss": 0.6967,
+      "step": 22994
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.515945915894181e-06,
+      "loss": 0.3193,
+      "step": 22995
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.511324499949753e-06,
+      "loss": 1.1538,
+      "step": 22996
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.506705413788109e-06,
+      "loss": 0.2954,
+      "step": 22997
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.5020886574832385e-06,
+      "loss": 1.2497,
+      "step": 22998
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.4974742311090326e-06,
+      "loss": 0.0034,
+      "step": 22999
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.492862134739433e-06,
+      "loss": 0.5755,
+      "step": 23000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.488252368448247e-06,
+      "loss": 0.694,
+      "step": 23001
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.483644932309333e-06,
+      "loss": 0.4134,
+      "step": 23002
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.479039826396447e-06,
+      "loss": 0.3212,
+      "step": 23003
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.474437050783347e-06,
+      "loss": 0.0032,
+      "step": 23004
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.4698366055437414e-06,
+      "loss": 0.531,
+      "step": 23005
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.465238490751288e-06,
+      "loss": 0.0024,
+      "step": 23006
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.460642706479611e-06,
+      "loss": 0.3168,
+      "step": 23007
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.456049252802352e-06,
+      "loss": 0.0003,
+      "step": 23008
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.451458129792984e-06,
+      "loss": 0.2661,
+      "step": 23009
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.4468693375251e-06,
+      "loss": 0.6359,
+      "step": 23010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.442282876072173e-06,
+      "loss": 0.3238,
+      "step": 23011
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.437698745507595e-06,
+      "loss": 0.867,
+      "step": 23012
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.433116945904841e-06,
+      "loss": 0.3608,
+      "step": 23013
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.4285374773372685e-06,
+      "loss": 0.4583,
+      "step": 23014
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.423960339878169e-06,
+      "loss": 1.559,
+      "step": 23015
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.419385533600883e-06,
+      "loss": 0.5436,
+      "step": 23016
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.414813058578636e-06,
+      "loss": 0.4422,
+      "step": 23017
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.410242914884671e-06,
+      "loss": 0.4893,
+      "step": 23018
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.4056751025921765e-06,
+      "loss": 0.7412,
+      "step": 23019
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.401109621774279e-06,
+      "loss": 0.0018,
+      "step": 23020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.3965464725040864e-06,
+      "loss": 0.492,
+      "step": 23021
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.391985654854707e-06,
+      "loss": 0.1097,
+      "step": 23022
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.387427168899132e-06,
+      "loss": 0.676,
+      "step": 23023
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.382871014710387e-06,
+      "loss": 0.4171,
+      "step": 23024
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.3783171923614135e-06,
+      "loss": 0.0507,
+      "step": 23025
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.373765701925153e-06,
+      "loss": 0.449,
+      "step": 23026
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.369216543474463e-06,
+      "loss": 0.5339,
+      "step": 23027
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.36466971708222e-06,
+      "loss": 0.0022,
+      "step": 23028
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.360125222821232e-06,
+      "loss": 0.3259,
+      "step": 23029
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.355583060764256e-06,
+      "loss": 0.826,
+      "step": 23030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.351043230984036e-06,
+      "loss": 0.2014,
+      "step": 23031
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.346505733553263e-06,
+      "loss": 2.4494,
+      "step": 23032
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.341970568544612e-06,
+      "loss": 0.3388,
+      "step": 23033
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.337437736030691e-06,
+      "loss": 0.3105,
+      "step": 23034
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.3329072360841265e-06,
+      "loss": 1.1364,
+      "step": 23035
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.328379068777393e-06,
+      "loss": 0.7183,
+      "step": 23036
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.3238532341830665e-06,
+      "loss": 0.2291,
+      "step": 23037
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.319329732373622e-06,
+      "loss": 0.0008,
+      "step": 23038
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.314808563421435e-06,
+      "loss": 0.4888,
+      "step": 23039
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.310289727398963e-06,
+      "loss": 0.4645,
+      "step": 23040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.305773224378584e-06,
+      "loss": 0.5741,
+      "step": 23041
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.301259054432538e-06,
+      "loss": 0.7482,
+      "step": 23042
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.296747217633184e-06,
+      "loss": 0.8803,
+      "step": 23043
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2922377140527816e-06,
+      "loss": 0.3476,
+      "step": 23044
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.28773054376349e-06,
+      "loss": 0.8096,
+      "step": 23045
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.283225706837517e-06,
+      "loss": 0.6374,
+      "step": 23046
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.27872320334699e-06,
+      "loss": 0.4116,
+      "step": 23047
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.274223033364016e-06,
+      "loss": 0.8386,
+      "step": 23048
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.269725196960638e-06,
+      "loss": 3.6146,
+      "step": 23049
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.265229694208916e-06,
+      "loss": 0.1214,
+      "step": 23050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2607365251808246e-06,
+      "loss": 0.2596,
+      "step": 23051
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.256245689948323e-06,
+      "loss": 0.0082,
+      "step": 23052
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.251757188583305e-06,
+      "loss": 0.2824,
+      "step": 23053
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.247271021157661e-06,
+      "loss": 0.0305,
+      "step": 23054
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.2427871877432185e-06,
+      "loss": 0.3778,
+      "step": 23055
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.238305688411786e-06,
+      "loss": 0.2241,
+      "step": 23056
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.2338265232351395e-06,
+      "loss": 0.495,
+      "step": 23057
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.229349692285006e-06,
+      "loss": 0.001,
+      "step": 23058
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.22487519563306e-06,
+      "loss": 0.6196,
+      "step": 23059
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.220403033350961e-06,
+      "loss": 0.2889,
+      "step": 23060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.215933205510302e-06,
+      "loss": 0.7192,
+      "step": 23061
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.2114657121826935e-06,
+      "loss": 0.5373,
+      "step": 23062
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.2070005534396934e-06,
+      "loss": 0.7466,
+      "step": 23063
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.202537729352728e-06,
+      "loss": 0.4628,
+      "step": 23064
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.198077239993342e-06,
+      "loss": 0.2497,
+      "step": 23065
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.193619085432926e-06,
+      "loss": 0.5815,
+      "step": 23066
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.189163265742856e-06,
+      "loss": 0.1322,
+      "step": 23067
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.184709780994527e-06,
+      "loss": 0.0767,
+      "step": 23068
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.180258631259231e-06,
+      "loss": 0.4881,
+      "step": 23069
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.175809816608228e-06,
+      "loss": 0.3318,
+      "step": 23070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.171363337112793e-06,
+      "loss": 0.7488,
+      "step": 23071
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.16691919284412e-06,
+      "loss": 0.2119,
+      "step": 23072
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.162477383873369e-06,
+      "loss": 0.6456,
+      "step": 23073
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.158037910271667e-06,
+      "loss": 0.7123,
+      "step": 23074
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.153600772110105e-06,
+      "loss": 0.0014,
+      "step": 23075
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.1491659694597616e-06,
+      "loss": 0.4471,
+      "step": 23076
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.144733502391611e-06,
+      "loss": 0.0293,
+      "step": 23077
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.140303370976666e-06,
+      "loss": 2.3891,
+      "step": 23078
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.135875575285852e-06,
+      "loss": 0.4058,
+      "step": 23079
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.131450115390095e-06,
+      "loss": 0.3568,
+      "step": 23080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.127026991360222e-06,
+      "loss": 0.1692,
+      "step": 23081
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.122606203267109e-06,
+      "loss": 0.3769,
+      "step": 23082
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.118187751181517e-06,
+      "loss": 0.4844,
+      "step": 23083
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.113771635174224e-06,
+      "loss": 0.2994,
+      "step": 23084
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.109357855315904e-06,
+      "loss": 0.2596,
+      "step": 23085
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.104946411677284e-06,
+      "loss": 1.6078,
+      "step": 23086
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.10053730432901e-06,
+      "loss": 0.2199,
+      "step": 23087
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.096130533341624e-06,
+      "loss": 0.0049,
+      "step": 23088
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.091726098785752e-06,
+      "loss": 0.3287,
+      "step": 23089
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.0873240007319394e-06,
+      "loss": 0.2678,
+      "step": 23090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.082924239250612e-06,
+      "loss": 0.3557,
+      "step": 23091
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.078526814412264e-06,
+      "loss": 0.5459,
+      "step": 23092
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.074131726287339e-06,
+      "loss": 0.0027,
+      "step": 23093
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.069738974946179e-06,
+      "loss": 0.6666,
+      "step": 23094
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.065348560459131e-06,
+      "loss": 0.2413,
+      "step": 23095
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.060960482896502e-06,
+      "loss": 0.0263,
+      "step": 23096
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.056574742328572e-06,
+      "loss": 0.5912,
+      "step": 23097
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.052191338825583e-06,
+      "loss": 0.6488,
+      "step": 23098
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.047810272457696e-06,
+      "loss": 0.741,
+      "step": 23099
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.043431543295089e-06,
+      "loss": 0.3446,
+      "step": 23100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.039055151407871e-06,
+      "loss": 0.2998,
+      "step": 23101
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.034681096866138e-06,
+      "loss": 0.6619,
+      "step": 23102
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.030309379739916e-06,
+      "loss": 0.633,
+      "step": 23103
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.025940000099215e-06,
+      "loss": 0.5037,
+      "step": 23104
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.0215729580140134e-06,
+      "loss": 0.0015,
+      "step": 23105
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.017208253554239e-06,
+      "loss": 0.4135,
+      "step": 23106
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.012845886789784e-06,
+      "loss": 0.4192,
+      "step": 23107
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.008485857790511e-06,
+      "loss": 0.4566,
+      "step": 23108
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.00412816662623e-06,
+      "loss": 0.266,
+      "step": 23109
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.999772813366719e-06,
+      "loss": 0.4122,
+      "step": 23110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.9954197980817395e-06,
+      "loss": 0.7386,
+      "step": 23111
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.991069120841017e-06,
+      "loss": 0.0228,
+      "step": 23112
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.986720781714148e-06,
+      "loss": 0.3225,
+      "step": 23113
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.982374780770842e-06,
+      "loss": 0.7412,
+      "step": 23114
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.978031118080644e-06,
+      "loss": 1.3353,
+      "step": 23115
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.973689793713131e-06,
+      "loss": 0.3711,
+      "step": 23116
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.969350807737831e-06,
+      "loss": 0.2677,
+      "step": 23117
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.965014160224206e-06,
+      "loss": 0.0091,
+      "step": 23118
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.960679851241699e-06,
+      "loss": 0.4883,
+      "step": 23119
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.956347880859773e-06,
+      "loss": 0.4321,
+      "step": 23120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.95201824914772e-06,
+      "loss": 1.5102,
+      "step": 23121
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.94769095617492e-06,
+      "loss": 0.002,
+      "step": 23122
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.943366002010651e-06,
+      "loss": 0.4629,
+      "step": 23123
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.939043386724172e-06,
+      "loss": 0.0016,
+      "step": 23124
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.934723110384713e-06,
+      "loss": 0.3103,
+      "step": 23125
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.930405173061452e-06,
+      "loss": 0.5212,
+      "step": 23126
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.926089574823515e-06,
+      "loss": 0.2093,
+      "step": 23127
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.921776315740049e-06,
+      "loss": 0.0694,
+      "step": 23128
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.917465395880082e-06,
+      "loss": 0.1051,
+      "step": 23129
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.9131568153126905e-06,
+      "loss": 0.0007,
+      "step": 23130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.90885057410682e-06,
+      "loss": 0.9273,
+      "step": 23131
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.9045466723314825e-06,
+      "loss": 0.4056,
+      "step": 23132
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.900245110055572e-06,
+      "loss": 0.9336,
+      "step": 23133
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.895945887347934e-06,
+      "loss": 2.2109,
+      "step": 23134
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.891649004277481e-06,
+      "loss": 0.2846,
+      "step": 23135
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.887354460913006e-06,
+      "loss": 0.3544,
+      "step": 23136
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.883062257323222e-06,
+      "loss": 0.5336,
+      "step": 23137
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.878772393576924e-06,
+      "loss": 2.3169,
+      "step": 23138
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8744848697428225e-06,
+      "loss": 0.2822,
+      "step": 23139
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.870199685889497e-06,
+      "loss": 1.5444,
+      "step": 23140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.865916842085625e-06,
+      "loss": 0.014,
+      "step": 23141
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.861636338399788e-06,
+      "loss": 0.8281,
+      "step": 23142
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.85735817490051e-06,
+      "loss": 0.4678,
+      "step": 23143
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.853082351656322e-06,
+      "loss": 0.5579,
+      "step": 23144
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.848808868735687e-06,
+      "loss": 0.0285,
+      "step": 23145
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.844537726207031e-06,
+      "loss": 0.619,
+      "step": 23146
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.840268924138751e-06,
+      "loss": 0.1333,
+      "step": 23147
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.836002462599208e-06,
+      "loss": 0.8457,
+      "step": 23148
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.831738341656732e-06,
+      "loss": 0.2294,
+      "step": 23149
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.827476561379584e-06,
+      "loss": 0.1474,
+      "step": 23150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.823217121836042e-06,
+      "loss": 1.4414,
+      "step": 23151
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.818960023094286e-06,
+      "loss": 0.4226,
+      "step": 23152
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.814705265222495e-06,
+      "loss": 0.1882,
+      "step": 23153
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8104528482888143e-06,
+      "loss": 0.5409,
+      "step": 23154
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8062027723613218e-06,
+      "loss": 0.323,
+      "step": 23155
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.8019550375080974e-06,
+      "loss": 0.7724,
+      "step": 23156
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7977096437971533e-06,
+      "loss": 0.0079,
+      "step": 23157
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.793466591296451e-06,
+      "loss": 0.0033,
+      "step": 23158
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7892258800739705e-06,
+      "loss": 0.3119,
+      "step": 23159
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.784987510197607e-06,
+      "loss": 0.3544,
+      "step": 23160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.780751481735239e-06,
+      "loss": 0.1824,
+      "step": 23161
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7765177947546633e-06,
+      "loss": 0.0512,
+      "step": 23162
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7722864493237417e-06,
+      "loss": 0.4705,
+      "step": 23163
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7680574455101697e-06,
+      "loss": 0.8889,
+      "step": 23164
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.763830783381694e-06,
+      "loss": 0.5884,
+      "step": 23165
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7596064630060095e-06,
+      "loss": 0.3864,
+      "step": 23166
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.755384484450746e-06,
+      "loss": 0.2509,
+      "step": 23167
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7511648477835e-06,
+      "loss": 0.2479,
+      "step": 23168
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7469475530718996e-06,
+      "loss": 0.2637,
+      "step": 23169
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.742732600383408e-06,
+      "loss": 0.5593,
+      "step": 23170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7385199897855712e-06,
+      "loss": 0.3694,
+      "step": 23171
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7343097213458185e-06,
+      "loss": 0.0227,
+      "step": 23172
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7301017951315793e-06,
+      "loss": 0.0034,
+      "step": 23173
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7258962112102333e-06,
+      "loss": 0.1867,
+      "step": 23174
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.721692969649143e-06,
+      "loss": 0.07,
+      "step": 23175
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.717492070515604e-06,
+      "loss": 0.0319,
+      "step": 23176
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.713293513876897e-06,
+      "loss": 0.0008,
+      "step": 23177
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.709097299800251e-06,
+      "loss": 0.2742,
+      "step": 23178
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.7049034283528456e-06,
+      "loss": 0.093,
+      "step": 23179
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.70071189960186e-06,
+      "loss": 0.3741,
+      "step": 23180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6965227136144083e-06,
+      "loss": 0.3449,
+      "step": 23181
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6923358704575855e-06,
+      "loss": 0.4376,
+      "step": 23182
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6881513701984057e-06,
+      "loss": 0.4324,
+      "step": 23183
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6839692129039145e-06,
+      "loss": 0.195,
+      "step": 23184
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6797893986410754e-06,
+      "loss": 0.0006,
+      "step": 23185
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.675611927476785e-06,
+      "loss": 1.1565,
+      "step": 23186
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6714367994779725e-06,
+      "loss": 0.595,
+      "step": 23187
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6672640147115184e-06,
+      "loss": 0.1378,
+      "step": 23188
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6630935732441856e-06,
+      "loss": 0.3948,
+      "step": 23189
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6589254751428043e-06,
+      "loss": 0.0112,
+      "step": 23190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.654759720474104e-06,
+      "loss": 0.3259,
+      "step": 23191
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.650596309304782e-06,
+      "loss": 1.5263,
+      "step": 23192
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6464352417015506e-06,
+      "loss": 0.4797,
+      "step": 23193
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.642276517730991e-06,
+      "loss": 1.1272,
+      "step": 23194
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6381201374596992e-06,
+      "loss": 0.3542,
+      "step": 23195
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6339661009542886e-06,
+      "loss": 0.0025,
+      "step": 23196
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6298144082812396e-06,
+      "loss": 0.5079,
+      "step": 23197
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6256650595070325e-06,
+      "loss": 0.0016,
+      "step": 23198
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6215180546981305e-06,
+      "loss": 0.0014,
+      "step": 23199
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6173733939209305e-06,
+      "loss": 3.4172,
+      "step": 23200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.613231077241796e-06,
+      "loss": 0.3795,
+      "step": 23201
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6090911047270908e-06,
+      "loss": 0.4369,
+      "step": 23202
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6049534764430787e-06,
+      "loss": 0.4862,
+      "step": 23203
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6008181924560233e-06,
+      "loss": 0.1834,
+      "step": 23204
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5966852528321545e-06,
+      "loss": 0.3805,
+      "step": 23205
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.592554657637653e-06,
+      "loss": 0.4398,
+      "step": 23206
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5884264069386494e-06,
+      "loss": 0.3014,
+      "step": 23207
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5843005008012735e-06,
+      "loss": 0.0507,
+      "step": 23208
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.58017693929159e-06,
+      "loss": 0.001,
+      "step": 23209
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5760557224756294e-06,
+      "loss": 0.3708,
+      "step": 23210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.571936850419371e-06,
+      "loss": 0.3098,
+      "step": 23211
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5678203231888135e-06,
+      "loss": 0.3446,
+      "step": 23212
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5637061408498367e-06,
+      "loss": 0.6704,
+      "step": 23213
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.559594303468305e-06,
+      "loss": 0.0007,
+      "step": 23214
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.555484811110132e-06,
+      "loss": 0.7599,
+      "step": 23215
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5513776638410817e-06,
+      "loss": 0.3411,
+      "step": 23216
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.547272861726902e-06,
+      "loss": 0.2518,
+      "step": 23217
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.54317040483339e-06,
+      "loss": 0.0039,
+      "step": 23218
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5390702932261928e-06,
+      "loss": 0.2612,
+      "step": 23219
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5349725269709582e-06,
+      "loss": 0.0117,
+      "step": 23220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5308771061333506e-06,
+      "loss": 0.5376,
+      "step": 23221
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5267840307789007e-06,
+      "loss": 0.0017,
+      "step": 23222
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.522693300973206e-06,
+      "loss": 0.664,
+      "step": 23223
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.518604916781731e-06,
+      "loss": 0.0698,
+      "step": 23224
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5145188782699562e-06,
+      "loss": 0.7515,
+      "step": 23225
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5104351855033298e-06,
+      "loss": 0.4907,
+      "step": 23226
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5063538385472155e-06,
+      "loss": 0.8489,
+      "step": 23227
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.5022748374670116e-06,
+      "loss": 0.3033,
+      "step": 23228
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4981981823279993e-06,
+      "loss": 0.1979,
+      "step": 23229
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4941238731954755e-06,
+      "loss": 1.9609,
+      "step": 23230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.490051910134706e-06,
+      "loss": 0.4362,
+      "step": 23231
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4859822932108373e-06,
+      "loss": 0.3988,
+      "step": 23232
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.481915022489101e-06,
+      "loss": 0.1722,
+      "step": 23233
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4778500980345956e-06,
+      "loss": 0.0009,
+      "step": 23234
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4737875199124186e-06,
+      "loss": 0.0269,
+      "step": 23235
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.469727288187618e-06,
+      "loss": 0.2795,
+      "step": 23236
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4656694029252584e-06,
+      "loss": 0.0767,
+      "step": 23237
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4616138641902545e-06,
+      "loss": 0.5969,
+      "step": 23238
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.457560672047588e-06,
+      "loss": 0.5328,
+      "step": 23239
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.453509826562173e-06,
+      "loss": 0.5773,
+      "step": 23240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4494613277988416e-06,
+      "loss": 0.0199,
+      "step": 23241
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4454151758224747e-06,
+      "loss": 0.1555,
+      "step": 23242
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4413713706978043e-06,
+      "loss": 0.2844,
+      "step": 23243
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.437329912489628e-06,
+      "loss": 0.501,
+      "step": 23244
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4332908012626613e-06,
+      "loss": 0.2903,
+      "step": 23245
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4292540370815692e-06,
+      "loss": 0.0005,
+      "step": 23246
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.425219620011016e-06,
+      "loss": 0.0017,
+      "step": 23247
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4211875501155674e-06,
+      "loss": 1.0193,
+      "step": 23248
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4171578274598378e-06,
+      "loss": 0.3537,
+      "step": 23249
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4131304521083257e-06,
+      "loss": 0.1815,
+      "step": 23250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.40910542412553e-06,
+      "loss": 0.3827,
+      "step": 23251
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4050827435759155e-06,
+      "loss": 0.2233,
+      "step": 23252
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.4010624105238972e-06,
+      "loss": 0.8238,
+      "step": 23253
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.39704442503384e-06,
+      "loss": 0.985,
+      "step": 23254
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3930287871700934e-06,
+      "loss": 0.3648,
+      "step": 23255
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3890154969969717e-06,
+      "loss": 0.4584,
+      "step": 23256
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3850045545787408e-06,
+      "loss": 0.0032,
+      "step": 23257
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3809959599796154e-06,
+      "loss": 0.2764,
+      "step": 23258
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3769897132637947e-06,
+      "loss": 0.3691,
+      "step": 23259
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3729858144954268e-06,
+      "loss": 0.144,
+      "step": 23260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.368984263738661e-06,
+      "loss": 0.0767,
+      "step": 23261
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3649850610575126e-06,
+      "loss": 1.3457,
+      "step": 23262
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3609882065160797e-06,
+      "loss": 0.2059,
+      "step": 23263
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.356993700178362e-06,
+      "loss": 0.0026,
+      "step": 23264
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3530015421082913e-06,
+      "loss": 0.4248,
+      "step": 23265
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3490117323697996e-06,
+      "loss": 0.2399,
+      "step": 23266
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.345024271026836e-06,
+      "loss": 0.2242,
+      "step": 23267
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3410391581431827e-06,
+      "loss": 0.0111,
+      "step": 23268
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.337056393782689e-06,
+      "loss": 0.001,
+      "step": 23269
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.33307597800912e-06,
+      "loss": 1.7803,
+      "step": 23270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3290979108862246e-06,
+      "loss": 0.6989,
+      "step": 23271
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3251221924777026e-06,
+      "loss": 0.5615,
+      "step": 23272
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3211488228472194e-06,
+      "loss": 0.0074,
+      "step": 23273
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.317177802058407e-06,
+      "loss": 0.291,
+      "step": 23274
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3132091301748485e-06,
+      "loss": 0.3151,
+      "step": 23275
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.3092428072601094e-06,
+      "loss": 0.6362,
+      "step": 23276
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.305278833377689e-06,
+      "loss": 0.0011,
+      "step": 23277
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.30131720859107e-06,
+      "loss": 0.3318,
+      "step": 23278
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2973579329636845e-06,
+      "loss": 0.0007,
+      "step": 23279
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2934010065589657e-06,
+      "loss": 0.0013,
+      "step": 23280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.289446429440229e-06,
+      "loss": 0.0019,
+      "step": 23281
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.285494201670841e-06,
+      "loss": 0.5887,
+      "step": 23282
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.281544323314084e-06,
+      "loss": 0.2661,
+      "step": 23283
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2775967944331737e-06,
+      "loss": 0.4856,
+      "step": 23284
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2736516150913606e-06,
+      "loss": 0.3921,
+      "step": 23285
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.269708785351827e-06,
+      "loss": 0.1293,
+      "step": 23286
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2657683052776716e-06,
+      "loss": 0.0025,
+      "step": 23287
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2618301749320287e-06,
+      "loss": 0.3874,
+      "step": 23288
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2578943943779634e-06,
+      "loss": 0.6277,
+      "step": 23289
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.253960963678476e-06,
+      "loss": 0.0011,
+      "step": 23290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2500298828965822e-06,
+      "loss": 0.3384,
+      "step": 23291
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.246101152095215e-06,
+      "loss": 0.2755,
+      "step": 23292
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2421747713372747e-06,
+      "loss": 0.3684,
+      "step": 23293
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2382507406856774e-06,
+      "loss": 0.6764,
+      "step": 23294
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2343290602032223e-06,
+      "loss": 0.4131,
+      "step": 23295
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2304097299527097e-06,
+      "loss": 0.0145,
+      "step": 23296
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2264927499969563e-06,
+      "loss": 1.4806,
+      "step": 23297
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2225781203986113e-06,
+      "loss": 1.3437,
+      "step": 23298
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.218665841220408e-06,
+      "loss": 1.6689,
+      "step": 23299
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.21475591252498e-06,
+      "loss": 0.4158,
+      "step": 23300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2108483343749434e-06,
+      "loss": 0.9208,
+      "step": 23301
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.2069431068328655e-06,
+      "loss": 0.3552,
+      "step": 23302
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.203040229961296e-06,
+      "loss": 0.8128,
+      "step": 23303
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1991397038227173e-06,
+      "loss": 0.7691,
+      "step": 23304
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.1952415284795975e-06,
+      "loss": 0.306,
+      "step": 23305
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.191345703994369e-06,
+      "loss": 0.2744,
+      "step": 23306
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1874522304294158e-06,
+      "loss": 0.4133,
+      "step": 23307
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1835611078470714e-06,
+      "loss": 0.0021,
+      "step": 23308
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.179672336309652e-06,
+      "loss": 0.3032,
+      "step": 23309
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1757859158794586e-06,
+      "loss": 0.6196,
+      "step": 23310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1719018466186574e-06,
+      "loss": 0.219,
+      "step": 23311
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1680201285895156e-06,
+      "loss": 0.0848,
+      "step": 23312
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1641407618541835e-06,
+      "loss": 0.0145,
+      "step": 23313
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.160263746474745e-06,
+      "loss": 0.3638,
+      "step": 23314
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1563890825133e-06,
+      "loss": 0.3014,
+      "step": 23315
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1525167700319322e-06,
+      "loss": 0.0048,
+      "step": 23316
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1486468090926087e-06,
+      "loss": 0.8322,
+      "step": 23317
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.144779199757297e-06,
+      "loss": 0.3206,
+      "step": 23318
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1409139420879638e-06,
+      "loss": 0.8462,
+      "step": 23319
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1370510361464764e-06,
+      "loss": 0.2967,
+      "step": 23320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1331904819947187e-06,
+      "loss": 0.476,
+      "step": 23321
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.129332279694491e-06,
+      "loss": 0.0018,
+      "step": 23322
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.125476429307594e-06,
+      "loss": 0.0049,
+      "step": 23323
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1216229308957618e-06,
+      "loss": 1.7734,
+      "step": 23324
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.117771784520695e-06,
+      "loss": 0.462,
+      "step": 23325
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1139229902440778e-06,
+      "loss": 0.7078,
+      "step": 23326
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1100765481275434e-06,
+      "loss": 0.2405,
+      "step": 23327
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1062324582326768e-06,
+      "loss": 0.3599,
+      "step": 23328
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.1023907206210453e-06,
+      "loss": 0.0359,
+      "step": 23329
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0985513353541657e-06,
+      "loss": 0.9491,
+      "step": 23330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.094714302493506e-06,
+      "loss": 0.3836,
+      "step": 23331
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0908796221005503e-06,
+      "loss": 2.2285,
+      "step": 23332
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0870472942366497e-06,
+      "loss": 0.6606,
+      "step": 23333
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0832173189632047e-06,
+      "loss": 2.8985,
+      "step": 23334
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0793896963415665e-06,
+      "loss": 0.3048,
+      "step": 23335
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.075564426432969e-06,
+      "loss": 0.8034,
+      "step": 23336
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0717415092987306e-06,
+      "loss": 0.6813,
+      "step": 23337
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.067920945000052e-06,
+      "loss": 0.463,
+      "step": 23338
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.064102733598067e-06,
+      "loss": 0.3872,
+      "step": 23339
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0602868751539945e-06,
+      "loss": 0.2557,
+      "step": 23340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0564733697288847e-06,
+      "loss": 0.0025,
+      "step": 23341
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.052662217383822e-06,
+      "loss": 0.3376,
+      "step": 23342
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.048853418179842e-06,
+      "loss": 1.8799,
+      "step": 23343
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.045046972177928e-06,
+      "loss": 0.7361,
+      "step": 23344
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.041242879439032e-06,
+      "loss": 0.2557,
+      "step": 23345
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.037441140024105e-06,
+      "loss": 0.1611,
+      "step": 23346
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.033641753993965e-06,
+      "loss": 0.4929,
+      "step": 23347
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.029844721409497e-06,
+      "loss": 0.1567,
+      "step": 23348
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0260500423315016e-06,
+      "loss": 0.2865,
+      "step": 23349
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0222577168207307e-06,
+      "loss": 0.1115,
+      "step": 23350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.018467744937919e-06,
+      "loss": 0.5224,
+      "step": 23351
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.014680126743768e-06,
+      "loss": 0.2482,
+      "step": 23352
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0108948622989125e-06,
+      "loss": 0.8585,
+      "step": 23353
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.0071119516639872e-06,
+      "loss": 0.3539,
+      "step": 23354
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.003331394899544e-06,
+      "loss": 0.3209,
+      "step": 23355
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9995531920661343e-06,
+      "loss": 0.6592,
+      "step": 23356
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.995777343224276e-06,
+      "loss": 0.0022,
+      "step": 23357
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.992003848434421e-06,
+      "loss": 0.0015,
+      "step": 23358
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9882327077570046e-06,
+      "loss": 0.0017,
+      "step": 23359
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.984463921252378e-06,
+      "loss": 1.3971,
+      "step": 23360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9806974889809432e-06,
+      "loss": 0.3205,
+      "step": 23361
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.976933411003002e-06,
+      "loss": 0.7031,
+      "step": 23362
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9731716873787894e-06,
+      "loss": 0.0395,
+      "step": 23363
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9694123181685904e-06,
+      "loss": 0.3134,
+      "step": 23364
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.965655303432607e-06,
+      "loss": 0.3628,
+      "step": 23365
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9619006432309743e-06,
+      "loss": 0.0039,
+      "step": 23366
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9581483376238444e-06,
+      "loss": 0.837,
+      "step": 23367
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9543983866712697e-06,
+      "loss": 0.311,
+      "step": 23368
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9506507904333175e-06,
+      "loss": 0.0013,
+      "step": 23369
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9469055489700246e-06,
+      "loss": 0.5481,
+      "step": 23370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.943162662341325e-06,
+      "loss": 0.3147,
+      "step": 23371
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.939422130607172e-06,
+      "loss": 2.287,
+      "step": 23372
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9356839538274832e-06,
+      "loss": 0.3988,
+      "step": 23373
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.931948132062095e-06,
+      "loss": 0.5832,
+      "step": 23374
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.928214665370826e-06,
+      "loss": 0.5006,
+      "step": 23375
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9244835538134947e-06,
+      "loss": 0.6899,
+      "step": 23376
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.92075479744982e-06,
+      "loss": 0.1235,
+      "step": 23377
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.917028396339521e-06,
+      "loss": 0.4015,
+      "step": 23378
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9133043505422836e-06,
+      "loss": 0.0641,
+      "step": 23379
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9095826601177264e-06,
+      "loss": 0.2301,
+      "step": 23380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9058633251254513e-06,
+      "loss": 0.0019,
+      "step": 23381
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.9021463456250116e-06,
+      "loss": 0.0069,
+      "step": 23382
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8984317216759258e-06,
+      "loss": 1.1702,
+      "step": 23383
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8947194533377126e-06,
+      "loss": 0.6001,
+      "step": 23384
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8910095406697754e-06,
+      "loss": 0.5649,
+      "step": 23385
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8873019837315323e-06,
+      "loss": 0.0019,
+      "step": 23386
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.883596782582387e-06,
+      "loss": 0.0021,
+      "step": 23387
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.879893937281624e-06,
+      "loss": 0.2272,
+      "step": 23388
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8761934478885974e-06,
+      "loss": 0.9866,
+      "step": 23389
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8724953144625085e-06,
+      "loss": 0.0016,
+      "step": 23390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.868799537062594e-06,
+      "loss": 0.3744,
+      "step": 23391
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.865106115748056e-06,
+      "loss": 0.9608,
+      "step": 23392
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8614150505780144e-06,
+      "loss": 0.449,
+      "step": 23393
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8577263416115713e-06,
+      "loss": 1.316,
+      "step": 23394
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.854039988907847e-06,
+      "loss": 0.314,
+      "step": 23395
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.850355992525827e-06,
+      "loss": 1.1128,
+      "step": 23396
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.846674352524497e-06,
+      "loss": 0.312,
+      "step": 23397
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8429950689628444e-06,
+      "loss": 0.2164,
+      "step": 23398
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.839318141899771e-06,
+      "loss": 0.3835,
+      "step": 23399
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.835643571394164e-06,
+      "loss": 0.6601,
+      "step": 23400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8319713575048587e-06,
+      "loss": 1.6321,
+      "step": 23401
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8283015002906585e-06,
+      "loss": 0.628,
+      "step": 23402
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8246339998103494e-06,
+      "loss": 0.0031,
+      "step": 23403
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.820968856122635e-06,
+      "loss": 0.6059,
+      "step": 23404
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8173060692862184e-06,
+      "loss": 1.1137,
+      "step": 23405
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8136456393597685e-06,
+      "loss": 0.4443,
+      "step": 23406
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8099875664018724e-06,
+      "loss": 0.3246,
+      "step": 23407
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.8063318504711496e-06,
+      "loss": 0.7597,
+      "step": 23408
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.802678491626087e-06,
+      "loss": 0.3518,
+      "step": 23409
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7990274899252207e-06,
+      "loss": 0.2049,
+      "step": 23410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7953788454270378e-06,
+      "loss": 0.7192,
+      "step": 23411
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7917325581899074e-06,
+      "loss": 0.7634,
+      "step": 23412
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.788088628272267e-06,
+      "loss": 0.0026,
+      "step": 23413
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7844470557324528e-06,
+      "loss": 0.3544,
+      "step": 23414
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.780807840628768e-06,
+      "loss": 0.6188,
+      "step": 23415
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7771709830195166e-06,
+      "loss": 0.5267,
+      "step": 23416
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7735364829629183e-06,
+      "loss": 0.0009,
+      "step": 23417
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7699043405171763e-06,
+      "loss": 0.4218,
+      "step": 23418
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7662745557404442e-06,
+      "loss": 0.7212,
+      "step": 23419
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.762647128690859e-06,
+      "loss": 0.374,
+      "step": 23420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7590220594265245e-06,
+      "loss": 1.1172,
+      "step": 23421
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7553993480054605e-06,
+      "loss": 0.1885,
+      "step": 23422
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7517789944856873e-06,
+      "loss": 1.9643,
+      "step": 23423
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7481609989251916e-06,
+      "loss": 0.0035,
+      "step": 23424
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7445453613819113e-06,
+      "loss": 0.0017,
+      "step": 23425
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7409320819137327e-06,
+      "loss": 0.337,
+      "step": 23426
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.737321160578526e-06,
+      "loss": 0.0034,
+      "step": 23427
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7337125974340955e-06,
+      "loss": 0.3169,
+      "step": 23428
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7301063925382448e-06,
+      "loss": 0.2011,
+      "step": 23429
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7265025459487276e-06,
+      "loss": 0.1641,
+      "step": 23430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.722901057723248e-06,
+      "loss": 0.6924,
+      "step": 23431
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.719301927919476e-06,
+      "loss": 0.336,
+      "step": 23432
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7157051565950493e-06,
+      "loss": 0.2882,
+      "step": 23433
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7121107438075383e-06,
+      "loss": 0.4624,
+      "step": 23434
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7085186896145473e-06,
+      "loss": 0.2541,
+      "step": 23435
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7049289940735797e-06,
+      "loss": 0.001,
+      "step": 23436
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.7013416572420897e-06,
+      "loss": 0.7172,
+      "step": 23437
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6977566791775816e-06,
+      "loss": 0.3571,
+      "step": 23438
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6941740599374097e-06,
+      "loss": 1.9062,
+      "step": 23439
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6905937995789608e-06,
+      "loss": 0.4023,
+      "step": 23440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.687015898159589e-06,
+      "loss": 0.23,
+      "step": 23441
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.683440355736566e-06,
+      "loss": 0.5284,
+      "step": 23442
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.679867172367145e-06,
+      "loss": 0.9129,
+      "step": 23443
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.676296348108564e-06,
+      "loss": 0.2759,
+      "step": 23444
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6727278830179942e-06,
+      "loss": 0.2258,
+      "step": 23445
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.669161777152573e-06,
+      "loss": 0.844,
+      "step": 23446
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6655980305694215e-06,
+      "loss": 0.0039,
+      "step": 23447
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6620366433255937e-06,
+      "loss": 0.3627,
+      "step": 23448
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6584776154781275e-06,
+      "loss": 0.0025,
+      "step": 23449
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.654920947084027e-06,
+      "loss": 0.0047,
+      "step": 23450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.651366638200214e-06,
+      "loss": 0.2025,
+      "step": 23451
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.647814688883626e-06,
+      "loss": 0.2994,
+      "step": 23452
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6442650991911507e-06,
+      "loss": 0.6725,
+      "step": 23453
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6407178691796094e-06,
+      "loss": 0.3016,
+      "step": 23454
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6371729989058234e-06,
+      "loss": 0.0272,
+      "step": 23455
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6336304884265634e-06,
+      "loss": 0.2165,
+      "step": 23456
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.630090337798535e-06,
+      "loss": 0.0102,
+      "step": 23457
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6265525470784254e-06,
+      "loss": 0.6338,
+      "step": 23458
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6230171163229063e-06,
+      "loss": 0.0045,
+      "step": 23459
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6194840455886156e-06,
+      "loss": 0.1004,
+      "step": 23460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6159533349320585e-06,
+      "loss": 0.0014,
+      "step": 23461
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6124249844098388e-06,
+      "loss": 0.1868,
+      "step": 23462
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.608898994078429e-06,
+      "loss": 0.0987,
+      "step": 23463
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.6053753639943e-06,
+      "loss": 1.418,
+      "step": 23464
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.60185409421389e-06,
+      "loss": 0.4264,
+      "step": 23465
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.598335184793554e-06,
+      "loss": 0.557,
+      "step": 23466
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.594818635789664e-06,
+      "loss": 1.0535,
+      "step": 23467
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5913044472585244e-06,
+      "loss": 0.4154,
+      "step": 23468
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5877926192564236e-06,
+      "loss": 0.6965,
+      "step": 23469
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5842831518395837e-06,
+      "loss": 0.0017,
+      "step": 23470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5807760450642256e-06,
+      "loss": 0.1977,
+      "step": 23471
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.577271298986472e-06,
+      "loss": 0.0053,
+      "step": 23472
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.573768913662477e-06,
+      "loss": 0.3695,
+      "step": 23473
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5702688891483137e-06,
+      "loss": 0.7638,
+      "step": 23474
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.566771225500036e-06,
+      "loss": 1.3754,
+      "step": 23475
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.56327592277365e-06,
+      "loss": 0.5311,
+      "step": 23476
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5597829810251434e-06,
+      "loss": 0.359,
+      "step": 23477
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5562924003104223e-06,
+      "loss": 0.9995,
+      "step": 23478
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.552804180685392e-06,
+      "loss": 0.0035,
+      "step": 23479
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.549318322205923e-06,
+      "loss": 0.6815,
+      "step": 23480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5458348249278225e-06,
+      "loss": 0.0032,
+      "step": 23481
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5423536889068785e-06,
+      "loss": 0.6323,
+      "step": 23482
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5388749141988463e-06,
+      "loss": 0.0021,
+      "step": 23483
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.535398500859415e-06,
+      "loss": 0.7284,
+      "step": 23484
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5319244489442723e-06,
+      "loss": 0.4885,
+      "step": 23485
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5284527585090253e-06,
+      "loss": 0.2038,
+      "step": 23486
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5249834296093118e-06,
+      "loss": 0.0125,
+      "step": 23487
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.521516462300638e-06,
+      "loss": 0.392,
+      "step": 23488
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.518051856638542e-06,
+      "loss": 0.2394,
+      "step": 23489
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5145896126785137e-06,
+      "loss": 0.6432,
+      "step": 23490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5111297304759915e-06,
+      "loss": 0.7481,
+      "step": 23491
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5076722100863643e-06,
+      "loss": 0.3585,
+      "step": 23492
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5042170515650216e-06,
+      "loss": 0.4277,
+      "step": 23493
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.5007642549672858e-06,
+      "loss": 0.0783,
+      "step": 23494
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4973138203484456e-06,
+      "loss": 0.0017,
+      "step": 23495
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4938657477637403e-06,
+      "loss": 0.5951,
+      "step": 23496
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.490420037268409e-06,
+      "loss": 0.334,
+      "step": 23497
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4869766889176245e-06,
+      "loss": 0.1335,
+      "step": 23498
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.483535702766509e-06,
+      "loss": 0.3234,
+      "step": 23499
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4800970788701857e-06,
+      "loss": 2.4446,
+      "step": 23500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.47666081728371e-06,
+      "loss": 0.0009,
+      "step": 23501
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4732269180621044e-06,
+      "loss": 0.3845,
+      "step": 23502
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4697953812603755e-06,
+      "loss": 0.327,
+      "step": 23503
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4663662069334454e-06,
+      "loss": 0.261,
+      "step": 23504
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.462939395136254e-06,
+      "loss": 0.2696,
+      "step": 23505
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.459514945923674e-06,
+      "loss": 0.27,
+      "step": 23506
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.456092859350528e-06,
+      "loss": 0.6853,
+      "step": 23507
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4526731354716054e-06,
+      "loss": 0.7313,
+      "step": 23508
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.449255774341713e-06,
+      "loss": 0.2745,
+      "step": 23509
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4458407760155228e-06,
+      "loss": 0.0008,
+      "step": 23510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4424281405477585e-06,
+      "loss": 0.8832,
+      "step": 23511
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4390178679930593e-06,
+      "loss": 0.0444,
+      "step": 23512
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.435609958406015e-06,
+      "loss": 0.3538,
+      "step": 23513
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4322044118412322e-06,
+      "loss": 0.6536,
+      "step": 23514
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4288012283532167e-06,
+      "loss": 0.0059,
+      "step": 23515
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.425400407996475e-06,
+      "loss": 1.2284,
+      "step": 23516
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4220019508254806e-06,
+      "loss": 0.6538,
+      "step": 23517
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.418605856894623e-06,
+      "loss": 0.2123,
+      "step": 23518
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.415212126258309e-06,
+      "loss": 0.0424,
+      "step": 23519
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.411820758970895e-06,
+      "loss": 0.3712,
+      "step": 23520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4084317550866538e-06,
+      "loss": 0.1678,
+      "step": 23521
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.4050451146598757e-06,
+      "loss": 0.4672,
+      "step": 23522
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.401660837744801e-06,
+      "loss": 0.4729,
+      "step": 23523
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3982789243956023e-06,
+      "loss": 1.6674,
+      "step": 23524
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3948993746664533e-06,
+      "loss": 0.4857,
+      "step": 23525
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.391522188611461e-06,
+      "loss": 1.7801,
+      "step": 23526
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.388147366284715e-06,
+      "loss": 0.8407,
+      "step": 23527
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3847749077402557e-06,
+      "loss": 1.7861,
+      "step": 23528
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3814048130320896e-06,
+      "loss": 0.2504,
+      "step": 23529
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3780370822141905e-06,
+      "loss": 0.2846,
+      "step": 23530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3746717153404648e-06,
+      "loss": 0.4648,
+      "step": 23531
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.37130871246482e-06,
+      "loss": 0.5525,
+      "step": 23532
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.367948073641096e-06,
+      "loss": 0.2474,
+      "step": 23533
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.36458979892315e-06,
+      "loss": 0.8166,
+      "step": 23534
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.361233888364689e-06,
+      "loss": 0.0086,
+      "step": 23535
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.357880342019519e-06,
+      "loss": 0.0361,
+      "step": 23536
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3545291599412985e-06,
+      "loss": 0.482,
+      "step": 23537
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3511803421837173e-06,
+      "loss": 0.7582,
+      "step": 23538
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.347833888800399e-06,
+      "loss": 0.0689,
+      "step": 23539
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3444897998449176e-06,
+      "loss": 0.0003,
+      "step": 23540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.34114807537083e-06,
+      "loss": 0.1574,
+      "step": 23541
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3378087154316605e-06,
+      "loss": 0.3461,
+      "step": 23542
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.334471720080866e-06,
+      "loss": 1.1128,
+      "step": 23543
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.331137089371904e-06,
+      "loss": 1.7236,
+      "step": 23544
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.327804823358148e-06,
+      "loss": 0.1191,
+      "step": 23545
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3244749220929727e-06,
+      "loss": 0.1533,
+      "step": 23546
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.321147385629718e-06,
+      "loss": 0.7651,
+      "step": 23547
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3178222140216417e-06,
+      "loss": 0.6514,
+      "step": 23548
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3144994073220014e-06,
+      "loss": 0.7016,
+      "step": 23549
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.311178965584004e-06,
+      "loss": 0.0756,
+      "step": 23550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3078608888608243e-06,
+      "loss": 0.3099,
+      "step": 23551
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.3045451772056024e-06,
+      "loss": 0.3182,
+      "step": 23552
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.301231830671446e-06,
+      "loss": 0.2364,
+      "step": 23553
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2979208493113797e-06,
+      "loss": 0.5922,
+      "step": 23554
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2946122331784443e-06,
+      "loss": 0.2701,
+      "step": 23555
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.291305982325614e-06,
+      "loss": 0.5066,
+      "step": 23556
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2880020968058467e-06,
+      "loss": 0.6216,
+      "step": 23557
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2847005766720494e-06,
+      "loss": 0.1782,
+      "step": 23558
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2814014219770638e-06,
+      "loss": 0.0031,
+      "step": 23559
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2781046327737474e-06,
+      "loss": 0.6437,
+      "step": 23560
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.274810209114891e-06,
+      "loss": 0.1784,
+      "step": 23561
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.27151815105322e-06,
+      "loss": 0.0145,
+      "step": 23562
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.268228458641508e-06,
+      "loss": 0.0103,
+      "step": 23563
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2649411319323964e-06,
+      "loss": 0.6912,
+      "step": 23564
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2616561709785097e-06,
+      "loss": 0.0393,
+      "step": 23565
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.258373575832506e-06,
+      "loss": 0.7139,
+      "step": 23566
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.25509334654691e-06,
+      "loss": 0.0017,
+      "step": 23567
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2518154831742462e-06,
+      "loss": 0.7676,
+      "step": 23568
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2485399857670394e-06,
+      "loss": 0.0032,
+      "step": 23569
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.245266854377714e-06,
+      "loss": 0.0006,
+      "step": 23570
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.241996089058695e-06,
+      "loss": 0.3186,
+      "step": 23571
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2387276898623408e-06,
+      "loss": 0.1866,
+      "step": 23572
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2354616568410087e-06,
+      "loss": 0.6824,
+      "step": 23573
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.232197990047008e-06,
+      "loss": 1.1149,
+      "step": 23574
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.228936689532562e-06,
+      "loss": 0.1937,
+      "step": 23575
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.225677755349947e-06,
+      "loss": 0.2899,
+      "step": 23576
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.222421187551321e-06,
+      "loss": 0.2378,
+      "step": 23577
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.219166986188825e-06,
+      "loss": 0.7309,
+      "step": 23578
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2159151513145845e-06,
+      "loss": 0.3522,
+      "step": 23579
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2126656829806742e-06,
+      "loss": 0.6444,
+      "step": 23580
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2094185812391194e-06,
+      "loss": 0.3029,
+      "step": 23581
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.2061738461419288e-06,
+      "loss": 0.3188,
+      "step": 23582
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.20293147774106e-06,
+      "loss": 0.2199,
+      "step": 23583
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.199691476088422e-06,
+      "loss": 1.1497,
+      "step": 23584
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1964538412359235e-06,
+      "loss": 0.2069,
+      "step": 23585
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.193218573235389e-06,
+      "loss": 0.0074,
+      "step": 23586
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.189985672138611e-06,
+      "loss": 0.0163,
+      "step": 23587
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.186755137997398e-06,
+      "loss": 0.1574,
+      "step": 23588
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.183526970863475e-06,
+      "loss": 0.3749,
+      "step": 23589
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1803011707885008e-06,
+      "loss": 0.5032,
+      "step": 23590
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1770777378241846e-06,
+      "loss": 2.7793,
+      "step": 23591
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.173856672022101e-06,
+      "loss": 0.7105,
+      "step": 23592
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.170637973433842e-06,
+      "loss": 0.227,
+      "step": 23593
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.167421642110967e-06,
+      "loss": 0.5264,
+      "step": 23594
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1642076781049678e-06,
+      "loss": 0.3473,
+      "step": 23595
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1609960814673032e-06,
+      "loss": 0.2781,
+      "step": 23596
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1577868522494157e-06,
+      "loss": 0.1793,
+      "step": 23597
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1545799905026972e-06,
+      "loss": 0.5924,
+      "step": 23598
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1513754962784904e-06,
+      "loss": 0.0564,
+      "step": 23599
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.148173369628103e-06,
+      "loss": 0.422,
+      "step": 23600
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.144973610602846e-06,
+      "loss": 0.4069,
+      "step": 23601
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1417762192539267e-06,
+      "loss": 3.9336,
+      "step": 23602
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.138581195632555e-06,
+      "loss": 0.0018,
+      "step": 23603
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.135388539789906e-06,
+      "loss": 0.7277,
+      "step": 23604
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.13219825177709e-06,
+      "loss": 0.2324,
+      "step": 23605
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.129010331645198e-06,
+      "loss": 0.001,
+      "step": 23606
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1258247794452897e-06,
+      "loss": 0.0051,
+      "step": 23607
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.122641595228358e-06,
+      "loss": 0.6016,
+      "step": 23608
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1194607790453954e-06,
+      "loss": 0.2999,
+      "step": 23609
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.116282330947344e-06,
+      "loss": 0.9378,
+      "step": 23610
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.113106250985064e-06,
+      "loss": 0.075,
+      "step": 23611
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1099325392094635e-06,
+      "loss": 0.8776,
+      "step": 23612
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.106761195671319e-06,
+      "loss": 0.5928,
+      "step": 23613
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1035922204214406e-06,
+      "loss": 0.6888,
+      "step": 23614
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.1004256135105867e-06,
+      "loss": 0.1322,
+      "step": 23615
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0972613749894506e-06,
+      "loss": 0.097,
+      "step": 23616
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0940995049086752e-06,
+      "loss": 0.7049,
+      "step": 23617
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.090940003318953e-06,
+      "loss": 0.0026,
+      "step": 23618
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0877828702708266e-06,
+      "loss": 0.452,
+      "step": 23619
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.084628105814873e-06,
+      "loss": 0.144,
+      "step": 23620
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.081475710001618e-06,
+      "loss": 0.2187,
+      "step": 23621
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0783256828815375e-06,
+      "loss": 0.4579,
+      "step": 23622
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0751780245050586e-06,
+      "loss": 1.1104,
+      "step": 23623
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0720327349226075e-06,
+      "loss": 0.4267,
+      "step": 23624
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0688898141845434e-06,
+      "loss": 0.0034,
+      "step": 23625
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0657492623412096e-06,
+      "loss": 0.3772,
+      "step": 23626
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.062611079442866e-06,
+      "loss": 0.7423,
+      "step": 23627
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0594752655397894e-06,
+      "loss": 0.3324,
+      "step": 23628
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0563418206821726e-06,
+      "loss": 0.9994,
+      "step": 23629
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.053210744920225e-06,
+      "loss": 1.4865,
+      "step": 23630
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0500820383040737e-06,
+      "loss": 0.2031,
+      "step": 23631
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0469557008838123e-06,
+      "loss": 0.258,
+      "step": 23632
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0438317327095e-06,
+      "loss": 0.9236,
+      "step": 23633
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0407101338311973e-06,
+      "loss": 0.8476,
+      "step": 23634
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.037590904298847e-06,
+      "loss": 1.0336,
+      "step": 23635
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0344740441624097e-06,
+      "loss": 0.2608,
+      "step": 23636
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.031359553471812e-06,
+      "loss": 0.5231,
+      "step": 23637
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0282474322769302e-06,
+      "loss": 0.6494,
+      "step": 23638
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0251376806275754e-06,
+      "loss": 0.2832,
+      "step": 23639
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0220302985735737e-06,
+      "loss": 1.2504,
+      "step": 23640
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0189252861646686e-06,
+      "loss": 0.1665,
+      "step": 23641
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0158226434505876e-06,
+      "loss": 0.0004,
+      "step": 23642
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0127223704810235e-06,
+      "loss": 0.106,
+      "step": 23643
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0096244673056037e-06,
+      "loss": 0.2164,
+      "step": 23644
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.006528933973939e-06,
+      "loss": 2.0098,
+      "step": 23645
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.0034357705356053e-06,
+      "loss": 0.2467,
+      "step": 23646
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2.000344977040147e-06,
+      "loss": 0.0004,
+      "step": 23647
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.997256553537041e-06,
+      "loss": 0.4772,
+      "step": 23648
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.994170500075748e-06,
+      "loss": 0.6054,
+      "step": 23649
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.991086816705695e-06,
+      "loss": 0.0004,
+      "step": 23650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9880055034762587e-06,
+      "loss": 0.0011,
+      "step": 23651
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9849265604367836e-06,
+      "loss": 0.0007,
+      "step": 23652
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9818499876365635e-06,
+      "loss": 0.4725,
+      "step": 23653
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9787757851248587e-06,
+      "loss": 0.2973,
+      "step": 23654
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.97570395295093e-06,
+      "loss": 0.0016,
+      "step": 23655
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9726344911639545e-06,
+      "loss": 0.6521,
+      "step": 23656
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9695673998130434e-06,
+      "loss": 0.3769,
+      "step": 23657
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9665026789473736e-06,
+      "loss": 0.6512,
+      "step": 23658
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9634403286160063e-06,
+      "loss": 0.1887,
+      "step": 23659
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9603803488679515e-06,
+      "loss": 0.7977,
+      "step": 23660
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.957322739752237e-06,
+      "loss": 0.2171,
+      "step": 23661
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9542675013178244e-06,
+      "loss": 0.1439,
+      "step": 23662
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.951214633613607e-06,
+      "loss": 0.1848,
+      "step": 23663
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.948164136688529e-06,
+      "loss": 0.428,
+      "step": 23664
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9451160105913853e-06,
+      "loss": 0.5721,
+      "step": 23665
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9420702553710197e-06,
+      "loss": 2.0039,
+      "step": 23666
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9390268710761937e-06,
+      "loss": 0.3232,
+      "step": 23667
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.935985857755651e-06,
+      "loss": 0.0005,
+      "step": 23668
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9329472154580526e-06,
+      "loss": 0.6411,
+      "step": 23669
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.929910944232127e-06,
+      "loss": 1.3245,
+      "step": 23670
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9268770441264348e-06,
+      "loss": 0.2426,
+      "step": 23671
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.923845515189587e-06,
+      "loss": 0.396,
+      "step": 23672
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.920816357470112e-06,
+      "loss": 0.2305,
+      "step": 23673
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.917789571016537e-06,
+      "loss": 0.4649,
+      "step": 23674
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.914765155877307e-06,
+      "loss": 0.0107,
+      "step": 23675
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.911743112100883e-06,
+      "loss": 0.2983,
+      "step": 23676
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.908723439735643e-06,
+      "loss": 0.8981,
+      "step": 23677
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9057061388299323e-06,
+      "loss": 0.5346,
+      "step": 23678
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.9026912094320946e-06,
+      "loss": 0.1974,
+      "step": 23679
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.899678651590375e-06,
+      "loss": 0.0075,
+      "step": 23680
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8966684653530516e-06,
+      "loss": 0.1065,
+      "step": 23681
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8936606507683027e-06,
+      "loss": 0.2483,
+      "step": 23682
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8906552078843062e-06,
+      "loss": 1.551,
+      "step": 23683
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8876521367491737e-06,
+      "loss": 0.8172,
+      "step": 23684
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8846514374110167e-06,
+      "loss": 0.5719,
+      "step": 23685
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8816531099178965e-06,
+      "loss": 2.3789,
+      "step": 23686
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8786571543177753e-06,
+      "loss": 0.3182,
+      "step": 23687
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8756635706586643e-06,
+      "loss": 0.3071,
+      "step": 23688
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8726723589885251e-06,
+      "loss": 0.6928,
+      "step": 23689
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.869683519355203e-06,
+      "loss": 0.0042,
+      "step": 23690
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8666970518065927e-06,
+      "loss": 0.4161,
+      "step": 23691
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8637129563905229e-06,
+      "loss": 0.5208,
+      "step": 23692
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8607312331547552e-06,
+      "loss": 0.3179,
+      "step": 23693
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8577518821470517e-06,
+      "loss": 0.2373,
+      "step": 23694
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8547749034151238e-06,
+      "loss": 0.6572,
+      "step": 23695
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8518002970066504e-06,
+      "loss": 0.4057,
+      "step": 23696
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8488280629692432e-06,
+      "loss": 0.2123,
+      "step": 23697
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8458582013505142e-06,
+      "loss": 0.517,
+      "step": 23698
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8428907121980252e-06,
+      "loss": 0.0011,
+      "step": 23699
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8399255955592719e-06,
+      "loss": 1.2473,
+      "step": 23700
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8369628514817658e-06,
+      "loss": 0.7109,
+      "step": 23701
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8340024800129527e-06,
+      "loss": 0.1494,
+      "step": 23702
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8310444812001946e-06,
+      "loss": 0.5419,
+      "step": 23703
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8280888550909034e-06,
+      "loss": 1.2129,
+      "step": 23704
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8251356017324082e-06,
+      "loss": 1.3144,
+      "step": 23705
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8221847211719543e-06,
+      "loss": 3.2754,
+      "step": 23706
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8192362134568373e-06,
+      "loss": 0.5146,
+      "step": 23707
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8162900786342859e-06,
+      "loss": 0.5897,
+      "step": 23708
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8133463167514296e-06,
+      "loss": 1.6313,
+      "step": 23709
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8104049278554633e-06,
+      "loss": 0.2446,
+      "step": 23710
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8074659119934498e-06,
+      "loss": 0.4573,
+      "step": 23711
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8045292692124513e-06,
+      "loss": 0.2821,
+      "step": 23712
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8015949995595135e-06,
+      "loss": 0.4405,
+      "step": 23713
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7986631030816156e-06,
+      "loss": 1.2949,
+      "step": 23714
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7957335798257033e-06,
+      "loss": 0.5738,
+      "step": 23715
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7928064298387224e-06,
+      "loss": 1.1349,
+      "step": 23716
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7898816531674854e-06,
+      "loss": 0.0076,
+      "step": 23717
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7869592498588714e-06,
+      "loss": 0.2784,
+      "step": 23718
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7840392199596765e-06,
+      "loss": 0.2701,
+      "step": 23719
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7811215635166464e-06,
+      "loss": 0.0316,
+      "step": 23720
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7782062805765106e-06,
+      "loss": 0.2072,
+      "step": 23721
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7752933711859652e-06,
+      "loss": 0.2529,
+      "step": 23722
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7723828353916226e-06,
+      "loss": 0.7837,
+      "step": 23723
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7694746732401121e-06,
+      "loss": 0.4247,
+      "step": 23724
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7665688847780135e-06,
+      "loss": 2.4303,
+      "step": 23725
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7636654700518393e-06,
+      "loss": 0.354,
+      "step": 23726
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7607644291080858e-06,
+      "loss": 0.2473,
+      "step": 23727
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7578657619932157e-06,
+      "loss": 0.211,
+      "step": 23728
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7549694687536587e-06,
+      "loss": 0.004,
+      "step": 23729
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7520755494357608e-06,
+      "loss": 0.6195,
+      "step": 23730
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7491840040859018e-06,
+      "loss": 0.2007,
+      "step": 23731
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7462948327503613e-06,
+      "loss": 0.4577,
+      "step": 23732
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.743408035475402e-06,
+      "loss": 0.1823,
+      "step": 23733
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7405236123072707e-06,
+      "loss": 0.3981,
+      "step": 23734
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.737641563292147e-06,
+      "loss": 0.9232,
+      "step": 23735
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.734761888476177e-06,
+      "loss": 0.2468,
+      "step": 23736
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.731884587905491e-06,
+      "loss": 0.2639,
+      "step": 23737
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7290096616261518e-06,
+      "loss": 0.4279,
+      "step": 23738
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7261371096842058e-06,
+      "loss": 0.4406,
+      "step": 23739
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7232669321256332e-06,
+      "loss": 0.0006,
+      "step": 23740
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7203991289964137e-06,
+      "loss": 0.3267,
+      "step": 23741
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7175337003424604e-06,
+      "loss": 0.4327,
+      "step": 23742
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7146706462096704e-06,
+      "loss": 0.9437,
+      "step": 23743
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7118099666438734e-06,
+      "loss": 0.2303,
+      "step": 23744
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7089516616908993e-06,
+      "loss": 0.1506,
+      "step": 23745
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7060957313965117e-06,
+      "loss": 0.2296,
+      "step": 23746
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.7032421758064406e-06,
+      "loss": 0.6584,
+      "step": 23747
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.700390994966383e-06,
+      "loss": 0.0015,
+      "step": 23748
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.697542188922002e-06,
+      "loss": 0.0016,
+      "step": 23749
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6946957577189114e-06,
+      "loss": 0.3254,
+      "step": 23750
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6918517014026745e-06,
+      "loss": 0.1524,
+      "step": 23751
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6890100200188717e-06,
+      "loss": 0.6275,
+      "step": 23752
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6861707136129831e-06,
+      "loss": 0.7749,
+      "step": 23753
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.683333782230506e-06,
+      "loss": 0.4586,
+      "step": 23754
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6804992259168203e-06,
+      "loss": 0.287,
+      "step": 23755
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.67766704471734e-06,
+      "loss": 0.6533,
+      "step": 23756
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6748372386774455e-06,
+      "loss": 0.1667,
+      "step": 23757
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6720098078424005e-06,
+      "loss": 0.003,
+      "step": 23758
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6691847522575353e-06,
+      "loss": 3.1797,
+      "step": 23759
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.666362071968047e-06,
+      "loss": 1.0033,
+      "step": 23760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6635417670191498e-06,
+      "loss": 0.3473,
+      "step": 23761
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.660723837456024e-06,
+      "loss": 0.6988,
+      "step": 23762
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6579082833237666e-06,
+      "loss": 0.1675,
+      "step": 23763
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6550951046674754e-06,
+      "loss": 0.5259,
+      "step": 23764
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.652284301532214e-06,
+      "loss": 0.3087,
+      "step": 23765
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6494758739629798e-06,
+      "loss": 0.6953,
+      "step": 23766
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6466698220047204e-06,
+      "loss": 0.1958,
+      "step": 23767
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6438661457024327e-06,
+      "loss": 0.0006,
+      "step": 23768
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6410648451009645e-06,
+      "loss": 0.1243,
+      "step": 23769
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6382659202451797e-06,
+      "loss": 0.7233,
+      "step": 23770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6354693711799094e-06,
+      "loss": 0.0017,
+      "step": 23771
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.632675197949934e-06,
+      "loss": 0.2199,
+      "step": 23772
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6298834006000017e-06,
+      "loss": 0.0013,
+      "step": 23773
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6270939791748095e-06,
+      "loss": 1.1515,
+      "step": 23774
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6243069337190383e-06,
+      "loss": 0.2274,
+      "step": 23775
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6215222642773028e-06,
+      "loss": 2.8691,
+      "step": 23776
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6187399708942006e-06,
+      "loss": 0.7109,
+      "step": 23777
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6159600536142957e-06,
+      "loss": 0.4673,
+      "step": 23778
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6131825124820862e-06,
+      "loss": 0.2296,
+      "step": 23779
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.61040734754207e-06,
+      "loss": 0.7939,
+      "step": 23780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.607634558838694e-06,
+      "loss": 1.9385,
+      "step": 23781
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6048641464163237e-06,
+      "loss": 0.3254,
+      "step": 23782
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.6020961103193396e-06,
+      "loss": 0.0046,
+      "step": 23783
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.59933045059209e-06,
+      "loss": 0.8873,
+      "step": 23784
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5965671672788393e-06,
+      "loss": 0.5919,
+      "step": 23785
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5938062604238355e-06,
+      "loss": 0.0075,
+      "step": 23786
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5910477300713098e-06,
+      "loss": 0.6246,
+      "step": 23787
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.588291576265427e-06,
+      "loss": 0.5454,
+      "step": 23788
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5855377990503016e-06,
+      "loss": 0.0082,
+      "step": 23789
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.582786398470065e-06,
+      "loss": 0.5551,
+      "step": 23790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.580037374568749e-06,
+      "loss": 0.2343,
+      "step": 23791
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5772907273904011e-06,
+      "loss": 0.4333,
+      "step": 23792
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.57454645697897e-06,
+      "loss": 0.7407,
+      "step": 23793
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.571804563378437e-06,
+      "loss": 0.6345,
+      "step": 23794
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5690650466327003e-06,
+      "loss": 0.7838,
+      "step": 23795
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5663279067856082e-06,
+      "loss": 0.2757,
+      "step": 23796
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.563593143881009e-06,
+      "loss": 0.148,
+      "step": 23797
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5608607579627008e-06,
+      "loss": 1.1062,
+      "step": 23798
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5581307490744322e-06,
+      "loss": 0.3948,
+      "step": 23799
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.555403117259918e-06,
+      "loss": 0.2388,
+      "step": 23800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.55267786256284e-06,
+      "loss": 1.0768,
+      "step": 23801
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.54995498502683e-06,
+      "loss": 0.3488,
+      "step": 23802
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5472344846955031e-06,
+      "loss": 0.0434,
+      "step": 23803
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5445163616124245e-06,
+      "loss": 0.0019,
+      "step": 23804
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5418006158211093e-06,
+      "loss": 0.2634,
+      "step": 23805
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.539087247365073e-06,
+      "loss": 0.435,
+      "step": 23806
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5363762562877134e-06,
+      "loss": 0.0673,
+      "step": 23807
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5336676426324968e-06,
+      "loss": 0.0878,
+      "step": 23808
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.530961406442771e-06,
+      "loss": 0.9548,
+      "step": 23809
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5282575477618685e-06,
+      "loss": 0.4312,
+      "step": 23810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5255560666331045e-06,
+      "loss": 0.5206,
+      "step": 23811
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5228569630997279e-06,
+      "loss": 0.5678,
+      "step": 23812
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5201602372049537e-06,
+      "loss": 0.155,
+      "step": 23813
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5174658889920143e-06,
+      "loss": 0.0086,
+      "step": 23814
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.514773918503992e-06,
+      "loss": 0.0021,
+      "step": 23815
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5120843257840021e-06,
+      "loss": 3.0929,
+      "step": 23816
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5093971108751602e-06,
+      "loss": 0.2876,
+      "step": 23817
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5067122738204652e-06,
+      "loss": 0.9865,
+      "step": 23818
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.504029814662916e-06,
+      "loss": 0.2468,
+      "step": 23819
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.5013497334454616e-06,
+      "loss": 0.3248,
+      "step": 23820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4986720302110178e-06,
+      "loss": 0.0037,
+      "step": 23821
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4959967050024835e-06,
+      "loss": 0.6538,
+      "step": 23822
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.493323757862691e-06,
+      "loss": 1.2105,
+      "step": 23823
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4906531888344397e-06,
+      "loss": 0.0032,
+      "step": 23824
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4879849979604952e-06,
+      "loss": 0.002,
+      "step": 23825
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.48531918528359e-06,
+      "loss": 0.3385,
+      "step": 23826
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4826557508463898e-06,
+      "loss": 0.212,
+      "step": 23827
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4799946946915775e-06,
+      "loss": 3.6064,
+      "step": 23828
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4773360168617688e-06,
+      "loss": 0.2979,
+      "step": 23829
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4746797173995129e-06,
+      "loss": 0.0008,
+      "step": 23830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4720257963473426e-06,
+      "loss": 0.2555,
+      "step": 23831
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4693742537477738e-06,
+      "loss": 0.0055,
+      "step": 23832
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4667250896432725e-06,
+      "loss": 0.4621,
+      "step": 23833
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4640783040762382e-06,
+      "loss": 0.6815,
+      "step": 23834
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4614338970890705e-06,
+      "loss": 3.3701,
+      "step": 23835
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.458791868724135e-06,
+      "loss": 0.1734,
+      "step": 23836
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4561522190236984e-06,
+      "loss": 0.7664,
+      "step": 23837
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.45351494803006e-06,
+      "loss": 0.2861,
+      "step": 23838
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.450880055785436e-06,
+      "loss": 0.4282,
+      "step": 23839
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4482475423320262e-06,
+      "loss": 0.2633,
+      "step": 23840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4456174077119963e-06,
+      "loss": 0.3655,
+      "step": 23841
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4429896519674466e-06,
+      "loss": 0.102,
+      "step": 23842
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4403642751404765e-06,
+      "loss": 0.003,
+      "step": 23843
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4377412772731022e-06,
+      "loss": 0.7539,
+      "step": 23844
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4351206584073405e-06,
+      "loss": 0.3529,
+      "step": 23845
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4325024185851574e-06,
+      "loss": 1.1414,
+      "step": 23846
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4298865578484864e-06,
+      "loss": 0.1804,
+      "step": 23847
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4272730762392104e-06,
+      "loss": 0.0872,
+      "step": 23848
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4246619737991793e-06,
+      "loss": 2.8438,
+      "step": 23849
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.422053250570193e-06,
+      "loss": 0.5248,
+      "step": 23850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4194469065940518e-06,
+      "loss": 0.0074,
+      "step": 23851
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4168429419124717e-06,
+      "loss": 0.6473,
+      "step": 23852
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4142413565671695e-06,
+      "loss": 0.4852,
+      "step": 23853
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4116421505997954e-06,
+      "loss": 0.5708,
+      "step": 23854
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4090453240519827e-06,
+      "loss": 0.0562,
+      "step": 23855
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4064508769652816e-06,
+      "loss": 1.2383,
+      "step": 23856
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.4038588093812754e-06,
+      "loss": 0.0005,
+      "step": 23857
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.401269121341464e-06,
+      "loss": 0.5821,
+      "step": 23858
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3986818128872811e-06,
+      "loss": 0.6832,
+      "step": 23859
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3960968840602272e-06,
+      "loss": 0.6008,
+      "step": 23860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.393514334901652e-06,
+      "loss": 1.0003,
+      "step": 23861
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3909341654528893e-06,
+      "loss": 0.3073,
+      "step": 23862
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.388356375755323e-06,
+      "loss": 0.3777,
+      "step": 23863
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3857809658501695e-06,
+      "loss": 0.094,
+      "step": 23864
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3832079357786963e-06,
+      "loss": 0.2267,
+      "step": 23865
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3806372855821202e-06,
+      "loss": 0.6822,
+      "step": 23866
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.378069015301575e-06,
+      "loss": 0.5309,
+      "step": 23867
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3755031249782111e-06,
+      "loss": 0.0031,
+      "step": 23868
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3729396146531125e-06,
+      "loss": 0.7553,
+      "step": 23869
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3703784843673293e-06,
+      "loss": 0.7793,
+      "step": 23870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3678197341618623e-06,
+      "loss": 0.5341,
+      "step": 23871
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3652633640777123e-06,
+      "loss": 0.7778,
+      "step": 23872
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362709374155796e-06,
+      "loss": 0.0039,
+      "step": 23873
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3601577644370144e-06,
+      "loss": 0.3078,
+      "step": 23874
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.357608534962218e-06,
+      "loss": 0.4793,
+      "step": 23875
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3550616857722573e-06,
+      "loss": 0.142,
+      "step": 23876
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3525172169078835e-06,
+      "loss": 0.2024,
+      "step": 23877
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3499751284098638e-06,
+      "loss": 0.1854,
+      "step": 23878
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3474354203189153e-06,
+      "loss": 0.003,
+      "step": 23879
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3448980926756558e-06,
+      "loss": 0.0004,
+      "step": 23880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3423631455207862e-06,
+      "loss": 1.5354,
+      "step": 23881
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.339830578894857e-06,
+      "loss": 0.4745,
+      "step": 23882
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3373003928384195e-06,
+      "loss": 0.0023,
+      "step": 23883
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3347725873920245e-06,
+      "loss": 0.0025,
+      "step": 23884
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3322471625961228e-06,
+      "loss": 0.761,
+      "step": 23885
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3297241184911656e-06,
+      "loss": 0.2558,
+      "step": 23886
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3272034551175537e-06,
+      "loss": 0.4056,
+      "step": 23887
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3246851725156548e-06,
+      "loss": 0.4404,
+      "step": 23888
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3221692707257869e-06,
+      "loss": 3.7245,
+      "step": 23889
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.319655749788251e-06,
+      "loss": 3.3761,
+      "step": 23890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3171446097432814e-06,
+      "loss": 0.3849,
+      "step": 23891
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3146358506311127e-06,
+      "loss": 0.0005,
+      "step": 23892
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3121294724918963e-06,
+      "loss": 0.4033,
+      "step": 23893
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3096254753657665e-06,
+      "loss": 2.2383,
+      "step": 23894
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3071238592928413e-06,
+      "loss": 0.5938,
+      "step": 23895
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.304624624313172e-06,
+      "loss": 0.1045,
+      "step": 23896
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3021277704667765e-06,
+      "loss": 0.7125,
+      "step": 23897
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2996332977936396e-06,
+      "loss": 2.491,
+      "step": 23898
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2971412063337128e-06,
+      "loss": 1.7324,
+      "step": 23899
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2946514961268973e-06,
+      "loss": 0.6577,
+      "step": 23900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2921641672130612e-06,
+      "loss": 0.7499,
+      "step": 23901
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2896792196320393e-06,
+      "loss": 0.4181,
+      "step": 23902
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2871966534236166e-06,
+      "loss": 1.8764,
+      "step": 23903
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2847164686275612e-06,
+      "loss": 0.1166,
+      "step": 23904
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2822386652835747e-06,
+      "loss": 0.6965,
+      "step": 23905
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.279763243431342e-06,
+      "loss": 0.0003,
+      "step": 23906
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2772902031104982e-06,
+      "loss": 0.2753,
+      "step": 23907
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2748195443606446e-06,
+      "loss": 0.1007,
+      "step": 23908
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.272351267221383e-06,
+      "loss": 0.5439,
+      "step": 23909
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2698853717321821e-06,
+      "loss": 1.043,
+      "step": 23910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2674218579325434e-06,
+      "loss": 0.1906,
+      "step": 23911
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.264960725861952e-06,
+      "loss": 0.4359,
+      "step": 23912
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2625019755597766e-06,
+      "loss": 0.5662,
+      "step": 23913
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2600456070654186e-06,
+      "loss": 0.0016,
+      "step": 23914
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2575916204181968e-06,
+      "loss": 0.3422,
+      "step": 23915
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2551400156574298e-06,
+      "loss": 0.267,
+      "step": 23916
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2526907928223362e-06,
+      "loss": 0.2876,
+      "step": 23917
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2502439519521845e-06,
+      "loss": 0.3047,
+      "step": 23918
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2477994930861101e-06,
+      "loss": 0.002,
+      "step": 23919
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2453574162632984e-06,
+      "loss": 0.0012,
+      "step": 23920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2429177215228347e-06,
+      "loss": 0.005,
+      "step": 23921
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2404804089037879e-06,
+      "loss": 0.5747,
+      "step": 23922
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2380454784451932e-06,
+      "loss": 0.0185,
+      "step": 23923
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2356129301860361e-06,
+      "loss": 0.3644,
+      "step": 23924
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2331827641652858e-06,
+      "loss": 0.6876,
+      "step": 23925
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2307549804218276e-06,
+      "loss": 1.3331,
+      "step": 23926
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2283295789945635e-06,
+      "loss": 0.7162,
+      "step": 23927
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2259065599223295e-06,
+      "loss": 0.6578,
+      "step": 23928
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2234859232439109e-06,
+      "loss": 0.0003,
+      "step": 23929
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2210676689980935e-06,
+      "loss": 0.5873,
+      "step": 23930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2186517972235966e-06,
+      "loss": 0.7951,
+      "step": 23931
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2162383079590887e-06,
+      "loss": 0.015,
+      "step": 23932
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2138272012432393e-06,
+      "loss": 0.2975,
+      "step": 23933
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2114184771146673e-06,
+      "loss": 0.3921,
+      "step": 23934
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.209012135611892e-06,
+      "loss": 2.0854,
+      "step": 23935
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2066081767735158e-06,
+      "loss": 0.5542,
+      "step": 23936
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2042066006379914e-06,
+      "loss": 0.1886,
+      "step": 23937
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.2018074072437712e-06,
+      "loss": 0.79,
+      "step": 23938
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1994105966293078e-06,
+      "loss": 0.1752,
+      "step": 23939
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1970161688329704e-06,
+      "loss": 0.1677,
+      "step": 23940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1946241238930786e-06,
+      "loss": 0.3794,
+      "step": 23941
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1922344618479685e-06,
+      "loss": 0.2504,
+      "step": 23942
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1898471827358924e-06,
+      "loss": 0.1943,
+      "step": 23943
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1874622865950867e-06,
+      "loss": 0.4464,
+      "step": 23944
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1850797734637208e-06,
+      "loss": 0.0019,
+      "step": 23945
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.182699643379964e-06,
+      "loss": 0.4155,
+      "step": 23946
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.180321896381936e-06,
+      "loss": 0.0028,
+      "step": 23947
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1779465325077065e-06,
+      "loss": 0.178,
+      "step": 23948
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1755735517953113e-06,
+      "loss": 0.4495,
+      "step": 23949
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1732029542827538e-06,
+      "loss": 0.131,
+      "step": 23950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1708347400079866e-06,
+      "loss": 0.738,
+      "step": 23951
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.168468909008946e-06,
+      "loss": 0.232,
+      "step": 23952
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1661054613235022e-06,
+      "loss": 0.0016,
+      "step": 23953
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1637443969895243e-06,
+      "loss": 0.4856,
+      "step": 23954
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.161385716044816e-06,
+      "loss": 0.8018,
+      "step": 23955
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.159029418527113e-06,
+      "loss": 0.0132,
+      "step": 23956
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.156675504474186e-06,
+      "loss": 0.554,
+      "step": 23957
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.154323973923721e-06,
+      "loss": 0.0008,
+      "step": 23958
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.151974826913371e-06,
+      "loss": 0.345,
+      "step": 23959
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1496280634807564e-06,
+      "loss": 0.2843,
+      "step": 23960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1472836836634468e-06,
+      "loss": 0.5886,
+      "step": 23961
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1449416874989958e-06,
+      "loss": 0.0104,
+      "step": 23962
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1426020750249065e-06,
+      "loss": 0.3637,
+      "step": 23963
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.140264846278649e-06,
+      "loss": 1.6752,
+      "step": 23964
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1379300012976267e-06,
+      "loss": 0.2997,
+      "step": 23965
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1355975401192596e-06,
+      "loss": 1.0503,
+      "step": 23966
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.133267462780868e-06,
+      "loss": 0.3743,
+      "step": 23967
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1309397693197886e-06,
+      "loss": 0.411,
+      "step": 23968
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1286144597732917e-06,
+      "loss": 1.4744,
+      "step": 23969
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1262915341786137e-06,
+      "loss": 0.0081,
+      "step": 23970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1239709925729256e-06,
+      "loss": 0.1498,
+      "step": 23971
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1216528349934305e-06,
+      "loss": 0.3523,
+      "step": 23972
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1193370614772324e-06,
+      "loss": 0.1787,
+      "step": 23973
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1170236720614012e-06,
+      "loss": 0.2413,
+      "step": 23974
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1147126667829908e-06,
+      "loss": 0.2026,
+      "step": 23975
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1124040456790217e-06,
+      "loss": 0.0017,
+      "step": 23976
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1100978087864642e-06,
+      "loss": 0.0016,
+      "step": 23977
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1077939561422055e-06,
+      "loss": 0.2211,
+      "step": 23978
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1054924877831995e-06,
+      "loss": 0.4559,
+      "step": 23979
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1031934037462665e-06,
+      "loss": 0.2057,
+      "step": 23980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.1008967040682103e-06,
+      "loss": 0.4065,
+      "step": 23981
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0986023887858354e-06,
+      "loss": 1.4298,
+      "step": 23982
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.096310457935895e-06,
+      "loss": 0.0028,
+      "step": 23983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0940209115550435e-06,
+      "loss": 0.7001,
+      "step": 23984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.091733749679985e-06,
+      "loss": 0.4199,
+      "step": 23985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0894489723473232e-06,
+      "loss": 0.0013,
+      "step": 23986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0871665795936457e-06,
+      "loss": 0.5593,
+      "step": 23987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0848865714555232e-06,
+      "loss": 0.029,
+      "step": 23988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0826089479694434e-06,
+      "loss": 1.065,
+      "step": 23989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0803337091718766e-06,
+      "loss": 2.0432,
+      "step": 23990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0780608550992942e-06,
+      "loss": 0.4346,
+      "step": 23991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0757903857880501e-06,
+      "loss": 0.2591,
+      "step": 23992
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.073522301274532e-06,
+      "loss": 3.3966,
+      "step": 23993
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0712566015950275e-06,
+      "loss": 0.7078,
+      "step": 23994
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0689932867858575e-06,
+      "loss": 0.7432,
+      "step": 23995
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0667323568832264e-06,
+      "loss": 0.6019,
+      "step": 23996
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.064473811923372e-06,
+      "loss": 0.527,
+      "step": 23997
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0622176519424486e-06,
+      "loss": 0.5221,
+      "step": 23998
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0599638769765772e-06,
+      "loss": 0.886,
+      "step": 23999
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0577124870618626e-06,
+      "loss": 0.4493,
+      "step": 24000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0554634822343588e-06,
+      "loss": 0.4275,
+      "step": 24001
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.053216862530054e-06,
+      "loss": 0.0028,
+      "step": 24002
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0509726279849529e-06,
+      "loss": 0.8951,
+      "step": 24003
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.048730778634993e-06,
+      "loss": 1.9031,
+      "step": 24004
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0464913145160626e-06,
+      "loss": 0.573,
+      "step": 24005
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.044254235664016e-06,
+      "loss": 0.5433,
+      "step": 24006
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0420195421146915e-06,
+      "loss": 0.3807,
+      "step": 24007
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.039787233903877e-06,
+      "loss": 0.01,
+      "step": 24008
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.037557311067294e-06,
+      "loss": 0.0012,
+      "step": 24009
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0353297736406807e-06,
+      "loss": 1.0378,
+      "step": 24010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0331046216596917e-06,
+      "loss": 0.0025,
+      "step": 24011
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0308818551599652e-06,
+      "loss": 0.0013,
+      "step": 24012
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.028661474177106e-06,
+      "loss": 0.0231,
+      "step": 24013
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0264434787466525e-06,
+      "loss": 0.7317,
+      "step": 24014
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.024227868904126e-06,
+      "loss": 0.2633,
+      "step": 24015
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0220146446849986e-06,
+      "loss": 0.0012,
+      "step": 24016
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0198038061247415e-06,
+      "loss": 1.0317,
+      "step": 24017
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0175953532587266e-06,
+      "loss": 0.6514,
+      "step": 24018
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0153892861223257e-06,
+      "loss": 0.2217,
+      "step": 24019
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0131856047508769e-06,
+      "loss": 0.6041,
+      "step": 24020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0109843091796687e-06,
+      "loss": 0.362,
+      "step": 24021
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.00878539944394e-06,
+      "loss": 0.5586,
+      "step": 24022
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.006588875578912e-06,
+      "loss": 0.3274,
+      "step": 24023
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.004394737619757e-06,
+      "loss": 0.1933,
+      "step": 24024
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0022029856015967e-06,
+      "loss": 0.4569,
+      "step": 24025
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0000136195595533e-06,
+      "loss": 0.3118,
+      "step": 24026
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.97826639528665e-07,
+      "loss": 0.2371,
+      "step": 24027
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.956420455439707e-07,
+      "loss": 0.5401,
+      "step": 24028
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.934598376404422e-07,
+      "loss": 0.3895,
+      "step": 24029
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.912800158530187e-07,
+      "loss": 0.6318,
+      "step": 24030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.89102580216622e-07,
+      "loss": 0.2535,
+      "step": 24031
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.869275307661239e-07,
+      "loss": 0.5785,
+      "step": 24032
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.847548675363304e-07,
+      "loss": 0.4667,
+      "step": 24033
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.825845905620632e-07,
+      "loss": 0.0018,
+      "step": 24034
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.80416699878045e-07,
+      "loss": 0.7586,
+      "step": 24035
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.782511955189975e-07,
+      "loss": 0.3367,
+      "step": 24036
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.760880775196267e-07,
+      "loss": 0.0024,
+      "step": 24037
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.73927345914538e-07,
+      "loss": 0.6601,
+      "step": 24038
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.71769000738354e-07,
+      "loss": 1.0945,
+      "step": 24039
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.696130420256298e-07,
+      "loss": 0.901,
+      "step": 24040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.674594698108885e-07,
+      "loss": 0.4199,
+      "step": 24041
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.65308284128602e-07,
+      "loss": 0.2053,
+      "step": 24042
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.631594850132595e-07,
+      "loss": 0.1698,
+      "step": 24043
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.610130724992338e-07,
+      "loss": 0.3007,
+      "step": 24044
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.58869046620897e-07,
+      "loss": 0.0279,
+      "step": 24045
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.567274074126053e-07,
+      "loss": 0.3178,
+      "step": 24046
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.54588154908631e-07,
+      "loss": 0.0097,
+      "step": 24047
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.524512891432634e-07,
+      "loss": 1.1889,
+      "step": 24048
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.503168101506753e-07,
+      "loss": 0.6717,
+      "step": 24049
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.481847179650892e-07,
+      "loss": 0.1171,
+      "step": 24050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.460550126206112e-07,
+      "loss": 0.5511,
+      "step": 24051
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.439276941513807e-07,
+      "loss": 1.0878,
+      "step": 24052
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.41802762591437e-07,
+      "loss": 0.1121,
+      "step": 24053
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.396802179748197e-07,
+      "loss": 1.0018,
+      "step": 24054
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.375600603355016e-07,
+      "loss": 0.8584,
+      "step": 24055
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.35442289707472e-07,
+      "loss": 4.7116,
+      "step": 24056
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.333269061246041e-07,
+      "loss": 0.7578,
+      "step": 24057
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.312139096207871e-07,
+      "loss": 0.2444,
+      "step": 24058
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.291033002298776e-07,
+      "loss": 0.3444,
+      "step": 24059
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.269950779856317e-07,
+      "loss": 0.6592,
+      "step": 24060
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.248892429218391e-07,
+      "loss": 0.2581,
+      "step": 24061
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.227857950722228e-07,
+      "loss": 0.824,
+      "step": 24062
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.20684734470456e-07,
+      "loss": 0.4135,
+      "step": 24063
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.185860611501783e-07,
+      "loss": 1.4463,
+      "step": 24064
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.164897751450129e-07,
+      "loss": 0.3857,
+      "step": 24065
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.143958764885163e-07,
+      "loss": 0.4046,
+      "step": 24066
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.123043652142281e-07,
+      "loss": 0.3062,
+      "step": 24067
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.102152413556385e-07,
+      "loss": 0.2967,
+      "step": 24068
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.081285049461873e-07,
+      "loss": 0.0022,
+      "step": 24069
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.060441560193143e-07,
+      "loss": 0.2715,
+      "step": 24070
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.039621946083764e-07,
+      "loss": 0.1655,
+      "step": 24071
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.0188262074673e-07,
+      "loss": 0.8416,
+      "step": 24072
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.998054344676487e-07,
+      "loss": 0.8098,
+      "step": 24073
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.977306358044389e-07,
+      "loss": 0.004,
+      "step": 24074
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.956582247903077e-07,
+      "loss": 0.6743,
+      "step": 24075
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.935882014584117e-07,
+      "loss": 1.1707,
+      "step": 24076
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.915205658419244e-07,
+      "loss": 0.6382,
+      "step": 24077
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.894553179739694e-07,
+      "loss": 0.333,
+      "step": 24078
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.873924578876035e-07,
+      "loss": 0.367,
+      "step": 24079
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.853319856158502e-07,
+      "loss": 0.2,
+      "step": 24080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.832739011917334e-07,
+      "loss": 0.4908,
+      "step": 24081
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.812182046481598e-07,
+      "loss": 0.0021,
+      "step": 24082
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.791648960181196e-07,
+      "loss": 0.2261,
+      "step": 24083
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.771139753344203e-07,
+      "loss": 0.0027,
+      "step": 24084
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.75065442629952e-07,
+      "loss": 0.132,
+      "step": 24085
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.730192979375216e-07,
+      "loss": 0.3969,
+      "step": 24086
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.709755412898534e-07,
+      "loss": 0.504,
+      "step": 24087
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.689341727197041e-07,
+      "loss": 0.0149,
+      "step": 24088
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.668951922597811e-07,
+      "loss": 0.4285,
+      "step": 24089
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.648585999426916e-07,
+      "loss": 0.6926,
+      "step": 24090
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.628243958010761e-07,
+      "loss": 0.3429,
+      "step": 24091
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.607925798675253e-07,
+      "loss": 0.2126,
+      "step": 24092
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.587631521745298e-07,
+      "loss": 0.0012,
+      "step": 24093
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.567361127546135e-07,
+      "loss": 0.5309,
+      "step": 24094
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.547114616402506e-07,
+      "loss": 0.0013,
+      "step": 24095
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.526891988638318e-07,
+      "loss": 0.5122,
+      "step": 24096
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.506693244577645e-07,
+      "loss": 0.1579,
+      "step": 24097
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.486518384543895e-07,
+      "loss": 0.3103,
+      "step": 24098
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.466367408860142e-07,
+      "loss": 0.2806,
+      "step": 24099
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.446240317848963e-07,
+      "loss": 0.2381,
+      "step": 24100
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.426137111832765e-07,
+      "loss": 0.7108,
+      "step": 24101
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.406057791133458e-07,
+      "loss": 0.4794,
+      "step": 24102
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.386002356072452e-07,
+      "loss": 1.0908,
+      "step": 24103
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.36597080697099e-07,
+      "loss": 1.0825,
+      "step": 24104
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.345963144150147e-07,
+      "loss": 0.2591,
+      "step": 24105
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.325979367929836e-07,
+      "loss": 0.2077,
+      "step": 24106
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.306019478630299e-07,
+      "loss": 0.3375,
+      "step": 24107
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.286083476571115e-07,
+      "loss": 0.3369,
+      "step": 24108
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.266171362071528e-07,
+      "loss": 0.3074,
+      "step": 24109
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.246283135450616e-07,
+      "loss": 0.9849,
+      "step": 24110
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.226418797026624e-07,
+      "loss": 0.0024,
+      "step": 24111
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.206578347117631e-07,
+      "loss": 0.9271,
+      "step": 24112
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.186761786041385e-07,
+      "loss": 0.7734,
+      "step": 24113
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.166969114115295e-07,
+      "loss": 0.8556,
+      "step": 24114
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.147200331656112e-07,
+      "loss": 0.4031,
+      "step": 24115
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.127455438980745e-07,
+      "loss": 0.3174,
+      "step": 24116
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.107734436405111e-07,
+      "loss": 0.0298,
+      "step": 24117
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.088037324245122e-07,
+      "loss": 1.5518,
+      "step": 24118
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.068364102816194e-07,
+      "loss": 1.4658,
+      "step": 24119
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.048714772433407e-07,
+      "loss": 0.209,
+      "step": 24120
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.029089333411176e-07,
+      "loss": 0.001,
+      "step": 24121
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 8.009487786064084e-07,
+      "loss": 0.7377,
+      "step": 24122
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.989910130705712e-07,
+      "loss": 0.2068,
+      "step": 24123
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.970356367649977e-07,
+      "loss": 0.4646,
+      "step": 24124
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.950826497209628e-07,
+      "loss": 2.3877,
+      "step": 24125
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.931320519697415e-07,
+      "loss": 0.001,
+      "step": 24126
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.911838435426088e-07,
+      "loss": 0.6509,
+      "step": 24127
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.892380244707231e-07,
+      "loss": 0.1047,
+      "step": 24128
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.872945947852427e-07,
+      "loss": 0.6753,
+      "step": 24129
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.853535545173262e-07,
+      "loss": 0.4182,
+      "step": 24130
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.834149036980153e-07,
+      "loss": 0.4923,
+      "step": 24131
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.814786423583852e-07,
+      "loss": 0.3484,
+      "step": 24132
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.795447705294278e-07,
+      "loss": 0.489,
+      "step": 24133
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.776132882421016e-07,
+      "loss": 0.2248,
+      "step": 24134
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.756841955273485e-07,
+      "loss": 0.2256,
+      "step": 24135
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.737574924160605e-07,
+      "loss": 0.792,
+      "step": 24136
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.718331789390964e-07,
+      "loss": 0.3281,
+      "step": 24137
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.699112551272646e-07,
+      "loss": 0.2043,
+      "step": 24138
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.67991721011324e-07,
+      "loss": 0.2371,
+      "step": 24139
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.660745766220333e-07,
+      "loss": 0.3947,
+      "step": 24140
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.641598219901013e-07,
+      "loss": 2.493,
+      "step": 24141
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.622474571461867e-07,
+      "loss": 0.1961,
+      "step": 24142
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.603374821208818e-07,
+      "loss": 0.4518,
+      "step": 24143
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.584298969448121e-07,
+      "loss": 0.161,
+      "step": 24144
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.565247016485033e-07,
+      "loss": 0.2442,
+      "step": 24145
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.546218962624639e-07,
+      "loss": 0.0888,
+      "step": 24146
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.5272148081717e-07,
+      "loss": 0.2533,
+      "step": 24147
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.50823455343047e-07,
+      "loss": 0.6452,
+      "step": 24148
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.48927819870504e-07,
+      "loss": 0.3045,
+      "step": 24149
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.470345744298834e-07,
+      "loss": 0.4171,
+      "step": 24150
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.451437190515108e-07,
+      "loss": 0.3395,
+      "step": 24151
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.432552537656455e-07,
+      "loss": 0.7558,
+      "step": 24152
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.413691786025633e-07,
+      "loss": 0.6297,
+      "step": 24153
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.394854935924399e-07,
+      "loss": 0.3062,
+      "step": 24154
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.376041987654513e-07,
+      "loss": 1.2552,
+      "step": 24155
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.357252941517234e-07,
+      "loss": 0.706,
+      "step": 24156
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.33848779781332e-07,
+      "loss": 0.0014,
+      "step": 24157
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.319746556843365e-07,
+      "loss": 0.3636,
+      "step": 24158
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.301029218907461e-07,
+      "loss": 0.3029,
+      "step": 24159
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.282335784305371e-07,
+      "loss": 0.2579,
+      "step": 24160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.26366625333652e-07,
+      "loss": 0.0056,
+      "step": 24161
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.245020626299669e-07,
+      "loss": 0.5962,
+      "step": 24162
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.226398903493413e-07,
+      "loss": 0.139,
+      "step": 24163
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.20780108521618e-07,
+      "loss": 0.6096,
+      "step": 24164
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.189227171765565e-07,
+      "loss": 1.1564,
+      "step": 24165
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.170677163439164e-07,
+      "loss": 0.2901,
+      "step": 24166
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.152151060533906e-07,
+      "loss": 0.6085,
+      "step": 24167
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.133648863346386e-07,
+      "loss": 0.2161,
+      "step": 24168
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.115170572173201e-07,
+      "loss": 0.3869,
+      "step": 24169
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.096716187309947e-07,
+      "loss": 0.1531,
+      "step": 24170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.078285709052223e-07,
+      "loss": 0.2869,
+      "step": 24171
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.059879137695124e-07,
+      "loss": 0.8849,
+      "step": 24172
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.041496473533581e-07,
+      "loss": 0.5311,
+      "step": 24173
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.02313771686186e-07,
+      "loss": 0.0741,
+      "step": 24174
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 7.00480286797389e-07,
+      "loss": 0.0008,
+      "step": 24175
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.98649192716344e-07,
+      "loss": 0.3188,
+      "step": 24176
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.968204894723606e-07,
+      "loss": 0.3791,
+      "step": 24177
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.949941770947154e-07,
+      "loss": 0.281,
+      "step": 24178
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.931702556126683e-07,
+      "loss": 2.4977,
+      "step": 24179
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.913487250554129e-07,
+      "loss": 0.0272,
+      "step": 24180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.895295854521421e-07,
+      "loss": 0.1906,
+      "step": 24181
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.877128368319662e-07,
+      "loss": 3.0022,
+      "step": 24182
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.858984792239619e-07,
+      "loss": 0.3324,
+      "step": 24183
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.840865126572393e-07,
+      "loss": 0.2939,
+      "step": 24184
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.822769371607584e-07,
+      "loss": 0.8252,
+      "step": 24185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.804697527635128e-07,
+      "loss": 0.5511,
+      "step": 24186
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.786649594944627e-07,
+      "loss": 0.55,
+      "step": 24187
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.768625573824849e-07,
+      "loss": 0.5297,
+      "step": 24188
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.750625464564397e-07,
+      "loss": 0.58,
+      "step": 24189
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.732649267451873e-07,
+      "loss": 0.4513,
+      "step": 24190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.714696982774714e-07,
+      "loss": 0.0008,
+      "step": 24191
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.696768610820691e-07,
+      "loss": 0.4957,
+      "step": 24192
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.678864151876572e-07,
+      "loss": 0.5398,
+      "step": 24193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.660983606229464e-07,
+      "loss": 2.7715,
+      "step": 24194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.643126974165469e-07,
+      "loss": 0.3115,
+      "step": 24195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.625294255970692e-07,
+      "loss": 0.3521,
+      "step": 24196
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.607485451930405e-07,
+      "loss": 0.4435,
+      "step": 24197
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.589700562330047e-07,
+      "loss": 0.0023,
+      "step": 24198
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.571939587454389e-07,
+      "loss": 0.1084,
+      "step": 24199
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.554202527587704e-07,
+      "loss": 0.5349,
+      "step": 24200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.536489383014098e-07,
+      "loss": 0.6202,
+      "step": 24201
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.518800154017345e-07,
+      "loss": 0.5879,
+      "step": 24202
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.501134840880718e-07,
+      "loss": 0.1223,
+      "step": 24203
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.483493443886656e-07,
+      "loss": 1.0961,
+      "step": 24204
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.465875963318267e-07,
+      "loss": 0.0011,
+      "step": 24205
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.448282399457327e-07,
+      "loss": 0.0011,
+      "step": 24206
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.430712752585443e-07,
+      "loss": 0.3404,
+      "step": 24207
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.413167022984389e-07,
+      "loss": 0.7492,
+      "step": 24208
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.39564521093494e-07,
+      "loss": 0.2227,
+      "step": 24209
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.378147316717375e-07,
+      "loss": 0.0085,
+      "step": 24210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.360673340612465e-07,
+      "loss": 1.2133,
+      "step": 24211
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.343223282899656e-07,
+      "loss": 1.8499,
+      "step": 24212
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.325797143858391e-07,
+      "loss": 0.2622,
+      "step": 24213
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.308394923767779e-07,
+      "loss": 0.334,
+      "step": 24214
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.291016622906597e-07,
+      "loss": 0.0011,
+      "step": 24215
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.273662241552957e-07,
+      "loss": 0.412,
+      "step": 24216
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.256331779984803e-07,
+      "loss": 0.9277,
+      "step": 24217
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.239025238479744e-07,
+      "loss": 0.1232,
+      "step": 24218
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.221742617315062e-07,
+      "loss": 0.1165,
+      "step": 24219
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.204483916767033e-07,
+      "loss": 1.1979,
+      "step": 24220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.187249137112438e-07,
+      "loss": 0.8435,
+      "step": 24221
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.170038278627221e-07,
+      "loss": 0.4454,
+      "step": 24222
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.152851341586829e-07,
+      "loss": 0.9525,
+      "step": 24223
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.135688326266708e-07,
+      "loss": 0.0731,
+      "step": 24224
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.118549232941306e-07,
+      "loss": 0.121,
+      "step": 24225
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.101434061885568e-07,
+      "loss": 0.2099,
+      "step": 24226
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.084342813373277e-07,
+      "loss": 0.2907,
+      "step": 24227
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.067275487678213e-07,
+      "loss": 0.2881,
+      "step": 24228
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.050232085073492e-07,
+      "loss": 0.2608,
+      "step": 24229
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.033212605832394e-07,
+      "loss": 0.3707,
+      "step": 24230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.016217050227201e-07,
+      "loss": 0.7549,
+      "step": 24231
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.99924541853003e-07,
+      "loss": 0.2662,
+      "step": 24232
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.982297711012828e-07,
+      "loss": 0.7116,
+      "step": 24233
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.965373927947048e-07,
+      "loss": 0.2385,
+      "step": 24234
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.948474069603304e-07,
+      "loss": 0.2795,
+      "step": 24235
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.931598136252714e-07,
+      "loss": 0.1405,
+      "step": 24236
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.914746128165226e-07,
+      "loss": 0.2687,
+      "step": 24237
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.897918045610795e-07,
+      "loss": 0.0088,
+      "step": 24238
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.881113888858868e-07,
+      "loss": 0.0008,
+      "step": 24239
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.864333658178566e-07,
+      "loss": 0.0018,
+      "step": 24240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.847577353838507e-07,
+      "loss": 0.1324,
+      "step": 24241
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.830844976107141e-07,
+      "loss": 0.4324,
+      "step": 24242
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.814136525252422e-07,
+      "loss": 0.0044,
+      "step": 24243
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.797452001541802e-07,
+      "loss": 0.2592,
+      "step": 24244
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.780791405242402e-07,
+      "loss": 0.2993,
+      "step": 24245
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.764154736621174e-07,
+      "loss": 0.2238,
+      "step": 24246
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.747541995944405e-07,
+      "loss": 0.5749,
+      "step": 24247
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.730953183478381e-07,
+      "loss": 0.044,
+      "step": 24248
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.714388299488559e-07,
+      "loss": 0.7556,
+      "step": 24249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.697847344239891e-07,
+      "loss": 1.5468,
+      "step": 24250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.681330317997834e-07,
+      "loss": 0.417,
+      "step": 24251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.664837221026674e-07,
+      "loss": 1.1196,
+      "step": 24252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.648368053590202e-07,
+      "loss": 0.1921,
+      "step": 24253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.631922815952539e-07,
+      "loss": 0.3555,
+      "step": 24254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.615501508376974e-07,
+      "loss": 0.0011,
+      "step": 24255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.599104131126298e-07,
+      "loss": 0.0122,
+      "step": 24256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.582730684463132e-07,
+      "loss": 0.388,
+      "step": 24257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.566381168649936e-07,
+      "loss": 0.3934,
+      "step": 24258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.550055583947999e-07,
+      "loss": 0.5974,
+      "step": 24259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.533753930619444e-07,
+      "loss": 0.1875,
+      "step": 24260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.517476208924732e-07,
+      "loss": 0.3747,
+      "step": 24261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.501222419124651e-07,
+      "loss": 0.2725,
+      "step": 24262
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.484992561479661e-07,
+      "loss": 0.4107,
+      "step": 24263
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.468786636249556e-07,
+      "loss": 0.0066,
+      "step": 24264
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.452604643693959e-07,
+      "loss": 1.136,
+      "step": 24265
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.43644658407183e-07,
+      "loss": 0.2735,
+      "step": 24266
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.420312457641962e-07,
+      "loss": 0.5677,
+      "step": 24267
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.404202264662649e-07,
+      "loss": 0.1667,
+      "step": 24268
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.388116005392018e-07,
+      "loss": 0.9129,
+      "step": 24269
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.372053680087696e-07,
+      "loss": 0.7974,
+      "step": 24270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.35601528900681e-07,
+      "loss": 1.4313,
+      "step": 24271
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.340000832406155e-07,
+      "loss": 0.2613,
+      "step": 24272
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.324010310542192e-07,
+      "loss": 0.3041,
+      "step": 24273
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.308043723670885e-07,
+      "loss": 0.6308,
+      "step": 24274
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.292101072048194e-07,
+      "loss": 0.2073,
+      "step": 24275
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.276182355929248e-07,
+      "loss": 0.3553,
+      "step": 24276
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.260287575569011e-07,
+      "loss": 0.1,
+      "step": 24277
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.244416731221945e-07,
+      "loss": 0.0035,
+      "step": 24278
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.228569823142347e-07,
+      "loss": 0.772,
+      "step": 24279
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.212746851583682e-07,
+      "loss": 0.2742,
+      "step": 24280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.196947816799413e-07,
+      "loss": 0.2642,
+      "step": 24281
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.181172719043004e-07,
+      "loss": 1.8968,
+      "step": 24282
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.165421558566419e-07,
+      "loss": 0.0013,
+      "step": 24283
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.149694335622289e-07,
+      "loss": 0.263,
+      "step": 24284
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.13399105046225e-07,
+      "loss": 0.8081,
+      "step": 24285
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.118311703337929e-07,
+      "loss": 1.0379,
+      "step": 24286
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.102656294500296e-07,
+      "loss": 0.4073,
+      "step": 24287
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.087024824200147e-07,
+      "loss": 0.1466,
+      "step": 24288
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.071417292687785e-07,
+      "loss": 0.2553,
+      "step": 24289
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.055833700213008e-07,
+      "loss": 0.7094,
+      "step": 24290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.040274047025617e-07,
+      "loss": 0.593,
+      "step": 24291
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.024738333374412e-07,
+      "loss": 0.0018,
+      "step": 24292
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.009226559508528e-07,
+      "loss": 0.623,
+      "step": 24293
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.993738725676099e-07,
+      "loss": 0.6309,
+      "step": 24294
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.978274832125429e-07,
+      "loss": 0.55,
+      "step": 24295
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.962834879103816e-07,
+      "loss": 0.4629,
+      "step": 24296
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.947418866858732e-07,
+      "loss": 0.5833,
+      "step": 24297
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.932026795636979e-07,
+      "loss": 1.1694,
+      "step": 24298
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.916658665685025e-07,
+      "loss": 0.3362,
+      "step": 24299
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.901314477248841e-07,
+      "loss": 1.6353,
+      "step": 24300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.885994230574564e-07,
+      "loss": 0.0013,
+      "step": 24301
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.870697925906997e-07,
+      "loss": 0.9004,
+      "step": 24302
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.855425563491278e-07,
+      "loss": 0.2777,
+      "step": 24303
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.84017714357221e-07,
+      "loss": 0.2412,
+      "step": 24304
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.824952666393767e-07,
+      "loss": 2.0332,
+      "step": 24305
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.809752132199585e-07,
+      "loss": 0.6935,
+      "step": 24306
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.79457554123347e-07,
+      "loss": 0.8742,
+      "step": 24307
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.779422893738061e-07,
+      "loss": 0.7964,
+      "step": 24308
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.76429418995633e-07,
+      "loss": 0.2073,
+      "step": 24309
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.7491894301304176e-07,
+      "loss": 0.5756,
+      "step": 24310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.7341086145019637e-07,
+      "loss": 0.5639,
+      "step": 24311
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.71905174331294e-07,
+      "loss": 0.9722,
+      "step": 24312
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.71905174331294e-07,
+      "loss": 1.9541,
+      "step": 24313
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.704018816804156e-07,
+      "loss": 0.0012,
+      "step": 24314
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.6890098352162506e-07,
+      "loss": 0.5203,
+      "step": 24315
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.6740247987896996e-07,
+      "loss": 0.2385,
+      "step": 24316
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.659063707764643e-07,
+      "loss": 0.1464,
+      "step": 24317
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.6441265623802236e-07,
+      "loss": 0.0044,
+      "step": 24318
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.629213362876083e-07,
+      "loss": 0.3682,
+      "step": 24319
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.6143241094908637e-07,
+      "loss": 0.0005,
+      "step": 24320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.599458802462874e-07,
+      "loss": 0.0033,
+      "step": 24321
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.584617442030425e-07,
+      "loss": 0.6558,
+      "step": 24322
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.5698000284309923e-07,
+      "loss": 0.3338,
+      "step": 24323
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.5550065619017194e-07,
+      "loss": 0.0027,
+      "step": 24324
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.540237042679917e-07,
+      "loss": 0.6333,
+      "step": 24325
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.5254914710015634e-07,
+      "loss": 0.0011,
+      "step": 24326
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.5107698471033015e-07,
+      "loss": 0.4035,
+      "step": 24327
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.49607217122061e-07,
+      "loss": 0.4076,
+      "step": 24328
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.481398443588802e-07,
+      "loss": 0.0023,
+      "step": 24329
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.4667486644430206e-07,
+      "loss": 0.4812,
+      "step": 24330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.452122834017746e-07,
+      "loss": 0.8785,
+      "step": 24331
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.4375209525471246e-07,
+      "loss": 0.8547,
+      "step": 24332
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.4229430202653014e-07,
+      "loss": 0.2809,
+      "step": 24333
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.4083890374052577e-07,
+      "loss": 0.1926,
+      "step": 24334
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.393859004200473e-07,
+      "loss": 0.1751,
+      "step": 24335
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.379352920883261e-07,
+      "loss": 1.0614,
+      "step": 24336
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.3648707876862697e-07,
+      "loss": 0.1541,
+      "step": 24337
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.350412604841147e-07,
+      "loss": 0.7431,
+      "step": 24338
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.3359783725795404e-07,
+      "loss": 0.1939,
+      "step": 24339
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.321568091132599e-07,
+      "loss": 0.0042,
+      "step": 24340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.307181760730971e-07,
+      "loss": 0.17,
+      "step": 24341
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.2928193816051394e-07,
+      "loss": 0.5927,
+      "step": 24342
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.278480953985086e-07,
+      "loss": 0.3611,
+      "step": 24343
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.2641664781004613e-07,
+      "loss": 0.244,
+      "step": 24344
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.2498759541802485e-07,
+      "loss": 0.3449,
+      "step": 24345
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.2356093824535975e-07,
+      "loss": 0.3545,
+      "step": 24346
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.2213667631488255e-07,
+      "loss": 0.0782,
+      "step": 24347
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.207148096494084e-07,
+      "loss": 0.1218,
+      "step": 24348
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.192953382716857e-07,
+      "loss": 0.2674,
+      "step": 24349
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.17878262204463e-07,
+      "loss": 0.3864,
+      "step": 24350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.16463581470422e-07,
+      "loss": 0.6975,
+      "step": 24351
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.150512960922281e-07,
+      "loss": 0.4616,
+      "step": 24352
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.1364140609249643e-07,
+      "loss": 0.2545,
+      "step": 24353
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.122339114938089e-07,
+      "loss": 0.2106,
+      "step": 24354
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.108288123186809e-07,
+      "loss": 0.6074,
+      "step": 24355
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.094261085896278e-07,
+      "loss": 0.125,
+      "step": 24356
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.080258003291148e-07,
+      "loss": 0.0016,
+      "step": 24357
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.066278875595574e-07,
+      "loss": 0.2239,
+      "step": 24358
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.052323703033544e-07,
+      "loss": 0.0006,
+      "step": 24359
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.038392485828379e-07,
+      "loss": 2.5408,
+      "step": 24360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.024485224203233e-07,
+      "loss": 0.2575,
+      "step": 24361
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.010601918380929e-07,
+      "loss": 0.0014,
+      "step": 24362
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.996742568583622e-07,
+      "loss": 0.3036,
+      "step": 24363
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.9829071750333007e-07,
+      "loss": 0.4247,
+      "step": 24364
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.969095737951455e-07,
+      "loss": 0.2297,
+      "step": 24365
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.955308257559242e-07,
+      "loss": 0.461,
+      "step": 24366
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.941544734077651e-07,
+      "loss": 1.3321,
+      "step": 24367
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.927805167727005e-07,
+      "loss": 0.5392,
+      "step": 24368
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.914089558727129e-07,
+      "loss": 0.327,
+      "step": 24369
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.9003979072978474e-07,
+      "loss": 0.8697,
+      "step": 24370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.886730213658318e-07,
+      "loss": 0.4528,
+      "step": 24371
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.873086478027532e-07,
+      "loss": 0.0063,
+      "step": 24372
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.8594667006238146e-07,
+      "loss": 0.08,
+      "step": 24373
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.8458708816654913e-07,
+      "loss": 0.4627,
+      "step": 24374
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.832299021369889e-07,
+      "loss": 0.2891,
+      "step": 24375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.818751119954833e-07,
+      "loss": 0.2365,
+      "step": 24376
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.805227177636816e-07,
+      "loss": 0.5591,
+      "step": 24377
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.791727194632832e-07,
+      "loss": 0.9267,
+      "step": 24378
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.7782511711587084e-07,
+      "loss": 0.1736,
+      "step": 24379
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.7647991074304386e-07,
+      "loss": 0.4401,
+      "step": 24380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.7513710036633505e-07,
+      "loss": 0.542,
+      "step": 24381
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.737966860072439e-07,
+      "loss": 0.237,
+      "step": 24382
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.7245866768723655e-07,
+      "loss": 0.0057,
+      "step": 24383
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.711230454277625e-07,
+      "loss": 0.0229,
+      "step": 24384
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.69789819250188e-07,
+      "loss": 2.219,
+      "step": 24385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.6845898917584604e-07,
+      "loss": 0.092,
+      "step": 24386
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.6713055522608615e-07,
+      "loss": 0.7895,
+      "step": 24387
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.6580451742215797e-07,
+      "loss": 0.2131,
+      "step": 24388
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.644808757853113e-07,
+      "loss": 0.0167,
+      "step": 24389
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.6315963033671234e-07,
+      "loss": 0.0045,
+      "step": 24390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.618407810975443e-07,
+      "loss": 0.0003,
+      "step": 24391
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.605243280889236e-07,
+      "loss": 0.9588,
+      "step": 24392
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5921027133193334e-07,
+      "loss": 1.5562,
+      "step": 24393
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5789861084760674e-07,
+      "loss": 0.3795,
+      "step": 24394
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5658934665696024e-07,
+      "loss": 1.179,
+      "step": 24395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5528247878094387e-07,
+      "loss": 1.0127,
+      "step": 24396
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5397800724049073e-07,
+      "loss": 0.0056,
+      "step": 24397
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5267593205650093e-07,
+      "loss": 0.3946,
+      "step": 24398
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5137625324980767e-07,
+      "loss": 0.8036,
+      "step": 24399
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5007897084124437e-07,
+      "loss": 0.4707,
+      "step": 24400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4878408485156105e-07,
+      "loss": 0.8204,
+      "step": 24401
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4749159530150785e-07,
+      "loss": 0.1061,
+      "step": 24402
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4620150221178477e-07,
+      "loss": 0.8397,
+      "step": 24403
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4491380560304206e-07,
+      "loss": 0.3412,
+      "step": 24404
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4362850549591316e-07,
+      "loss": 0.4534,
+      "step": 24405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4234560191096495e-07,
+      "loss": 0.465,
+      "step": 24406
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.4106509486873103e-07,
+      "loss": 0.2575,
+      "step": 24407
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.397869843897616e-07,
+      "loss": 0.0013,
+      "step": 24408
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.385112704944737e-07,
+      "loss": 0.1625,
+      "step": 24409
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.372379532033176e-07,
+      "loss": 0.6248,
+      "step": 24410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.359670325367103e-07,
+      "loss": 1.5986,
+      "step": 24411
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.346985085149523e-07,
+      "loss": 0.7869,
+      "step": 24412
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.334323811583772e-07,
+      "loss": 1.0188,
+      "step": 24413
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.321686504872856e-07,
+      "loss": 0.6464,
+      "step": 24414
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.3090731652187785e-07,
+      "loss": 0.2619,
+      "step": 24415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.296483792823712e-07,
+      "loss": 0.2392,
+      "step": 24416
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2839183878891615e-07,
+      "loss": 0.7385,
+      "step": 24417
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2713769506164664e-07,
+      "loss": 0.8886,
+      "step": 24418
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2588594812062997e-07,
+      "loss": 0.2783,
+      "step": 24419
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2463659798593335e-07,
+      "loss": 0.0054,
+      "step": 24420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.233896446775408e-07,
+      "loss": 0.4001,
+      "step": 24421
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.221450882154364e-07,
+      "loss": 0.4851,
+      "step": 24422
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2090292861955415e-07,
+      "loss": 0.2446,
+      "step": 24423
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.196631659097615e-07,
+      "loss": 0.0018,
+      "step": 24424
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.184258001059259e-07,
+      "loss": 0.7047,
+      "step": 24425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1719083122784815e-07,
+      "loss": 0.0851,
+      "step": 24426
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.159582592953458e-07,
+      "loss": 0.3096,
+      "step": 24427
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1472808432810303e-07,
+      "loss": 0.2124,
+      "step": 24428
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.135003063458541e-07,
+      "loss": 1.5709,
+      "step": 24429
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1227492536824996e-07,
+      "loss": 0.1481,
+      "step": 24430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1105194141489153e-07,
+      "loss": 0.001,
+      "step": 24431
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.098313545054132e-07,
+      "loss": 0.7503,
+      "step": 24432
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0861316465931594e-07,
+      "loss": 0.0149,
+      "step": 24433
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0739737189613424e-07,
+      "loss": 0.0169,
+      "step": 24434
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0618397623531907e-07,
+      "loss": 3.1153,
+      "step": 24435
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0497297769632166e-07,
+      "loss": 0.9079,
+      "step": 24436
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.037643762985098e-07,
+      "loss": 0.1589,
+      "step": 24437
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.025581720612513e-07,
+      "loss": 0.4961,
+      "step": 24438
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.013543650038475e-07,
+      "loss": 0.0019,
+      "step": 24439
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0015295514559943e-07,
+      "loss": 0.0012,
+      "step": 24440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.989539425057419e-07,
+      "loss": 0.5534,
+      "step": 24441
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9775732710345944e-07,
+      "loss": 0.0123,
+      "step": 24442
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9656310895792015e-07,
+      "loss": 2.4297,
+      "step": 24443
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.95371288088242e-07,
+      "loss": 0.8762,
+      "step": 24444
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.941818645135263e-07,
+      "loss": 0.5527,
+      "step": 24445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.92994838252808e-07,
+      "loss": 0.1581,
+      "step": 24446
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9181020932510515e-07,
+      "loss": 0.0167,
+      "step": 24447
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9062797774936917e-07,
+      "loss": 0.1952,
+      "step": 24448
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.894481435445517e-07,
+      "loss": 0.3619,
+      "step": 24449
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8827070672953756e-07,
+      "loss": 0.374,
+      "step": 24450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8709566732317836e-07,
+      "loss": 0.7559,
+      "step": 24451
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.85923025344309e-07,
+      "loss": 0.2332,
+      "step": 24452
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8475278081168116e-07,
+      "loss": 0.1458,
+      "step": 24453
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8358493374404656e-07,
+      "loss": 0.4367,
+      "step": 24454
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8241948416010685e-07,
+      "loss": 0.3133,
+      "step": 24455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.812564320785138e-07,
+      "loss": 0.2613,
+      "step": 24456
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.800957775179191e-07,
+      "loss": 1.024,
+      "step": 24457
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.789375204968913e-07,
+      "loss": 0.7664,
+      "step": 24458
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.777816610339656e-07,
+      "loss": 0.2233,
+      "step": 24459
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.766281991476771e-07,
+      "loss": 0.4095,
+      "step": 24460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.754771348564777e-07,
+      "loss": 0.2646,
+      "step": 24461
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7432846817881935e-07,
+      "loss": 0.1558,
+      "step": 24462
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7318219913307073e-07,
+      "loss": 0.5127,
+      "step": 24463
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.720383277376004e-07,
+      "loss": 0.0005,
+      "step": 24464
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.708968540107437e-07,
+      "loss": 0.1551,
+      "step": 24465
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6975777797073606e-07,
+      "loss": 0.7542,
+      "step": 24466
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6862109963586286e-07,
+      "loss": 0.1915,
+      "step": 24467
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6748681902430955e-07,
+      "loss": 0.0676,
+      "step": 24468
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.663549361542283e-07,
+      "loss": 0.2746,
+      "step": 24469
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.652254510437546e-07,
+      "loss": 0.4974,
+      "step": 24470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.640983637109739e-07,
+      "loss": 0.4643,
+      "step": 24471
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6297367417393855e-07,
+      "loss": 0.1148,
+      "step": 24472
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.618513824506507e-07,
+      "loss": 0.0027,
+      "step": 24473
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.607314885590961e-07,
+      "loss": 0.5751,
+      "step": 24474
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.596139925171936e-07,
+      "loss": 0.002,
+      "step": 24475
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5849889434284564e-07,
+      "loss": 1.025,
+      "step": 24476
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.573861940539046e-07,
+      "loss": 0.4553,
+      "step": 24477
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.562758916681895e-07,
+      "loss": 0.8984,
+      "step": 24478
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.551679872035028e-07,
+      "loss": 0.3041,
+      "step": 24479
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.540624806775471e-07,
+      "loss": 0.449,
+      "step": 24480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.529593721080414e-07,
+      "loss": 0.0908,
+      "step": 24481
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.51858661512655e-07,
+      "loss": 0.8688,
+      "step": 24482
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.507603489090071e-07,
+      "loss": 0.6213,
+      "step": 24483
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.496644343147003e-07,
+      "loss": 1.8233,
+      "step": 24484
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4857091774727057e-07,
+      "loss": 0.0211,
+      "step": 24485
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.474797992242372e-07,
+      "loss": 0.1693,
+      "step": 24486
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.463910787630696e-07,
+      "loss": 0.0047,
+      "step": 24487
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4530475638118717e-07,
+      "loss": 0.9326,
+      "step": 24488
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.442208320960093e-07,
+      "loss": 0.1362,
+      "step": 24489
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4313930592487206e-07,
+      "loss": 0.8642,
+      "step": 24490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.420601778851117e-07,
+      "loss": 0.5091,
+      "step": 24491
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.409834479939976e-07,
+      "loss": 0.6134,
+      "step": 24492
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.399091162687827e-07,
+      "loss": 0.6094,
+      "step": 24493
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3883718272666997e-07,
+      "loss": 0.0399,
+      "step": 24494
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3776764738481223e-07,
+      "loss": 0.3856,
+      "step": 24495
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3670051026034587e-07,
+      "loss": 0.2871,
+      "step": 24496
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3563577137035716e-07,
+      "loss": 0.1022,
+      "step": 24497
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3457343073189916e-07,
+      "loss": 1.7642,
+      "step": 24498
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3351348836197493e-07,
+      "loss": 0.3005,
+      "step": 24499
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3245594427757085e-07,
+      "loss": 0.605,
+      "step": 24500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3140079849560676e-07,
+      "loss": 0.1262,
+      "step": 24501
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3034805103300246e-07,
+      "loss": 0.0934,
+      "step": 24502
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2929770190657782e-07,
+      "loss": 0.6977,
+      "step": 24503
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2824975113320265e-07,
+      "loss": 0.3807,
+      "step": 24504
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.272041987296136e-07,
+      "loss": 0.3528,
+      "step": 24505
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2616104471258056e-07,
+      "loss": 0.5434,
+      "step": 24506
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.251202890987902e-07,
+      "loss": 0.1586,
+      "step": 24507
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2408193190492918e-07,
+      "loss": 0.196,
+      "step": 24508
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2304597314760087e-07,
+      "loss": 2.6149,
+      "step": 24509
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.220124128434253e-07,
+      "loss": 0.4087,
+      "step": 24510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2098125100892261e-07,
+      "loss": 0.5492,
+      "step": 24511
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1995248766062956e-07,
+      "loss": 0.335,
+      "step": 24512
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1892612281499967e-07,
+      "loss": 0.7751,
+      "step": 24513
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1790215648846976e-07,
+      "loss": 0.6875,
+      "step": 24514
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1688058869746007e-07,
+      "loss": 0.2203,
+      "step": 24515
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.158614194583075e-07,
+      "loss": 0.0029,
+      "step": 24516
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1484464878734898e-07,
+      "loss": 0.2952,
+      "step": 24517
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1383027670083818e-07,
+      "loss": 0.6203,
+      "step": 24518
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.128183032150621e-07,
+      "loss": 0.5609,
+      "step": 24519
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1180872834617446e-07,
+      "loss": 0.0039,
+      "step": 24520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1080155211039563e-07,
+      "loss": 0.0892,
+      "step": 24521
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0979677452381273e-07,
+      "loss": 0.3771,
+      "step": 24522
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0879439560252954e-07,
+      "loss": 0.0021,
+      "step": 24523
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0779441536261654e-07,
+      "loss": 0.2312,
+      "step": 24524
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0679683382006096e-07,
+      "loss": 0.8883,
+      "step": 24525
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0580165099083334e-07,
+      "loss": 0.3959,
+      "step": 24526
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.048088668908876e-07,
+      "loss": 0.0012,
+      "step": 24527
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0381848153612768e-07,
+      "loss": 0.2289,
+      "step": 24528
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.028304949423909e-07,
+      "loss": 2.03,
+      "step": 24529
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.018449071255146e-07,
+      "loss": 0.0034,
+      "step": 24530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.008617181012695e-07,
+      "loss": 0.5339,
+      "step": 24531
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9988092788540967e-07,
+      "loss": 0.0134,
+      "step": 24532
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9890253649365583e-07,
+      "loss": 0.2806,
+      "step": 24533
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9792654394164553e-07,
+      "loss": 0.2847,
+      "step": 24534
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9695295024501623e-07,
+      "loss": 0.3213,
+      "step": 24535
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9598175541935545e-07,
+      "loss": 0.3083,
+      "step": 24536
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.950129594802341e-07,
+      "loss": 0.0073,
+      "step": 24537
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.940465624431564e-07,
+      "loss": 0.0026,
+      "step": 24538
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9308256432359336e-07,
+      "loss": 0.2564,
+      "step": 24539
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.921209651369826e-07,
+      "loss": 0.2167,
+      "step": 24540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9116176489872847e-07,
+      "loss": 0.4373,
+      "step": 24541
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9020496362418535e-07,
+      "loss": 0.0144,
+      "step": 24542
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8925056132867433e-07,
+      "loss": 1.6242,
+      "step": 24543
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8829855802748317e-07,
+      "loss": 2.0645,
+      "step": 24544
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.873489537358497e-07,
+      "loss": 0.5174,
+      "step": 24545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8640174846897836e-07,
+      "loss": 0.0014,
+      "step": 24546
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.854569422420571e-07,
+      "loss": 0.6522,
+      "step": 24547
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8451453507020708e-07,
+      "loss": 0.7406,
+      "step": 24548
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8357452696849962e-07,
+      "loss": 0.8261,
+      "step": 24549
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8263691795202261e-07,
+      "loss": 1.6,
+      "step": 24550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8170170803574746e-07,
+      "loss": 0.5951,
+      "step": 24551
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8076889723469546e-07,
+      "loss": 2.9039,
+      "step": 24552
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7983848556377135e-07,
+      "loss": 0.6425,
+      "step": 24553
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7891047303787987e-07,
+      "loss": 0.3226,
+      "step": 24554
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7798485967190913e-07,
+      "loss": 0.2598,
+      "step": 24555
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7706164548064727e-07,
+      "loss": 0.9653,
+      "step": 24556
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.761408304788825e-07,
+      "loss": 0.1968,
+      "step": 24557
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7522241468136966e-07,
+      "loss": 0.246,
+      "step": 24558
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7430639810283032e-07,
+      "loss": 0.3772,
+      "step": 24559
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7339278075788608e-07,
+      "loss": 0.3631,
+      "step": 24560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7248156266122525e-07,
+      "loss": 0.2342,
+      "step": 24561
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7157274382740282e-07,
+      "loss": 0.5653,
+      "step": 24562
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7066632427099047e-07,
+      "loss": 0.2143,
+      "step": 24563
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6976230400649327e-07,
+      "loss": 0.0038,
+      "step": 24564
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6886068304838295e-07,
+      "loss": 0.4795,
+      "step": 24565
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.67961461411098e-07,
+      "loss": 0.3453,
+      "step": 24566
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.670646391090602e-07,
+      "loss": 0.3409,
+      "step": 24567
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6617021615659143e-07,
+      "loss": 0.3407,
+      "step": 24568
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.652781925680635e-07,
+      "loss": 0.038,
+      "step": 24569
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.643885683577151e-07,
+      "loss": 0.0129,
+      "step": 24570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6350134353981802e-07,
+      "loss": 1.0948,
+      "step": 24571
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6261651812857767e-07,
+      "loss": 0.04,
+      "step": 24572
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6173409213814935e-07,
+      "loss": 0.0014,
+      "step": 24573
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6085406558268843e-07,
+      "loss": 2.8914,
+      "step": 24574
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5997643847626696e-07,
+      "loss": 0.625,
+      "step": 24575
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5910121083294037e-07,
+      "loss": 0.5201,
+      "step": 24576
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5822838266673078e-07,
+      "loss": 1.7032,
+      "step": 24577
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5735795399161034e-07,
+      "loss": 1.328,
+      "step": 24578
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.564899248215179e-07,
+      "loss": 0.1936,
+      "step": 24579
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5562429517037568e-07,
+      "loss": 0.0006,
+      "step": 24580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.547610650520059e-07,
+      "loss": 0.2339,
+      "step": 24581
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.539002344802476e-07,
+      "loss": 0.514,
+      "step": 24582
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5304180346890627e-07,
+      "loss": 0.7232,
+      "step": 24583
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.521857720316877e-07,
+      "loss": 0.0004,
+      "step": 24584
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.513321401823475e-07,
+      "loss": 0.5848,
+      "step": 24585
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5048090793452482e-07,
+      "loss": 0.2686,
+      "step": 24586
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4963207530185873e-07,
+      "loss": 0.0036,
+      "step": 24587
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4878564229793832e-07,
+      "loss": 0.0015,
+      "step": 24588
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4794160893631945e-07,
+      "loss": 0.6221,
+      "step": 24589
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4709997523050798e-07,
+      "loss": 0.3884,
+      "step": 24590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4626074119400978e-07,
+      "loss": 0.4607,
+      "step": 24591
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4542390684024741e-07,
+      "loss": 0.682,
+      "step": 24592
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4458947218261018e-07,
+      "loss": 0.7354,
+      "step": 24593
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.437574372344874e-07,
+      "loss": 0.7524,
+      "step": 24594
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4292780200916843e-07,
+      "loss": 0.3412,
+      "step": 24595
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4210056651995927e-07,
+      "loss": 0.2943,
+      "step": 24596
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.41275730780116e-07,
+      "loss": 1.3174,
+      "step": 24597
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4045329480282807e-07,
+      "loss": 1.2934,
+      "step": 24598
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3963325860126827e-07,
+      "loss": 0.3015,
+      "step": 24599
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.388156221885761e-07,
+      "loss": 0.252,
+      "step": 24600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.380003855778411e-07,
+      "loss": 0.0094,
+      "step": 24601
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3718754878213611e-07,
+      "loss": 0.4612,
+      "step": 24602
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3637711181443413e-07,
+      "loss": 0.006,
+      "step": 24603
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3556907468775803e-07,
+      "loss": 0.3411,
+      "step": 24604
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3476343741501417e-07,
+      "loss": 0.6831,
+      "step": 24605
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3396020000912556e-07,
+      "loss": 0.4949,
+      "step": 24606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3315936248294856e-07,
+      "loss": 0.2505,
+      "step": 24607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3236092484928962e-07,
+      "loss": 0.0013,
+      "step": 24608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.315648871209718e-07,
+      "loss": 0.0104,
+      "step": 24609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3077124931071826e-07,
+      "loss": 0.8778,
+      "step": 24610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2998001143123548e-07,
+      "loss": 0.9296,
+      "step": 24611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.291911734951967e-07,
+      "loss": 0.695,
+      "step": 24612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.284047355152251e-07,
+      "loss": 0.3504,
+      "step": 24613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.27620697503944e-07,
+      "loss": 0.0023,
+      "step": 24614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2683905947389327e-07,
+      "loss": 2.5735,
+      "step": 24615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2605982143757965e-07,
+      "loss": 0.7385,
+      "step": 24616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.252829834074931e-07,
+      "loss": 0.0067,
+      "step": 24617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2450854539605703e-07,
+      "loss": 0.1558,
+      "step": 24618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2373650741569484e-07,
+      "loss": 0.2447,
+      "step": 24619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2296686947876332e-07,
+      "loss": 0.559,
+      "step": 24620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2219963159756928e-07,
+      "loss": 0.7506,
+      "step": 24621
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2143479378441955e-07,
+      "loss": 0.2521,
+      "step": 24622
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.2067235605155435e-07,
+      "loss": 0.0005,
+      "step": 24623
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1991231841119718e-07,
+      "loss": 0.5572,
+      "step": 24624
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1915468087548841e-07,
+      "loss": 2.7703,
+      "step": 24625
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1839944345658492e-07,
+      "loss": 0.3217,
+      "step": 24626
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1764660616657706e-07,
+      "loss": 0.8501,
+      "step": 24627
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.168961690175052e-07,
+      "loss": 0.0018,
+      "step": 24628
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1614813202140972e-07,
+      "loss": 0.0009,
+      "step": 24629
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1540249519026434e-07,
+      "loss": 1.2907,
+      "step": 24630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.146592585359929e-07,
+      "loss": 0.3249,
+      "step": 24631
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.139184220705025e-07,
+      "loss": 0.3003,
+      "step": 24632
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.13179985805667e-07,
+      "loss": 0.7622,
+      "step": 24633
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1244394975331028e-07,
+      "loss": 0.694,
+      "step": 24634
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1171031392520624e-07,
+      "loss": 0.668,
+      "step": 24635
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1097907833311215e-07,
+      "loss": 0.3131,
+      "step": 24636
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.102502429887353e-07,
+      "loss": 0.2428,
+      "step": 24637
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.095238079037497e-07,
+      "loss": 0.7177,
+      "step": 24638
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0879977308977938e-07,
+      "loss": 0.1599,
+      "step": 24639
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0807813855843173e-07,
+      "loss": 0.2598,
+      "step": 24640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.073589043212475e-07,
+      "loss": 0.0031,
+      "step": 24641
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0664207038976746e-07,
+      "loss": 0.5499,
+      "step": 24642
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0592763677544913e-07,
+      "loss": 0.0136,
+      "step": 24643
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0521560348975e-07,
+      "loss": 0.0009,
+      "step": 24644
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.045059705440443e-07,
+      "loss": 0.6539,
+      "step": 24645
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0379873794972293e-07,
+      "loss": 0.1605,
+      "step": 24646
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0309390571811016e-07,
+      "loss": 0.0094,
+      "step": 24647
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0239147386046365e-07,
+      "loss": 0.5055,
+      "step": 24648
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0169144238807436e-07,
+      "loss": 1.6777,
+      "step": 24649
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.009938113121167e-07,
+      "loss": 0.8293,
+      "step": 24650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0029858064378172e-07,
+      "loss": 0.3197,
+      "step": 24651
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.960575039421048e-08,
+      "loss": 0.0008,
+      "step": 24652
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.891532057446083e-08,
+      "loss": 0.9328,
+      "step": 24653
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.82272911956239e-08,
+      "loss": 0.4021,
+      "step": 24654
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.754166226870752e-08,
+      "loss": 0.504,
+      "step": 24655
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.685843380468628e-08,
+      "loss": 0.0645,
+      "step": 24656
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.61776058145014e-08,
+      "loss": 0.5602,
+      "step": 24657
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.549917830906084e-08,
+      "loss": 0.3508,
+      "step": 24658
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.482315129922257e-08,
+      "loss": 0.7848,
+      "step": 24659
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.414952479582794e-08,
+      "loss": 0.2579,
+      "step": 24660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.347829880963498e-08,
+      "loss": 0.3755,
+      "step": 24661
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.280947335141842e-08,
+      "loss": 0.675,
+      "step": 24662
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.214304843186971e-08,
+      "loss": 0.0021,
+      "step": 24663
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.147902406168028e-08,
+      "loss": 0.3385,
+      "step": 24664
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.0817400251475e-08,
+      "loss": 0.6586,
+      "step": 24665
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.015817701184536e-08,
+      "loss": 1.2821,
+      "step": 24666
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.950135435333294e-08,
+      "loss": 0.3715,
+      "step": 24667
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.884693228649597e-08,
+      "loss": 0.738,
+      "step": 24668
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.81949108217761e-08,
+      "loss": 2.0217,
+      "step": 24669
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.754528996963162e-08,
+      "loss": 0.9212,
+      "step": 24670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.68980697404542e-08,
+      "loss": 0.3217,
+      "step": 24671
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.625325014463558e-08,
+      "loss": 0.2473,
+      "step": 24672
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.561083119246748e-08,
+      "loss": 2.1992,
+      "step": 24673
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.497081289425834e-08,
+      "loss": 0.2402,
+      "step": 24674
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.433319526024996e-08,
+      "loss": 0.351,
+      "step": 24675
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.369797830063419e-08,
+      "loss": 0.293,
+      "step": 24676
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.306516202561952e-08,
+      "loss": 0.2295,
+      "step": 24677
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.243474644533121e-08,
+      "loss": 0.6935,
+      "step": 24678
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.180673156984452e-08,
+      "loss": 0.5524,
+      "step": 24679
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.118111740923472e-08,
+      "loss": 0.4848,
+      "step": 24680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.055790397351047e-08,
+      "loss": 0.0087,
+      "step": 24681
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.993709127264713e-08,
+      "loss": 0.2933,
+      "step": 24682
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.931867931658675e-08,
+      "loss": 0.2128,
+      "step": 24683
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.870266811525471e-08,
+      "loss": 0.7205,
+      "step": 24684
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.808905767849316e-08,
+      "loss": 1.5108,
+      "step": 24685
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.747784801612755e-08,
+      "loss": 0.3053,
+      "step": 24686
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.686903913796672e-08,
+      "loss": 3.5908,
+      "step": 24687
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.62626310537362e-08,
+      "loss": 0.1868,
+      "step": 24688
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.565862377314491e-08,
+      "loss": 0.1153,
+      "step": 24689
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.50570173058851e-08,
+      "loss": 0.3674,
+      "step": 24690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.445781166158238e-08,
+      "loss": 0.1558,
+      "step": 24691
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.386100684982909e-08,
+      "loss": 1.0448,
+      "step": 24692
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.326660288020092e-08,
+      "loss": 0.2707,
+      "step": 24693
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.267459976217361e-08,
+      "loss": 0.8014,
+      "step": 24694
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.208499750527285e-08,
+      "loss": 0.3553,
+      "step": 24695
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.149779611890782e-08,
+      "loss": 0.8085,
+      "step": 24696
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.09129956125043e-08,
+      "loss": 0.3494,
+      "step": 24697
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.03305959954048e-08,
+      "loss": 0.3754,
+      "step": 24698
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.975059727695187e-08,
+      "loss": 0.031,
+      "step": 24699
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.917299946643806e-08,
+      "loss": 0.788,
+      "step": 24700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.859780257308934e-08,
+      "loss": 0.0915,
+      "step": 24701
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.802500660614829e-08,
+      "loss": 0.1047,
+      "step": 24702
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.745461157475762e-08,
+      "loss": 0.6478,
+      "step": 24703
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.688661748807667e-08,
+      "loss": 0.3438,
+      "step": 24704
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.632102435518149e-08,
+      "loss": 0.0016,
+      "step": 24705
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.575783218513153e-08,
+      "loss": 1.4733,
+      "step": 24706
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.519704098695289e-08,
+      "loss": 0.883,
+      "step": 24707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.463865076962171e-08,
+      "loss": 1.6055,
+      "step": 24708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.408266154208086e-08,
+      "loss": 0.2401,
+      "step": 24709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.352907331323986e-08,
+      "loss": 0.5447,
+      "step": 24710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.297788609195831e-08,
+      "loss": 0.2966,
+      "step": 24711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.24290998870458e-08,
+      "loss": 0.1682,
+      "step": 24712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.188271470732863e-08,
+      "loss": 0.4645,
+      "step": 24713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.13387305615165e-08,
+      "loss": 0.3794,
+      "step": 24714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.07971474583524e-08,
+      "loss": 0.0005,
+      "step": 24715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.025796540647943e-08,
+      "loss": 0.293,
+      "step": 24716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.972118441455732e-08,
+      "loss": 2.8483,
+      "step": 24717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.918680449117918e-08,
+      "loss": 0.7065,
+      "step": 24718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.865482564487156e-08,
+      "loss": 0.7889,
+      "step": 24719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.8125247884194257e-08,
+      "loss": 1.5,
+      "step": 24720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.7598071217607175e-08,
+      "loss": 0.1966,
+      "step": 24721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.7073295653553565e-08,
+      "loss": 0.3213,
+      "step": 24722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.655092120044336e-08,
+      "loss": 0.121,
+      "step": 24723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.603094786663653e-08,
+      "loss": 0.235,
+      "step": 24724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.5513375660443116e-08,
+      "loss": 0.6272,
+      "step": 24725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.499820459018977e-08,
+      "loss": 1.9922,
+      "step": 24726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.4485434664086616e-08,
+      "loss": 0.0743,
+      "step": 24727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3975065890377035e-08,
+      "loss": 0.002,
+      "step": 24728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.3467098277204525e-08,
+      "loss": 0.1671,
+      "step": 24729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.296153183272922e-08,
+      "loss": 0.0043,
+      "step": 24730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.2458366565028e-08,
+      "loss": 1.0503,
+      "step": 24731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.195760248217773e-08,
+      "loss": 0.3409,
+      "step": 24732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.145923959217202e-08,
+      "loss": 0.426,
+      "step": 24733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.096327790300447e-08,
+      "loss": 0.3437,
+      "step": 24734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.0469717422635394e-08,
+      "loss": 0.1933,
+      "step": 24735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.997855815894181e-08,
+      "loss": 0.3771,
+      "step": 24736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.948980011980075e-08,
+      "loss": 1.5459,
+      "step": 24737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.900344331302264e-08,
+      "loss": 0.0008,
+      "step": 24738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.851948774643455e-08,
+      "loss": 0.4125,
+      "step": 24739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.803793342774698e-08,
+      "loss": 0.0095,
+      "step": 24740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7558780364670427e-08,
+      "loss": 0.395,
+      "step": 24741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.708202856491539e-08,
+      "loss": 0.2467,
+      "step": 24742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6607678036092444e-08,
+      "loss": 0.0009,
+      "step": 24743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.613572878577887e-08,
+      "loss": 0.5368,
+      "step": 24744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.566618082156859e-08,
+      "loss": 0.5837,
+      "step": 24745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.519903415097226e-08,
+      "loss": 0.1403,
+      "step": 24746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4734288781450576e-08,
+      "loss": 0.647,
+      "step": 24747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4271944720464245e-08,
+      "loss": 0.1306,
+      "step": 24748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3812001975407354e-08,
+      "loss": 0.6214,
+      "step": 24749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.3354460553640666e-08,
+      "loss": 0.2824,
+      "step": 24750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.289932046250833e-08,
+      "loss": 1.0784,
+      "step": 24751
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.244658170928783e-08,
+      "loss": 0.002,
+      "step": 24752
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.199624430122339e-08,
+      "loss": 0.0012,
+      "step": 24753
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.154830824554256e-08,
+      "loss": 1.0963,
+      "step": 24754
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.1102773549406274e-08,
+      "loss": 0.3534,
+      "step": 24755
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0659640219958825e-08,
+      "loss": 0.8543,
+      "step": 24756
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0218908264277874e-08,
+      "loss": 1.2395,
+      "step": 24757
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9780577689441096e-08,
+      "loss": 0.6494,
+      "step": 24758
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9344648502459553e-08,
+      "loss": 0.0021,
+      "step": 24759
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.891112071031099e-08,
+      "loss": 0.2348,
+      "step": 24760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8479994319939865e-08,
+      "loss": 1.1155,
+      "step": 24761
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8051269338257304e-08,
+      "loss": 0.067,
+      "step": 24762
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.762494577210784e-08,
+      "loss": 0.5789,
+      "step": 24763
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7201023628352645e-08,
+      "loss": 0.1141,
+      "step": 24764
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6779502913736325e-08,
+      "loss": 0.2295,
+      "step": 24765
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6360383635053445e-08,
+      "loss": 0.9971,
+      "step": 24766
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.594366579899866e-08,
+      "loss": 0.2213,
+      "step": 24767
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.552934941221663e-08,
+      "loss": 0.875,
+      "step": 24768
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5117434481385376e-08,
+      "loss": 0.2095,
+      "step": 24769
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4707921013082953e-08,
+      "loss": 0.9813,
+      "step": 24770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.430080901385412e-08,
+      "loss": 0.7624,
+      "step": 24771
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3896098490243664e-08,
+      "loss": 0.8089,
+      "step": 24772
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.349378944871306e-08,
+      "loss": 0.009,
+      "step": 24773
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.309388189572382e-08,
+      "loss": 0.0108,
+      "step": 24774
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.269637583765416e-08,
+      "loss": 2.3535,
+      "step": 24775
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2301271280882335e-08,
+      "loss": 0.2984,
+      "step": 24776
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.19085682317366e-08,
+      "loss": 0.3338,
+      "step": 24777
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.151826669651192e-08,
+      "loss": 0.2505,
+      "step": 24778
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.11303666814533e-08,
+      "loss": 0.0056,
+      "step": 24779
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0744868192755786e-08,
+      "loss": 0.2282,
+      "step": 24780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.036177123661443e-08,
+      "loss": 0.8762,
+      "step": 24781
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.998107581915765e-08,
+      "loss": 0.3113,
+      "step": 24782
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.960278194648058e-08,
+      "loss": 0.516,
+      "step": 24783
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9226889624628382e-08,
+      "loss": 0.005,
+      "step": 24784
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8853398859646216e-08,
+      "loss": 0.3281,
+      "step": 24785
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8482309657495984e-08,
+      "loss": 0.4478,
+      "step": 24786
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.811362202413958e-08,
+      "loss": 0.67,
+      "step": 24787
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7747335965455642e-08,
+      "loss": 0.3731,
+      "step": 24788
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7383451487322793e-08,
+      "loss": 0.0011,
+      "step": 24789
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.702196859556971e-08,
+      "loss": 0.7522,
+      "step": 24790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.66628872959751e-08,
+      "loss": 0.7466,
+      "step": 24791
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.630620759430102e-08,
+      "loss": 0.6596,
+      "step": 24792
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5951929496242918e-08,
+      "loss": 0.2057,
+      "step": 24793
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5600053007496236e-08,
+      "loss": 0.3669,
+      "step": 24794
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5250578133689804e-08,
+      "loss": 0.0034,
+      "step": 24795
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4903504880402492e-08,
+      "loss": 0.1881,
+      "step": 24796
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4558833253213172e-08,
+      "loss": 0.4602,
+      "step": 24797
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4216563257617448e-08,
+      "loss": 0.9811,
+      "step": 24798
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.387669489912758e-08,
+      "loss": 2.8555,
+      "step": 24799
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3539228183172552e-08,
+      "loss": 0.0015,
+      "step": 24800
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 24926,
+  "num_train_epochs": 1,
+  "save_steps": 200,
+  "total_flos": 5.469857088436077e+18,
+  "trial_name": null,
+  "trial_params": null
+}