diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,75016 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.49328637247067414,
+  "global_step": 12500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099996,
+      "loss": 19.0668,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099992,
+      "loss": 20.4419,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099988,
+      "loss": 18.403,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009998399999999999,
+      "loss": 19.0156,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009998000000000001,
+      "loss": 12.2304,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099976,
+      "loss": 11.9098,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009997200000000002,
+      "loss": 11.8655,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099968,
+      "loss": 11.103,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099964,
+      "loss": 8.8224,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009996,
+      "loss": 6.0737,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099956,
+      "loss": 5.9995,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099952,
+      "loss": 5.8613,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099948,
+      "loss": 6.0743,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099944,
+      "loss": 5.7928,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009994,
+      "loss": 5.1918,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009993600000000001,
+      "loss": 4.59,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099932,
+      "loss": 6.1942,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099928,
+      "loss": 4.9524,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099924,
+      "loss": 4.2394,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009992,
+      "loss": 4.9486,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099916,
+      "loss": 4.149,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099912,
+      "loss": 3.6541,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099908,
+      "loss": 3.6981,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099904,
+      "loss": 3.3391,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000999,
+      "loss": 3.319,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099896,
+      "loss": 3.3612,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009989200000000001,
+      "loss": 3.3739,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099888,
+      "loss": 3.2076,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099884,
+      "loss": 3.138,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009988,
+      "loss": 3.1239,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099876,
+      "loss": 3.1168,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099872,
+      "loss": 2.9912,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099868,
+      "loss": 2.731,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099864,
+      "loss": 2.6893,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009986,
+      "loss": 2.7064,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009985600000000001,
+      "loss": 2.6795,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099852,
+      "loss": 2.5955,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009984800000000002,
+      "loss": 2.6869,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099844,
+      "loss": 2.5575,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009984,
+      "loss": 2.5116,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099836,
+      "loss": 2.5504,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099832,
+      "loss": 2.5918,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099828,
+      "loss": 2.4815,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099824,
+      "loss": 2.5396,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009982,
+      "loss": 2.5002,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099816,
+      "loss": 2.6185,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009981200000000001,
+      "loss": 2.5308,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099808,
+      "loss": 2.5187,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099804,
+      "loss": 2.5162,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000998,
+      "loss": 2.4158,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099796,
+      "loss": 2.412,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099792,
+      "loss": 2.363,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099788,
+      "loss": 2.3681,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099784,
+      "loss": 2.4234,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009978,
+      "loss": 2.3174,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099776,
+      "loss": 2.561,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009977200000000001,
+      "loss": 2.1715,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009976800000000001,
+      "loss": 2.2928,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099764,
+      "loss": 2.1592,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009976,
+      "loss": 2.2872,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099756,
+      "loss": 2.3656,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099752,
+      "loss": 2.3061,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099748,
+      "loss": 2.2957,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099744,
+      "loss": 2.1918,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009974,
+      "loss": 2.2743,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099736,
+      "loss": 2.3673,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099732,
+      "loss": 2.3333,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009972800000000001,
+      "loss": 2.1511,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099724,
+      "loss": 2.2897,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009972,
+      "loss": 2.1991,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099716,
+      "loss": 2.4041,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099712,
+      "loss": 2.3083,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099708,
+      "loss": 2.1394,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099704,
+      "loss": 2.2134,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000997,
+      "loss": 2.1536,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099696,
+      "loss": 2.2372,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099692,
+      "loss": 2.2853,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099688,
+      "loss": 2.1525,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099684,
+      "loss": 2.2202,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009968,
+      "loss": 2.1938,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099676,
+      "loss": 2.1177,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099672,
+      "loss": 2.1325,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099668,
+      "loss": 2.1128,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099664,
+      "loss": 2.0839,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009966,
+      "loss": 2.1249,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099656,
+      "loss": 2.1128,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099652,
+      "loss": 2.1077,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009964800000000001,
+      "loss": 2.3383,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099644,
+      "loss": 2.1341,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009964,
+      "loss": 2.1339,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099636,
+      "loss": 2.0521,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099632,
+      "loss": 2.0864,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099628,
+      "loss": 2.0886,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099624,
+      "loss": 2.0998,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009962,
+      "loss": 2.1523,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099616,
+      "loss": 1.9835,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099612,
+      "loss": 2.168,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009960799999999999,
+      "loss": 2.0925,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009960400000000001,
+      "loss": 2.0238,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000996,
+      "loss": 2.1484,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099596,
+      "loss": 2.0307,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099592,
+      "loss": 2.1454,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099588,
+      "loss": 2.0175,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099584,
+      "loss": 2.204,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009958,
+      "loss": 2.137,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099576,
+      "loss": 2.0411,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099572,
+      "loss": 2.0774,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099568,
+      "loss": 2.1539,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099564,
+      "loss": 2.0218,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009956000000000001,
+      "loss": 2.0275,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099556,
+      "loss": 2.1035,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099552,
+      "loss": 2.085,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099548,
+      "loss": 1.9915,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099544,
+      "loss": 2.0149,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009954,
+      "loss": 2.0525,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099536,
+      "loss": 1.9819,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099532,
+      "loss": 2.0125,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099528,
+      "loss": 2.1065,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099524,
+      "loss": 2.0173,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009952,
+      "loss": 2.0433,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0009951600000000001,
+      "loss": 1.8824,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099512,
+      "loss": 2.0183,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099508,
+      "loss": 2.0724,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099504,
+      "loss": 2.0274,
+      "step": 124
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.000995,
+      "loss": 2.0761,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.00099496,
+      "loss": 1.9915,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099492,
+      "loss": 2.0835,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099488,
+      "loss": 1.9794,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009948399999999999,
+      "loss": 2.0202,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009948000000000001,
+      "loss": 1.8476,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099476,
+      "loss": 2.0747,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009947200000000002,
+      "loss": 1.9736,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099468,
+      "loss": 2.0299,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099464,
+      "loss": 1.9613,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009946,
+      "loss": 1.9457,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099456,
+      "loss": 1.9559,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099452,
+      "loss": 1.951,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099448,
+      "loss": 1.9472,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099444,
+      "loss": 2.0451,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009943999999999999,
+      "loss": 2.0725,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009943600000000001,
+      "loss": 1.9357,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099432,
+      "loss": 1.993,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009942800000000002,
+      "loss": 1.9384,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099424,
+      "loss": 2.0588,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009942,
+      "loss": 1.9828,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099416,
+      "loss": 1.9088,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099412,
+      "loss": 1.7772,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099408,
+      "loss": 1.9326,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099404,
+      "loss": 1.8666,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000994,
+      "loss": 2.033,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099396,
+      "loss": 1.9549,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009939200000000001,
+      "loss": 1.9506,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099388,
+      "loss": 1.8825,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099384,
+      "loss": 1.8297,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009938,
+      "loss": 1.9007,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099376,
+      "loss": 1.9688,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099372,
+      "loss": 1.9418,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099368,
+      "loss": 1.8894,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099364,
+      "loss": 1.9187,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009936,
+      "loss": 1.928,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099356,
+      "loss": 1.9035,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099352,
+      "loss": 1.9674,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009934800000000001,
+      "loss": 1.9573,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099344,
+      "loss": 1.9853,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009934,
+      "loss": 1.9652,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099336,
+      "loss": 1.9257,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099332,
+      "loss": 2.05,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099328,
+      "loss": 1.8644,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099324,
+      "loss": 1.8845,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009932,
+      "loss": 1.9303,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099316,
+      "loss": 1.9039,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009931200000000001,
+      "loss": 1.938,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099308,
+      "loss": 1.9876,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009930400000000002,
+      "loss": 1.94,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000993,
+      "loss": 1.9371,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099296,
+      "loss": 2.0145,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099292,
+      "loss": 1.8677,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099288,
+      "loss": 1.942,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099284,
+      "loss": 1.8749,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009928,
+      "loss": 2.0096,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099276,
+      "loss": 1.9652,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099272,
+      "loss": 1.8538,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009926800000000001,
+      "loss": 1.9325,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099264,
+      "loss": 1.7823,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009926000000000002,
+      "loss": 1.9284,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099256,
+      "loss": 1.8414,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099252,
+      "loss": 1.863,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099248,
+      "loss": 1.8595,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099244,
+      "loss": 1.8983,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009924,
+      "loss": 1.9036,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099236,
+      "loss": 1.9746,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099232,
+      "loss": 1.8353,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009922800000000001,
+      "loss": 1.9268,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009922400000000001,
+      "loss": 1.911,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009922,
+      "loss": 1.8752,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099216,
+      "loss": 1.91,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099212,
+      "loss": 1.7393,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099208,
+      "loss": 1.9226,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099204,
+      "loss": 1.8594,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000992,
+      "loss": 1.9901,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099196,
+      "loss": 1.8485,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099192,
+      "loss": 1.8505,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099188,
+      "loss": 1.9161,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009918400000000001,
+      "loss": 1.9281,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009918,
+      "loss": 1.8522,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099176,
+      "loss": 1.9497,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099172,
+      "loss": 1.7709,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099168,
+      "loss": 1.8637,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099164,
+      "loss": 1.915,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009916,
+      "loss": 1.9414,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099156,
+      "loss": 1.8757,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099152,
+      "loss": 1.9752,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099148,
+      "loss": 1.9248,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009914400000000001,
+      "loss": 1.7613,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009914,
+      "loss": 1.896,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099136,
+      "loss": 1.8521,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099132,
+      "loss": 1.915,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099128,
+      "loss": 1.8605,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099124,
+      "loss": 1.8321,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009912,
+      "loss": 1.7759,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099116,
+      "loss": 1.8582,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099112,
+      "loss": 1.9002,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099108,
+      "loss": 1.9351,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099104,
+      "loss": 1.7915,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000991,
+      "loss": 1.83,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099096,
+      "loss": 1.8527,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099092,
+      "loss": 1.942,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099088,
+      "loss": 1.8221,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099084,
+      "loss": 1.8932,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009908,
+      "loss": 1.9213,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099076,
+      "loss": 1.8731,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099072,
+      "loss": 1.8509,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099068,
+      "loss": 1.7678,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009906399999999999,
+      "loss": 1.7626,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009906000000000001,
+      "loss": 1.8313,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099056,
+      "loss": 1.8645,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099052,
+      "loss": 1.8625,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099048,
+      "loss": 1.8664,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099044,
+      "loss": 1.912,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009904,
+      "loss": 1.868,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099036,
+      "loss": 1.8103,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099032,
+      "loss": 1.7844,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099028,
+      "loss": 1.751,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099024,
+      "loss": 1.7728,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009901999999999999,
+      "loss": 1.8475,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009901600000000001,
+      "loss": 1.8265,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099012,
+      "loss": 1.8343,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099008,
+      "loss": 1.746,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099004,
+      "loss": 1.8824,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00099,
+      "loss": 1.884,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098996,
+      "loss": 1.8938,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098992,
+      "loss": 1.7204,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098988,
+      "loss": 1.8862,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098984,
+      "loss": 1.7757,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009898,
+      "loss": 1.7692,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098976,
+      "loss": 1.7589,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009897200000000001,
+      "loss": 1.8249,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098968,
+      "loss": 1.8404,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098964,
+      "loss": 1.8215,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009896,
+      "loss": 1.8697,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098956,
+      "loss": 1.7824,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098952,
+      "loss": 1.7807,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098948,
+      "loss": 1.8721,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098944,
+      "loss": 1.864,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009893999999999999,
+      "loss": 1.7812,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009893600000000001,
+      "loss": 1.8458,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098932,
+      "loss": 1.683,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009892800000000002,
+      "loss": 1.7996,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098924,
+      "loss": 1.8823,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009892,
+      "loss": 1.6717,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098916,
+      "loss": 1.792,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098912,
+      "loss": 1.7125,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098908,
+      "loss": 1.7432,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098904,
+      "loss": 1.7945,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000989,
+      "loss": 1.7455,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009889599999999999,
+      "loss": 1.8314,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009889200000000001,
+      "loss": 1.8562,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098888,
+      "loss": 1.8987,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009888400000000002,
+      "loss": 1.8199,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009888,
+      "loss": 1.8358,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098876,
+      "loss": 1.8548,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098872,
+      "loss": 1.8504,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098868,
+      "loss": 1.7497,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098864,
+      "loss": 1.7261,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009886,
+      "loss": 1.7651,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098856,
+      "loss": 1.7832,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098852,
+      "loss": 1.9436,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009884800000000001,
+      "loss": 1.7851,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098844,
+      "loss": 1.8317,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009884,
+      "loss": 1.7422,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098836,
+      "loss": 1.7519,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098832,
+      "loss": 1.7455,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098828,
+      "loss": 1.8193,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098824,
+      "loss": 1.7148,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009882,
+      "loss": 1.7962,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098816,
+      "loss": 1.7387,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098812,
+      "loss": 1.8204,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098808,
+      "loss": 1.7832,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009880400000000001,
+      "loss": 1.7427,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000988,
+      "loss": 1.732,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098796,
+      "loss": 1.7103,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098792,
+      "loss": 1.7262,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098788,
+      "loss": 1.8969,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098784,
+      "loss": 1.7385,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009878,
+      "loss": 1.701,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098776,
+      "loss": 1.8269,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098772,
+      "loss": 1.6532,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009876800000000001,
+      "loss": 1.7443,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098764,
+      "loss": 1.7862,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009876000000000002,
+      "loss": 1.7974,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098756,
+      "loss": 1.7926,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098752,
+      "loss": 1.7541,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098748,
+      "loss": 1.7988,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098744,
+      "loss": 1.9052,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009874,
+      "loss": 1.8117,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098736,
+      "loss": 1.725,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098732,
+      "loss": 1.8268,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098728,
+      "loss": 1.7783,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009872400000000001,
+      "loss": 1.7836,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009872,
+      "loss": 1.7441,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009871600000000002,
+      "loss": 1.7552,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098712,
+      "loss": 1.735,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098708,
+      "loss": 1.6794,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098704,
+      "loss": 1.694,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000987,
+      "loss": 1.8065,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098696,
+      "loss": 1.7431,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098692,
+      "loss": 1.6418,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098688,
+      "loss": 1.7754,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009868400000000001,
+      "loss": 1.766,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009868000000000001,
+      "loss": 1.8102,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098676,
+      "loss": 1.6544,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098672,
+      "loss": 1.7719,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098668,
+      "loss": 1.673,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098664,
+      "loss": 1.6828,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009866,
+      "loss": 1.8121,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098656,
+      "loss": 1.8876,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098652,
+      "loss": 1.6456,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098648,
+      "loss": 1.6855,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098644,
+      "loss": 1.7584,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009864000000000001,
+      "loss": 1.6908,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098636,
+      "loss": 1.7188,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098632,
+      "loss": 1.8353,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098628,
+      "loss": 1.7576,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098624,
+      "loss": 1.6525,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009862,
+      "loss": 1.7959,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098616,
+      "loss": 1.8444,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098612,
+      "loss": 1.6246,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098608,
+      "loss": 1.7638,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098604,
+      "loss": 1.7503,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009860000000000001,
+      "loss": 1.8471,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098596,
+      "loss": 1.8188,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098592,
+      "loss": 1.7107,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098588,
+      "loss": 1.7706,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098584,
+      "loss": 1.7611,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009858,
+      "loss": 1.7723,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098576,
+      "loss": 1.8155,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098572,
+      "loss": 1.7695,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098568,
+      "loss": 1.7355,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098564,
+      "loss": 1.7122,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009856,
+      "loss": 1.8011,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098556,
+      "loss": 1.7723,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098552,
+      "loss": 1.7535,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098548,
+      "loss": 1.7696,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098544,
+      "loss": 1.8203,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009854,
+      "loss": 1.751,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098536,
+      "loss": 1.816,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098532,
+      "loss": 1.7724,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098528,
+      "loss": 1.7698,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098524,
+      "loss": 1.7209,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009851999999999999,
+      "loss": 1.69,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009851600000000001,
+      "loss": 1.7772,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098512,
+      "loss": 1.8179,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098508,
+      "loss": 1.7558,
+      "step": 373
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098504,
+      "loss": 1.8269,
+      "step": 374
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.000985,
+      "loss": 1.819,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098496,
+      "loss": 1.6876,
+      "step": 376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098492,
+      "loss": 1.7792,
+      "step": 377
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098488,
+      "loss": 1.8178,
+      "step": 378
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.00098484,
+      "loss": 1.7331,
+      "step": 379
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0009848,
+      "loss": 1.7469,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009847599999999999,
+      "loss": 1.7632,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009847200000000001,
+      "loss": 1.6696,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098468,
+      "loss": 1.7413,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098464,
+      "loss": 1.7309,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009846,
+      "loss": 1.7342,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098456,
+      "loss": 1.6844,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098452,
+      "loss": 1.7343,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098448,
+      "loss": 1.8045,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098444,
+      "loss": 1.7243,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009844,
+      "loss": 1.6841,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098436,
+      "loss": 1.7019,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098432,
+      "loss": 1.6347,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009842800000000001,
+      "loss": 1.7131,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098424,
+      "loss": 1.7077,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009842,
+      "loss": 1.6273,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098416,
+      "loss": 1.7544,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098412,
+      "loss": 1.644,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098408,
+      "loss": 1.6763,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098404,
+      "loss": 1.7344,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000984,
+      "loss": 1.6056,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009839599999999999,
+      "loss": 1.7976,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098392,
+      "loss": 1.761,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098388,
+      "loss": 1.7367,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009838400000000001,
+      "loss": 1.787,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009838,
+      "loss": 1.7622,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098376,
+      "loss": 1.6962,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098372,
+      "loss": 1.8486,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098368,
+      "loss": 1.7922,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098364,
+      "loss": 1.8008,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009836,
+      "loss": 1.7346,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098356,
+      "loss": 1.7328,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009835199999999999,
+      "loss": 1.7144,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009834800000000001,
+      "loss": 1.7184,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098344,
+      "loss": 1.7332,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009834000000000002,
+      "loss": 1.7763,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098336,
+      "loss": 1.7242,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098332,
+      "loss": 1.6967,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098328,
+      "loss": 1.7296,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098324,
+      "loss": 1.7671,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009832,
+      "loss": 1.6657,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098316,
+      "loss": 1.6949,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098312,
+      "loss": 1.6751,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009830799999999999,
+      "loss": 1.8305,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009830400000000001,
+      "loss": 1.7155,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000983,
+      "loss": 1.6582,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098296,
+      "loss": 1.7235,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098292,
+      "loss": 1.7375,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098288,
+      "loss": 1.6156,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098284,
+      "loss": 1.6923,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009828,
+      "loss": 1.7168,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098276,
+      "loss": 1.7575,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098272,
+      "loss": 1.8581,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098268,
+      "loss": 1.7231,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098264,
+      "loss": 1.6655,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009826000000000001,
+      "loss": 1.7207,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098256,
+      "loss": 1.6879,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098252,
+      "loss": 1.6553,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098248,
+      "loss": 1.6617,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098244,
+      "loss": 1.7436,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009824,
+      "loss": 1.7165,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098236,
+      "loss": 1.7024,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098232,
+      "loss": 1.6531,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098228,
+      "loss": 1.6484,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009822400000000001,
+      "loss": 1.6182,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009822,
+      "loss": 1.6498,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009821600000000002,
+      "loss": 1.8428,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098212,
+      "loss": 1.6951,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098208,
+      "loss": 1.6236,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098204,
+      "loss": 1.5958,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000982,
+      "loss": 1.6325,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098196,
+      "loss": 1.6678,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098192,
+      "loss": 1.706,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098188,
+      "loss": 1.74,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098184,
+      "loss": 1.6997,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009818000000000001,
+      "loss": 1.6552,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098176,
+      "loss": 1.6308,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009817200000000002,
+      "loss": 1.7047,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098168,
+      "loss": 1.7936,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098164,
+      "loss": 1.7015,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009816,
+      "loss": 1.6301,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098156,
+      "loss": 1.7414,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098152,
+      "loss": 1.5759,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098148,
+      "loss": 1.7785,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098144,
+      "loss": 1.6892,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009814000000000001,
+      "loss": 1.7025,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009813600000000001,
+      "loss": 1.7412,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098132,
+      "loss": 1.6815,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098128,
+      "loss": 1.8024,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098124,
+      "loss": 1.655,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009812,
+      "loss": 1.5335,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098116,
+      "loss": 1.5589,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098112,
+      "loss": 1.811,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098108,
+      "loss": 1.6784,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098104,
+      "loss": 1.7283,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000981,
+      "loss": 1.6789,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009809600000000001,
+      "loss": 1.6897,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009809200000000001,
+      "loss": 1.7384,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098088,
+      "loss": 1.7103,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098084,
+      "loss": 1.7065,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009808,
+      "loss": 1.6673,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098076,
+      "loss": 1.635,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098072,
+      "loss": 1.7449,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098068,
+      "loss": 1.6638,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098064,
+      "loss": 1.7055,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009806,
+      "loss": 1.7875,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009805600000000001,
+      "loss": 1.7642,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098052,
+      "loss": 1.7175,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098048,
+      "loss": 1.6093,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098044,
+      "loss": 1.7152,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009804,
+      "loss": 1.7898,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098036,
+      "loss": 1.7949,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098032,
+      "loss": 1.6678,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098028,
+      "loss": 1.6013,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098024,
+      "loss": 1.7795,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009802,
+      "loss": 1.7655,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098016,
+      "loss": 1.6958,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098012,
+      "loss": 1.7183,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098008,
+      "loss": 1.645,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098004,
+      "loss": 1.5412,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00098,
+      "loss": 1.6183,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097996,
+      "loss": 1.7159,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097992,
+      "loss": 1.7736,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097988,
+      "loss": 1.6251,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097984,
+      "loss": 1.5801,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009798,
+      "loss": 1.6564,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097976,
+      "loss": 1.5818,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009797200000000001,
+      "loss": 1.6519,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097968,
+      "loss": 1.7008,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097964,
+      "loss": 1.6177,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009796,
+      "loss": 1.6258,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097956,
+      "loss": 1.6282,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097952,
+      "loss": 1.6557,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097948,
+      "loss": 1.6377,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097944,
+      "loss": 1.6249,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009794,
+      "loss": 1.7492,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097936,
+      "loss": 1.7184,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009793199999999999,
+      "loss": 1.7108,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009792800000000001,
+      "loss": 1.6718,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097924,
+      "loss": 1.6823,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009792,
+      "loss": 1.6749,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097916,
+      "loss": 1.747,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097912,
+      "loss": 1.6413,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097908,
+      "loss": 1.656,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097904,
+      "loss": 1.7008,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000979,
+      "loss": 1.638,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097896,
+      "loss": 1.6759,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097892,
+      "loss": 1.7061,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097888,
+      "loss": 1.6898,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009788400000000001,
+      "loss": 1.5889,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009788,
+      "loss": 1.6206,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097876,
+      "loss": 1.6586,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097872,
+      "loss": 1.6765,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097868,
+      "loss": 1.7636,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097864,
+      "loss": 1.6999,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009786,
+      "loss": 1.7617,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097856,
+      "loss": 1.6234,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009785199999999999,
+      "loss": 1.7233,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097848,
+      "loss": 1.6569,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097844,
+      "loss": 1.7188,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009784000000000001,
+      "loss": 1.6473,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097836,
+      "loss": 1.7264,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097832,
+      "loss": 1.7152,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097828,
+      "loss": 1.6669,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097824,
+      "loss": 1.6049,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009782,
+      "loss": 1.6875,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097816,
+      "loss": 1.7777,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097812,
+      "loss": 1.7224,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009780799999999999,
+      "loss": 1.644,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009780400000000001,
+      "loss": 1.648,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000978,
+      "loss": 1.699,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009779600000000002,
+      "loss": 1.6407,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097792,
+      "loss": 1.68,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097788,
+      "loss": 1.6209,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097784,
+      "loss": 1.6523,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009778,
+      "loss": 1.7113,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097776,
+      "loss": 1.5899,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097772,
+      "loss": 1.7023,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097768,
+      "loss": 1.678,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009776399999999999,
+      "loss": 1.7008,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009776000000000001,
+      "loss": 1.7676,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097756,
+      "loss": 1.6234,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097752,
+      "loss": 1.6652,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097748,
+      "loss": 1.7125,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097744,
+      "loss": 1.6402,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009774,
+      "loss": 1.6616,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097736,
+      "loss": 1.6378,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097732,
+      "loss": 1.5998,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097728,
+      "loss": 1.7217,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097724,
+      "loss": 1.7031,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009772,
+      "loss": 1.5978,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009771600000000001,
+      "loss": 1.7293,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097712,
+      "loss": 1.6405,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097708,
+      "loss": 1.6801,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097704,
+      "loss": 1.6024,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000977,
+      "loss": 1.5837,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097696,
+      "loss": 1.6725,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097692,
+      "loss": 1.6908,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097688,
+      "loss": 1.7802,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097684,
+      "loss": 1.6643,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009768,
+      "loss": 1.6921,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097676,
+      "loss": 1.5575,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009767200000000001,
+      "loss": 1.6486,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097668,
+      "loss": 1.6948,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097664,
+      "loss": 1.5602,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009766,
+      "loss": 1.6373,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097656,
+      "loss": 1.5043,
+      "step": 586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097652,
+      "loss": 1.6767,
+      "step": 587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009764800000000001,
+      "loss": 1.6199,
+      "step": 588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097644,
+      "loss": 1.599,
+      "step": 589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009764000000000001,
+      "loss": 1.7423,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097636,
+      "loss": 1.6379,
+      "step": 591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097632,
+      "loss": 1.7446,
+      "step": 592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097628,
+      "loss": 1.6995,
+      "step": 593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097624,
+      "loss": 1.6459,
+      "step": 594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009762,
+      "loss": 1.6468,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097616,
+      "loss": 1.7061,
+      "step": 596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097612,
+      "loss": 1.7105,
+      "step": 597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009760799999999999,
+      "loss": 1.6424,
+      "step": 598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009760400000000001,
+      "loss": 1.7182,
+      "step": 599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000976,
+      "loss": 1.6495,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009759600000000001,
+      "loss": 1.5984,
+      "step": 601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097592,
+      "loss": 1.594,
+      "step": 602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097588,
+      "loss": 1.5655,
+      "step": 603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009758400000000001,
+      "loss": 1.7122,
+      "step": 604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009758,
+      "loss": 1.6149,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097576,
+      "loss": 1.6176,
+      "step": 606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097572,
+      "loss": 1.7522,
+      "step": 607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097568,
+      "loss": 1.6586,
+      "step": 608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097564,
+      "loss": 1.5094,
+      "step": 609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009756000000000001,
+      "loss": 1.6177,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097556,
+      "loss": 1.6442,
+      "step": 611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009755200000000001,
+      "loss": 1.6602,
+      "step": 612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097548,
+      "loss": 1.6542,
+      "step": 613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009754399999999999,
+      "loss": 1.735,
+      "step": 614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009754000000000001,
+      "loss": 1.729,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097536,
+      "loss": 1.6787,
+      "step": 616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097532,
+      "loss": 1.6901,
+      "step": 617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097528,
+      "loss": 1.665,
+      "step": 618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097524,
+      "loss": 1.5688,
+      "step": 619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009752,
+      "loss": 1.6906,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009751600000000001,
+      "loss": 1.6599,
+      "step": 621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097512,
+      "loss": 1.6632,
+      "step": 622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009750799999999999,
+      "loss": 1.6762,
+      "step": 623
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097504,
+      "loss": 1.6518,
+      "step": 624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.000975,
+      "loss": 1.5746,
+      "step": 625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009749600000000001,
+      "loss": 1.6234,
+      "step": 626
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097492,
+      "loss": 1.6091,
+      "step": 627
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097488,
+      "loss": 1.7598,
+      "step": 628
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097484,
+      "loss": 1.5814,
+      "step": 629
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009748000000000001,
+      "loss": 1.6594,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097476,
+      "loss": 1.5693,
+      "step": 631
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0009747200000000001,
+      "loss": 1.5506,
+      "step": 632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00097468,
+      "loss": 1.5907,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009746399999999999,
+      "loss": 1.6157,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009746,
+      "loss": 1.72,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097456,
+      "loss": 1.6557,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009745200000000001,
+      "loss": 1.5659,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097448,
+      "loss": 1.5823,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097444,
+      "loss": 1.5987,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009744,
+      "loss": 1.7122,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009743600000000001,
+      "loss": 1.6363,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097432,
+      "loss": 1.6447,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097428,
+      "loss": 1.6404,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097424,
+      "loss": 1.6064,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009741999999999999,
+      "loss": 1.5922,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009741600000000001,
+      "loss": 1.6497,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097412,
+      "loss": 1.6853,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097408,
+      "loss": 1.5678,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097404,
+      "loss": 1.6228,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000974,
+      "loss": 1.6823,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009739600000000001,
+      "loss": 1.5582,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009739200000000001,
+      "loss": 1.5961,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097388,
+      "loss": 1.5777,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097384,
+      "loss": 1.7106,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009738,
+      "loss": 1.6143,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097376,
+      "loss": 1.6236,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009737200000000001,
+      "loss": 1.6005,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097368,
+      "loss": 1.5814,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097364,
+      "loss": 1.6462,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009736,
+      "loss": 1.6788,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097356,
+      "loss": 1.6309,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009735200000000001,
+      "loss": 1.6644,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097348,
+      "loss": 1.674,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097344,
+      "loss": 1.7638,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009734,
+      "loss": 1.6257,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097336,
+      "loss": 1.745,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097332,
+      "loss": 1.6082,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009732800000000001,
+      "loss": 1.6178,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097324,
+      "loss": 1.5228,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009732,
+      "loss": 1.6507,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097316,
+      "loss": 1.6938,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097312,
+      "loss": 1.6937,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097308,
+      "loss": 1.6408,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097304,
+      "loss": 1.6924,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000973,
+      "loss": 1.7378,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097296,
+      "loss": 1.6493,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009729200000000001,
+      "loss": 1.6294,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097288,
+      "loss": 1.6165,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009728400000000001,
+      "loss": 1.6694,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009728,
+      "loss": 1.606,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097276,
+      "loss": 1.7249,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097272,
+      "loss": 1.5379,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097268,
+      "loss": 1.5954,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097264,
+      "loss": 1.5838,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009726,
+      "loss": 1.6711,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097256,
+      "loss": 1.6279,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097252,
+      "loss": 1.6293,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009724800000000001,
+      "loss": 1.5986,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097244,
+      "loss": 1.6783,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009724000000000001,
+      "loss": 1.5407,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097236,
+      "loss": 1.6276,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009723199999999999,
+      "loss": 1.711,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009722800000000001,
+      "loss": 1.5057,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097224,
+      "loss": 1.6425,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009722,
+      "loss": 1.6542,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097216,
+      "loss": 1.6744,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097212,
+      "loss": 1.6917,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009720800000000001,
+      "loss": 1.6531,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009720400000000001,
+      "loss": 1.5782,
+      "step": 699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000972,
+      "loss": 1.6456,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009719600000000001,
+      "loss": 1.6371,
+      "step": 701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097192,
+      "loss": 1.638,
+      "step": 702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009718799999999999,
+      "loss": 1.6823,
+      "step": 703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009718400000000001,
+      "loss": 1.4966,
+      "step": 704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009718,
+      "loss": 1.6872,
+      "step": 705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097176,
+      "loss": 1.5801,
+      "step": 706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097172,
+      "loss": 1.5632,
+      "step": 707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097168,
+      "loss": 1.6042,
+      "step": 708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097164,
+      "loss": 1.5584,
+      "step": 709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009716000000000001,
+      "loss": 1.6592,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097156,
+      "loss": 1.6048,
+      "step": 711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097152,
+      "loss": 1.5822,
+      "step": 712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097148,
+      "loss": 1.5302,
+      "step": 713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097144,
+      "loss": 1.615,
+      "step": 714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009714000000000001,
+      "loss": 1.6484,
+      "step": 715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097136,
+      "loss": 1.646,
+      "step": 716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097132,
+      "loss": 1.6007,
+      "step": 717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097128,
+      "loss": 1.5183,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009712400000000001,
+      "loss": 1.6283,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009712,
+      "loss": 1.5339,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097116,
+      "loss": 1.5514,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097112,
+      "loss": 1.6831,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097108,
+      "loss": 1.5621,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097104,
+      "loss": 1.5843,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000971,
+      "loss": 1.5687,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009709600000000001,
+      "loss": 1.5934,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097092,
+      "loss": 1.6653,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097088,
+      "loss": 1.5562,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097084,
+      "loss": 1.5783,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009708000000000001,
+      "loss": 1.6,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097076,
+      "loss": 1.6231,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097072,
+      "loss": 1.5695,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097068,
+      "loss": 1.6248,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009706399999999999,
+      "loss": 1.6454,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009706000000000001,
+      "loss": 1.6497,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097056,
+      "loss": 1.5898,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009705200000000001,
+      "loss": 1.6337,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097048,
+      "loss": 1.5933,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097044,
+      "loss": 1.5426,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009704000000000001,
+      "loss": 1.6195,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097036,
+      "loss": 1.6773,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097032,
+      "loss": 1.5766,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097028,
+      "loss": 1.6111,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097024,
+      "loss": 1.5909,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009702,
+      "loss": 1.6918,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009701600000000001,
+      "loss": 1.6462,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097012,
+      "loss": 1.6215,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009700800000000001,
+      "loss": 1.6819,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00097004,
+      "loss": 1.6502,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009699999999999999,
+      "loss": 1.665,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009699600000000001,
+      "loss": 1.636,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096992,
+      "loss": 1.6504,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096988,
+      "loss": 1.5686,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096984,
+      "loss": 1.5673,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009698,
+      "loss": 1.5166,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096976,
+      "loss": 1.5816,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009697200000000001,
+      "loss": 1.5579,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096968,
+      "loss": 1.6163,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096964,
+      "loss": 1.6277,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009696,
+      "loss": 1.6458,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096956,
+      "loss": 1.6385,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009695200000000001,
+      "loss": 1.5273,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096948,
+      "loss": 1.5314,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096944,
+      "loss": 1.5955,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009694,
+      "loss": 1.6774,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009693600000000001,
+      "loss": 1.5437,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096932,
+      "loss": 1.6309,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009692800000000001,
+      "loss": 1.6485,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096924,
+      "loss": 1.633,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009691999999999999,
+      "loss": 1.5608,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096916,
+      "loss": 1.4974,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096912,
+      "loss": 1.5017,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009690800000000001,
+      "loss": 1.5736,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096904,
+      "loss": 1.6503,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000969,
+      "loss": 1.6027,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096896,
+      "loss": 1.4814,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009689200000000001,
+      "loss": 1.5389,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096888,
+      "loss": 1.501,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096884,
+      "loss": 1.6395,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009688,
+      "loss": 1.5737,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009687599999999999,
+      "loss": 1.4467,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009687200000000001,
+      "loss": 1.5392,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096868,
+      "loss": 1.576,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096864,
+      "loss": 1.653,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009686,
+      "loss": 1.613,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096856,
+      "loss": 1.5606,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009685200000000001,
+      "loss": 1.5973,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009684800000000001,
+      "loss": 1.5585,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096844,
+      "loss": 1.6452,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009684,
+      "loss": 1.657,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096836,
+      "loss": 1.5909,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009683199999999999,
+      "loss": 1.6693,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009682800000000001,
+      "loss": 1.5661,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096824,
+      "loss": 1.6092,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009682,
+      "loss": 1.6238,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096816,
+      "loss": 1.5533,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096812,
+      "loss": 1.6824,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009680800000000001,
+      "loss": 1.6502,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096804,
+      "loss": 1.6948,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000968,
+      "loss": 1.6346,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096796,
+      "loss": 1.6982,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096792,
+      "loss": 1.543,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096788,
+      "loss": 1.6236,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009678400000000001,
+      "loss": 1.5657,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009678,
+      "loss": 1.573,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096776,
+      "loss": 1.5974,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096772,
+      "loss": 1.6036,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096768,
+      "loss": 1.508,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096764,
+      "loss": 1.6026,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009676,
+      "loss": 1.6511,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096756,
+      "loss": 1.69,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096752,
+      "loss": 1.6271,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009674800000000001,
+      "loss": 1.6856,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096744,
+      "loss": 1.5805,
+      "step": 814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009674000000000001,
+      "loss": 1.6424,
+      "step": 815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096736,
+      "loss": 1.5865,
+      "step": 816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096732,
+      "loss": 1.5963,
+      "step": 817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096728,
+      "loss": 1.5373,
+      "step": 818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096724,
+      "loss": 1.4875,
+      "step": 819
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009672,
+      "loss": 1.689,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096716,
+      "loss": 1.6447,
+      "step": 821
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096712,
+      "loss": 1.6136,
+      "step": 822
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096708,
+      "loss": 1.5218,
+      "step": 823
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009670400000000001,
+      "loss": 1.5323,
+      "step": 824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000967,
+      "loss": 1.6226,
+      "step": 825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009669600000000001,
+      "loss": 1.6898,
+      "step": 826
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096692,
+      "loss": 1.567,
+      "step": 827
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009668799999999999,
+      "loss": 1.6052,
+      "step": 828
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009668400000000001,
+      "loss": 1.5005,
+      "step": 829
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009668,
+      "loss": 1.6615,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096676,
+      "loss": 1.5036,
+      "step": 831
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096672,
+      "loss": 1.5069,
+      "step": 832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096668,
+      "loss": 1.5825,
+      "step": 833
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009666400000000001,
+      "loss": 1.5873,
+      "step": 834
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009666000000000001,
+      "loss": 1.5682,
+      "step": 835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096656,
+      "loss": 1.6468,
+      "step": 836
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009665200000000001,
+      "loss": 1.5826,
+      "step": 837
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096648,
+      "loss": 1.6523,
+      "step": 838
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009664399999999999,
+      "loss": 1.539,
+      "step": 839
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009664000000000001,
+      "loss": 1.5654,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096636,
+      "loss": 1.5931,
+      "step": 841
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096632,
+      "loss": 1.6461,
+      "step": 842
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096628,
+      "loss": 1.5837,
+      "step": 843
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096624,
+      "loss": 1.6573,
+      "step": 844
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009662,
+      "loss": 1.5526,
+      "step": 845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009661600000000001,
+      "loss": 1.625,
+      "step": 846
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096612,
+      "loss": 1.522,
+      "step": 847
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096608,
+      "loss": 1.5099,
+      "step": 848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096604,
+      "loss": 1.5851,
+      "step": 849
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000966,
+      "loss": 1.5683,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009659600000000001,
+      "loss": 1.5757,
+      "step": 851
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096592,
+      "loss": 1.6099,
+      "step": 852
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096588,
+      "loss": 1.5108,
+      "step": 853
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096584,
+      "loss": 1.5513,
+      "step": 854
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009658000000000001,
+      "loss": 1.5233,
+      "step": 855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096576,
+      "loss": 1.6504,
+      "step": 856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096572,
+      "loss": 1.5611,
+      "step": 857
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096568,
+      "loss": 1.5139,
+      "step": 858
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096564,
+      "loss": 1.5657,
+      "step": 859
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009656,
+      "loss": 1.5537,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096556,
+      "loss": 1.6217,
+      "step": 861
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009655200000000001,
+      "loss": 1.5746,
+      "step": 862
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096548,
+      "loss": 1.6474,
+      "step": 863
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096544,
+      "loss": 1.7091,
+      "step": 864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009654,
+      "loss": 1.5678,
+      "step": 865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009653600000000001,
+      "loss": 1.5993,
+      "step": 866
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096532,
+      "loss": 1.5907,
+      "step": 867
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096528,
+      "loss": 1.6334,
+      "step": 868
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096524,
+      "loss": 1.5255,
+      "step": 869
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009651999999999999,
+      "loss": 1.5192,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009651600000000001,
+      "loss": 1.6153,
+      "step": 871
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096512,
+      "loss": 1.7463,
+      "step": 872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009650800000000001,
+      "loss": 1.661,
+      "step": 873
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096504,
+      "loss": 1.6048,
+      "step": 874
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000965,
+      "loss": 1.5723,
+      "step": 875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009649600000000001,
+      "loss": 1.5959,
+      "step": 876
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096492,
+      "loss": 1.6736,
+      "step": 877
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096488,
+      "loss": 1.6751,
+      "step": 878
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096484,
+      "loss": 1.4925,
+      "step": 879
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009648,
+      "loss": 1.584,
+      "step": 880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009647599999999999,
+      "loss": 1.5126,
+      "step": 881
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009647200000000001,
+      "loss": 1.6067,
+      "step": 882
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00096468,
+      "loss": 1.528,
+      "step": 883
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009646400000000001,
+      "loss": 1.5042,
+      "step": 884
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009646,
+      "loss": 1.5649,
+      "step": 885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0009645599999999999,
+      "loss": 1.4919,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009645200000000001,
+      "loss": 1.6297,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096448,
+      "loss": 1.5516,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096444,
+      "loss": 1.5505,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009644,
+      "loss": 1.5665,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096436,
+      "loss": 1.6069,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096432,
+      "loss": 1.6066,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009642800000000001,
+      "loss": 1.5876,
+      "step": 893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096424,
+      "loss": 1.5578,
+      "step": 894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009642,
+      "loss": 1.6901,
+      "step": 895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096416,
+      "loss": 1.5969,
+      "step": 896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096412,
+      "loss": 1.6628,
+      "step": 897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009640800000000001,
+      "loss": 1.5531,
+      "step": 898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096404,
+      "loss": 1.5121,
+      "step": 899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000964,
+      "loss": 1.5819,
+      "step": 900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096396,
+      "loss": 1.5578,
+      "step": 901
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009639200000000001,
+      "loss": 1.6058,
+      "step": 902
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096388,
+      "loss": 1.5545,
+      "step": 903
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009638400000000001,
+      "loss": 1.5312,
+      "step": 904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009638,
+      "loss": 1.5989,
+      "step": 905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009637599999999999,
+      "loss": 1.6469,
+      "step": 906
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096372,
+      "loss": 1.6416,
+      "step": 907
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096368,
+      "loss": 1.5109,
+      "step": 908
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009636400000000001,
+      "loss": 1.5884,
+      "step": 909
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009636,
+      "loss": 1.5809,
+      "step": 910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096356,
+      "loss": 1.4752,
+      "step": 911
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096352,
+      "loss": 1.6014,
+      "step": 912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009634800000000001,
+      "loss": 1.5321,
+      "step": 913
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096344,
+      "loss": 1.5991,
+      "step": 914
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009634000000000001,
+      "loss": 1.624,
+      "step": 915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096336,
+      "loss": 1.5084,
+      "step": 916
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009633199999999999,
+      "loss": 1.6731,
+      "step": 917
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009632800000000001,
+      "loss": 1.552,
+      "step": 918
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096324,
+      "loss": 1.587,
+      "step": 919
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009632,
+      "loss": 1.4795,
+      "step": 920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096316,
+      "loss": 1.5259,
+      "step": 921
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096312,
+      "loss": 1.6974,
+      "step": 922
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009630800000000001,
+      "loss": 1.6566,
+      "step": 923
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009630400000000001,
+      "loss": 1.6069,
+      "step": 924
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000963,
+      "loss": 1.4652,
+      "step": 925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096296,
+      "loss": 1.4688,
+      "step": 926
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096292,
+      "loss": 1.5037,
+      "step": 927
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009628799999999999,
+      "loss": 1.5906,
+      "step": 928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009628400000000001,
+      "loss": 1.6544,
+      "step": 929
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009628,
+      "loss": 1.5093,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096276,
+      "loss": 1.6223,
+      "step": 931
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096272,
+      "loss": 1.5487,
+      "step": 932
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096268,
+      "loss": 1.6319,
+      "step": 933
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009626400000000001,
+      "loss": 1.5618,
+      "step": 934
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009626,
+      "loss": 1.5294,
+      "step": 935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096256,
+      "loss": 1.6889,
+      "step": 936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096252,
+      "loss": 1.57,
+      "step": 937
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096248,
+      "loss": 1.6513,
+      "step": 938
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096244,
+      "loss": 1.5718,
+      "step": 939
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009624000000000001,
+      "loss": 1.6018,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096236,
+      "loss": 1.3905,
+      "step": 941
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096232,
+      "loss": 1.6062,
+      "step": 942
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096228,
+      "loss": 1.4965,
+      "step": 943
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009622400000000001,
+      "loss": 1.67,
+      "step": 944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009622000000000001,
+      "loss": 1.5506,
+      "step": 945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096216,
+      "loss": 1.528,
+      "step": 946
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096212,
+      "loss": 1.6317,
+      "step": 947
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096208,
+      "loss": 1.6388,
+      "step": 948
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096204,
+      "loss": 1.623,
+      "step": 949
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000962,
+      "loss": 1.606,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009619600000000001,
+      "loss": 1.527,
+      "step": 951
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096192,
+      "loss": 1.5162,
+      "step": 952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096188,
+      "loss": 1.7119,
+      "step": 953
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096184,
+      "loss": 1.5301,
+      "step": 954
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009618,
+      "loss": 1.4859,
+      "step": 955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096176,
+      "loss": 1.5095,
+      "step": 956
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096172,
+      "loss": 1.6099,
+      "step": 957
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096168,
+      "loss": 1.5535,
+      "step": 958
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096164,
+      "loss": 1.5587,
+      "step": 959
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009616000000000001,
+      "loss": 1.5555,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096156,
+      "loss": 1.5194,
+      "step": 961
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009615200000000001,
+      "loss": 1.5637,
+      "step": 962
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096148,
+      "loss": 1.554,
+      "step": 963
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009614399999999999,
+      "loss": 1.4749,
+      "step": 964
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009614000000000001,
+      "loss": 1.6291,
+      "step": 965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096136,
+      "loss": 1.6509,
+      "step": 966
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096132,
+      "loss": 1.5714,
+      "step": 967
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096128,
+      "loss": 1.5506,
+      "step": 968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096124,
+      "loss": 1.4494,
+      "step": 969
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009612,
+      "loss": 1.5774,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009611600000000001,
+      "loss": 1.5042,
+      "step": 971
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096112,
+      "loss": 1.5229,
+      "step": 972
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009610800000000001,
+      "loss": 1.5401,
+      "step": 973
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096104,
+      "loss": 1.5862,
+      "step": 974
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009609999999999999,
+      "loss": 1.6234,
+      "step": 975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009609600000000001,
+      "loss": 1.5579,
+      "step": 976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096092,
+      "loss": 1.5531,
+      "step": 977
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096088,
+      "loss": 1.5241,
+      "step": 978
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096084,
+      "loss": 1.5507,
+      "step": 979
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009608,
+      "loss": 1.5398,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096076,
+      "loss": 1.6652,
+      "step": 981
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009607200000000001,
+      "loss": 1.5258,
+      "step": 982
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096068,
+      "loss": 1.5265,
+      "step": 983
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096064,
+      "loss": 1.5812,
+      "step": 984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009606,
+      "loss": 1.4891,
+      "step": 985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096056,
+      "loss": 1.6061,
+      "step": 986
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009605200000000001,
+      "loss": 1.7032,
+      "step": 987
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096048,
+      "loss": 1.5203,
+      "step": 988
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096044,
+      "loss": 1.5778,
+      "step": 989
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009604,
+      "loss": 1.4938,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009603600000000001,
+      "loss": 1.5031,
+      "step": 991
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096032,
+      "loss": 1.5154,
+      "step": 992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009602800000000001,
+      "loss": 1.5816,
+      "step": 993
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096024,
+      "loss": 1.5786,
+      "step": 994
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009602,
+      "loss": 1.554,
+      "step": 995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096016,
+      "loss": 1.4629,
+      "step": 996
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096012,
+      "loss": 1.5551,
+      "step": 997
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009600800000000001,
+      "loss": 1.6078,
+      "step": 998
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096004,
+      "loss": 1.4771,
+      "step": 999
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00096,
+      "loss": 1.6341,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095996,
+      "loss": 1.5058,
+      "step": 1001
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009599200000000001,
+      "loss": 1.5647,
+      "step": 1002
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095988,
+      "loss": 1.4994,
+      "step": 1003
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095984,
+      "loss": 1.4857,
+      "step": 1004
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009598,
+      "loss": 1.6171,
+      "step": 1005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009597599999999999,
+      "loss": 1.4861,
+      "step": 1006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009597200000000001,
+      "loss": 1.5892,
+      "step": 1007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095968,
+      "loss": 1.5243,
+      "step": 1008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009596400000000001,
+      "loss": 1.5914,
+      "step": 1009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009596,
+      "loss": 1.5042,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095956,
+      "loss": 1.5516,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009595200000000001,
+      "loss": 1.5014,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095948,
+      "loss": 1.5515,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095944,
+      "loss": 1.6454,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009594,
+      "loss": 1.5604,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095936,
+      "loss": 1.5833,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009593199999999999,
+      "loss": 1.4538,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009592800000000001,
+      "loss": 1.5718,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095924,
+      "loss": 1.6315,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009592000000000001,
+      "loss": 1.5156,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095916,
+      "loss": 1.6062,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095912,
+      "loss": 1.5672,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009590800000000001,
+      "loss": 1.5835,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095904,
+      "loss": 1.5655,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000959,
+      "loss": 1.5342,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095896,
+      "loss": 1.5405,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095892,
+      "loss": 1.5386,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095888,
+      "loss": 1.5244,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009588400000000001,
+      "loss": 1.5624,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009588,
+      "loss": 1.5136,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095876,
+      "loss": 1.6599,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095872,
+      "loss": 1.5811,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095868,
+      "loss": 1.4536,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009586400000000001,
+      "loss": 1.5543,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009586,
+      "loss": 1.5844,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095856,
+      "loss": 1.455,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095852,
+      "loss": 1.5109,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095848,
+      "loss": 1.611,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095844,
+      "loss": 1.5971,
+      "step": 1039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009584000000000001,
+      "loss": 1.5074,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095836,
+      "loss": 1.4857,
+      "step": 1041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009583199999999999,
+      "loss": 1.4855,
+      "step": 1042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095828,
+      "loss": 1.4811,
+      "step": 1043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095824,
+      "loss": 1.5364,
+      "step": 1044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009582000000000001,
+      "loss": 1.4994,
+      "step": 1045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095816,
+      "loss": 1.5367,
+      "step": 1046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095812,
+      "loss": 1.5972,
+      "step": 1047
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095808,
+      "loss": 1.6086,
+      "step": 1048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009580400000000001,
+      "loss": 1.6772,
+      "step": 1049
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000958,
+      "loss": 1.5355,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009579600000000001,
+      "loss": 1.5049,
+      "step": 1051
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095792,
+      "loss": 1.5004,
+      "step": 1052
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009578799999999999,
+      "loss": 1.5254,
+      "step": 1053
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009578400000000001,
+      "loss": 1.595,
+      "step": 1054
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009578,
+      "loss": 1.5472,
+      "step": 1055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095776,
+      "loss": 1.5921,
+      "step": 1056
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095772,
+      "loss": 1.5007,
+      "step": 1057
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095768,
+      "loss": 1.5447,
+      "step": 1058
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095764,
+      "loss": 1.5468,
+      "step": 1059
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009576000000000001,
+      "loss": 1.5914,
+      "step": 1060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095756,
+      "loss": 1.5271,
+      "step": 1061
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095752,
+      "loss": 1.5621,
+      "step": 1062
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095748,
+      "loss": 1.4689,
+      "step": 1063
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009574399999999999,
+      "loss": 1.5458,
+      "step": 1064
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009574000000000001,
+      "loss": 1.6401,
+      "step": 1065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095736,
+      "loss": 1.6332,
+      "step": 1066
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095732,
+      "loss": 1.5472,
+      "step": 1067
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095728,
+      "loss": 1.5106,
+      "step": 1068
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095724,
+      "loss": 1.5031,
+      "step": 1069
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009572000000000001,
+      "loss": 1.5055,
+      "step": 1070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095716,
+      "loss": 1.515,
+      "step": 1071
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095712,
+      "loss": 1.6089,
+      "step": 1072
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095708,
+      "loss": 1.4891,
+      "step": 1073
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095704,
+      "loss": 1.5408,
+      "step": 1074
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000957,
+      "loss": 1.5904,
+      "step": 1075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009569600000000001,
+      "loss": 1.5685,
+      "step": 1076
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095692,
+      "loss": 1.5231,
+      "step": 1077
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095688,
+      "loss": 1.516,
+      "step": 1078
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095684,
+      "loss": 1.5133,
+      "step": 1079
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009568000000000001,
+      "loss": 1.5296,
+      "step": 1080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009567600000000001,
+      "loss": 1.5802,
+      "step": 1081
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095672,
+      "loss": 1.4756,
+      "step": 1082
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095668,
+      "loss": 1.5699,
+      "step": 1083
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095664,
+      "loss": 1.6045,
+      "step": 1084
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009566,
+      "loss": 1.5347,
+      "step": 1085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095656,
+      "loss": 1.6114,
+      "step": 1086
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009565200000000001,
+      "loss": 1.5277,
+      "step": 1087
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095648,
+      "loss": 1.5334,
+      "step": 1088
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095644,
+      "loss": 1.5123,
+      "step": 1089
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009564,
+      "loss": 1.5974,
+      "step": 1090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095636,
+      "loss": 1.5013,
+      "step": 1091
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095632,
+      "loss": 1.4566,
+      "step": 1092
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095628,
+      "loss": 1.5138,
+      "step": 1093
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095624,
+      "loss": 1.5463,
+      "step": 1094
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009562,
+      "loss": 1.5128,
+      "step": 1095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009561600000000001,
+      "loss": 1.5712,
+      "step": 1096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095612,
+      "loss": 1.5957,
+      "step": 1097
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009560800000000001,
+      "loss": 1.5266,
+      "step": 1098
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095604,
+      "loss": 1.6436,
+      "step": 1099
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009559999999999999,
+      "loss": 1.5596,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009559600000000001,
+      "loss": 1.4791,
+      "step": 1101
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095592,
+      "loss": 1.5779,
+      "step": 1102
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095588,
+      "loss": 1.6566,
+      "step": 1103
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095584,
+      "loss": 1.578,
+      "step": 1104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009558,
+      "loss": 1.5753,
+      "step": 1105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095576,
+      "loss": 1.5664,
+      "step": 1106
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009557200000000001,
+      "loss": 1.5993,
+      "step": 1107
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095568,
+      "loss": 1.5571,
+      "step": 1108
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009556400000000001,
+      "loss": 1.4981,
+      "step": 1109
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009556,
+      "loss": 1.5093,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009555599999999999,
+      "loss": 1.4295,
+      "step": 1111
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009555200000000001,
+      "loss": 1.517,
+      "step": 1112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095548,
+      "loss": 1.5485,
+      "step": 1113
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095544,
+      "loss": 1.4787,
+      "step": 1114
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009554,
+      "loss": 1.6507,
+      "step": 1115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095536,
+      "loss": 1.593,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095532,
+      "loss": 1.5231,
+      "step": 1117
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009552800000000001,
+      "loss": 1.5019,
+      "step": 1118
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095524,
+      "loss": 1.5113,
+      "step": 1119
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009552,
+      "loss": 1.5171,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095516,
+      "loss": 1.5838,
+      "step": 1121
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095512,
+      "loss": 1.631,
+      "step": 1122
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009550800000000001,
+      "loss": 1.544,
+      "step": 1123
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095504,
+      "loss": 1.4872,
+      "step": 1124
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000955,
+      "loss": 1.5721,
+      "step": 1125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095496,
+      "loss": 1.5617,
+      "step": 1126
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095492,
+      "loss": 1.5244,
+      "step": 1127
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095488,
+      "loss": 1.5201,
+      "step": 1128
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009548400000000001,
+      "loss": 1.5396,
+      "step": 1129
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009548,
+      "loss": 1.48,
+      "step": 1130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095476,
+      "loss": 1.5349,
+      "step": 1131
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095472,
+      "loss": 1.4931,
+      "step": 1132
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095468,
+      "loss": 1.6049,
+      "step": 1133
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009546400000000001,
+      "loss": 1.4964,
+      "step": 1134
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009546,
+      "loss": 1.4512,
+      "step": 1135
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095456,
+      "loss": 1.5789,
+      "step": 1136
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095452,
+      "loss": 1.5677,
+      "step": 1137
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009544800000000001,
+      "loss": 1.5708,
+      "step": 1138
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00095444,
+      "loss": 1.4271,
+      "step": 1139
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0009544,
+      "loss": 1.4967,
+      "step": 1140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095436,
+      "loss": 1.4567,
+      "step": 1141
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009543199999999999,
+      "loss": 1.4348,
+      "step": 1142
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009542800000000001,
+      "loss": 1.4827,
+      "step": 1143
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095424,
+      "loss": 1.5512,
+      "step": 1144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009542000000000001,
+      "loss": 1.5569,
+      "step": 1145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095416,
+      "loss": 1.586,
+      "step": 1146
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095412,
+      "loss": 1.5167,
+      "step": 1147
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095408,
+      "loss": 1.585,
+      "step": 1148
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095404,
+      "loss": 1.4576,
+      "step": 1149
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000954,
+      "loss": 1.5795,
+      "step": 1150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095396,
+      "loss": 1.5241,
+      "step": 1151
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095392,
+      "loss": 1.4565,
+      "step": 1152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009538799999999999,
+      "loss": 1.6478,
+      "step": 1153
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009538400000000001,
+      "loss": 1.5141,
+      "step": 1154
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009538,
+      "loss": 1.5734,
+      "step": 1155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009537600000000001,
+      "loss": 1.6081,
+      "step": 1156
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095372,
+      "loss": 1.5146,
+      "step": 1157
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095368,
+      "loss": 1.5358,
+      "step": 1158
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009536400000000001,
+      "loss": 1.5719,
+      "step": 1159
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009536,
+      "loss": 1.5412,
+      "step": 1160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095356,
+      "loss": 1.4992,
+      "step": 1161
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095352,
+      "loss": 1.3991,
+      "step": 1162
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095348,
+      "loss": 1.4768,
+      "step": 1163
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095344,
+      "loss": 1.6108,
+      "step": 1164
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009534000000000001,
+      "loss": 1.6465,
+      "step": 1165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095336,
+      "loss": 1.5418,
+      "step": 1166
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095332,
+      "loss": 1.5782,
+      "step": 1167
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095328,
+      "loss": 1.5318,
+      "step": 1168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095324,
+      "loss": 1.5392,
+      "step": 1169
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009532000000000001,
+      "loss": 1.593,
+      "step": 1170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095316,
+      "loss": 1.469,
+      "step": 1171
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095312,
+      "loss": 1.5771,
+      "step": 1172
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095308,
+      "loss": 1.471,
+      "step": 1173
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095304,
+      "loss": 1.4893,
+      "step": 1174
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000953,
+      "loss": 1.6152,
+      "step": 1175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009529600000000001,
+      "loss": 1.5645,
+      "step": 1176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095292,
+      "loss": 1.5592,
+      "step": 1177
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009528799999999999,
+      "loss": 1.5908,
+      "step": 1178
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095284,
+      "loss": 1.57,
+      "step": 1179
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009528,
+      "loss": 1.5667,
+      "step": 1180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009527600000000001,
+      "loss": 1.4972,
+      "step": 1181
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095272,
+      "loss": 1.6021,
+      "step": 1182
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095268,
+      "loss": 1.4837,
+      "step": 1183
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095264,
+      "loss": 1.6738,
+      "step": 1184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009526000000000001,
+      "loss": 1.547,
+      "step": 1185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095256,
+      "loss": 1.5218,
+      "step": 1186
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009525200000000001,
+      "loss": 1.4951,
+      "step": 1187
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095248,
+      "loss": 1.6311,
+      "step": 1188
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009524399999999999,
+      "loss": 1.5002,
+      "step": 1189
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009524000000000001,
+      "loss": 1.5636,
+      "step": 1190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095236,
+      "loss": 1.4968,
+      "step": 1191
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009523200000000001,
+      "loss": 1.5661,
+      "step": 1192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095228,
+      "loss": 1.4822,
+      "step": 1193
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095224,
+      "loss": 1.4591,
+      "step": 1194
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009522,
+      "loss": 1.6195,
+      "step": 1195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009521600000000001,
+      "loss": 1.5469,
+      "step": 1196
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095212,
+      "loss": 1.4876,
+      "step": 1197
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095208,
+      "loss": 1.6462,
+      "step": 1198
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095204,
+      "loss": 1.577,
+      "step": 1199
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009519999999999999,
+      "loss": 1.521,
+      "step": 1200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009519600000000001,
+      "loss": 1.5487,
+      "step": 1201
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095192,
+      "loss": 1.5449,
+      "step": 1202
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095188,
+      "loss": 1.4237,
+      "step": 1203
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095184,
+      "loss": 1.4846,
+      "step": 1204
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009518,
+      "loss": 1.6197,
+      "step": 1205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009517600000000001,
+      "loss": 1.5623,
+      "step": 1206
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009517200000000001,
+      "loss": 1.5528,
+      "step": 1207
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095168,
+      "loss": 1.5221,
+      "step": 1208
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095164,
+      "loss": 1.5365,
+      "step": 1209
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009516,
+      "loss": 1.4985,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095156,
+      "loss": 1.5904,
+      "step": 1211
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009515200000000001,
+      "loss": 1.5414,
+      "step": 1212
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095148,
+      "loss": 1.6505,
+      "step": 1213
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095144,
+      "loss": 1.499,
+      "step": 1214
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009514,
+      "loss": 1.661,
+      "step": 1215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095136,
+      "loss": 1.4437,
+      "step": 1216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009513200000000001,
+      "loss": 1.4985,
+      "step": 1217
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095128,
+      "loss": 1.4806,
+      "step": 1218
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095124,
+      "loss": 1.4747,
+      "step": 1219
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009512,
+      "loss": 1.4865,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095116,
+      "loss": 1.4831,
+      "step": 1221
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095112,
+      "loss": 1.4851,
+      "step": 1222
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009510800000000001,
+      "loss": 1.4228,
+      "step": 1223
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095104,
+      "loss": 1.4863,
+      "step": 1224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000951,
+      "loss": 1.4176,
+      "step": 1225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095096,
+      "loss": 1.437,
+      "step": 1226
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095092,
+      "loss": 1.47,
+      "step": 1227
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095088,
+      "loss": 1.3556,
+      "step": 1228
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095084,
+      "loss": 1.567,
+      "step": 1229
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009508,
+      "loss": 1.5568,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095076,
+      "loss": 1.4899,
+      "step": 1231
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009507200000000001,
+      "loss": 1.6236,
+      "step": 1232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095068,
+      "loss": 1.6152,
+      "step": 1233
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009506400000000001,
+      "loss": 1.5436,
+      "step": 1234
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009506,
+      "loss": 1.471,
+      "step": 1235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095056,
+      "loss": 1.5831,
+      "step": 1236
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095052,
+      "loss": 1.4628,
+      "step": 1237
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095048,
+      "loss": 1.6251,
+      "step": 1238
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095044,
+      "loss": 1.4881,
+      "step": 1239
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009504,
+      "loss": 1.6044,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095036,
+      "loss": 1.4964,
+      "step": 1241
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095032,
+      "loss": 1.6101,
+      "step": 1242
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009502800000000001,
+      "loss": 1.609,
+      "step": 1243
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095024,
+      "loss": 1.5789,
+      "step": 1244
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009502000000000001,
+      "loss": 1.456,
+      "step": 1245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095016,
+      "loss": 1.462,
+      "step": 1246
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009501199999999999,
+      "loss": 1.6271,
+      "step": 1247
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009500800000000001,
+      "loss": 1.4935,
+      "step": 1248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095004,
+      "loss": 1.4697,
+      "step": 1249
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00095,
+      "loss": 1.4994,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094996,
+      "loss": 1.5881,
+      "step": 1251
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094992,
+      "loss": 1.5313,
+      "step": 1252
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094988,
+      "loss": 1.4963,
+      "step": 1253
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009498400000000001,
+      "loss": 1.6061,
+      "step": 1254
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009498,
+      "loss": 1.4339,
+      "step": 1255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009497600000000001,
+      "loss": 1.5469,
+      "step": 1256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094972,
+      "loss": 1.5234,
+      "step": 1257
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094968,
+      "loss": 1.5188,
+      "step": 1258
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009496400000000001,
+      "loss": 1.5567,
+      "step": 1259
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009496,
+      "loss": 1.4672,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094956,
+      "loss": 1.4124,
+      "step": 1261
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094952,
+      "loss": 1.5082,
+      "step": 1262
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094948,
+      "loss": 1.5931,
+      "step": 1263
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094944,
+      "loss": 1.4624,
+      "step": 1264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009494000000000001,
+      "loss": 1.5531,
+      "step": 1265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094936,
+      "loss": 1.5542,
+      "step": 1266
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094932,
+      "loss": 1.6052,
+      "step": 1267
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094928,
+      "loss": 1.5057,
+      "step": 1268
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094924,
+      "loss": 1.5675,
+      "step": 1269
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009492000000000001,
+      "loss": 1.4559,
+      "step": 1270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094916,
+      "loss": 1.5459,
+      "step": 1271
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094912,
+      "loss": 1.5724,
+      "step": 1272
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094908,
+      "loss": 1.5583,
+      "step": 1273
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009490400000000001,
+      "loss": 1.5253,
+      "step": 1274
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000949,
+      "loss": 1.5275,
+      "step": 1275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094896,
+      "loss": 1.5379,
+      "step": 1276
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094892,
+      "loss": 1.5411,
+      "step": 1277
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009488799999999999,
+      "loss": 1.551,
+      "step": 1278
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009488400000000001,
+      "loss": 1.4218,
+      "step": 1279
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009488,
+      "loss": 1.5443,
+      "step": 1280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009487600000000001,
+      "loss": 1.5395,
+      "step": 1281
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094872,
+      "loss": 1.498,
+      "step": 1282
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094868,
+      "loss": 1.4815,
+      "step": 1283
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094864,
+      "loss": 1.4756,
+      "step": 1284
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009486000000000001,
+      "loss": 1.4774,
+      "step": 1285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094856,
+      "loss": 1.3922,
+      "step": 1286
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094852,
+      "loss": 1.528,
+      "step": 1287
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094848,
+      "loss": 1.5891,
+      "step": 1288
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009484399999999999,
+      "loss": 1.5458,
+      "step": 1289
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009484000000000001,
+      "loss": 1.4874,
+      "step": 1290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094836,
+      "loss": 1.4948,
+      "step": 1291
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009483200000000001,
+      "loss": 1.5485,
+      "step": 1292
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094828,
+      "loss": 1.5903,
+      "step": 1293
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094824,
+      "loss": 1.5398,
+      "step": 1294
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009482000000000001,
+      "loss": 1.5058,
+      "step": 1295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094816,
+      "loss": 1.4457,
+      "step": 1296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094812,
+      "loss": 1.4675,
+      "step": 1297
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094808,
+      "loss": 1.4808,
+      "step": 1298
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094804,
+      "loss": 1.4931,
+      "step": 1299
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000948,
+      "loss": 1.5326,
+      "step": 1300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009479600000000001,
+      "loss": 1.5789,
+      "step": 1301
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094792,
+      "loss": 1.5442,
+      "step": 1302
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094788,
+      "loss": 1.5456,
+      "step": 1303
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094784,
+      "loss": 1.5118,
+      "step": 1304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009477999999999999,
+      "loss": 1.5636,
+      "step": 1305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009477600000000001,
+      "loss": 1.4632,
+      "step": 1306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094772,
+      "loss": 1.5401,
+      "step": 1307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094768,
+      "loss": 1.4173,
+      "step": 1308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094764,
+      "loss": 1.5764,
+      "step": 1309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009476,
+      "loss": 1.5103,
+      "step": 1310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094756,
+      "loss": 1.5279,
+      "step": 1311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009475200000000001,
+      "loss": 1.5186,
+      "step": 1312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094748,
+      "loss": 1.5792,
+      "step": 1313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094744,
+      "loss": 1.5519,
+      "step": 1314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009474,
+      "loss": 1.5062,
+      "step": 1315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094736,
+      "loss": 1.5722,
+      "step": 1316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009473200000000001,
+      "loss": 1.4989,
+      "step": 1317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094728,
+      "loss": 1.516,
+      "step": 1318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094724,
+      "loss": 1.4795,
+      "step": 1319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009472,
+      "loss": 1.5131,
+      "step": 1320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009471600000000001,
+      "loss": 1.4869,
+      "step": 1321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094712,
+      "loss": 1.4872,
+      "step": 1322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009470800000000001,
+      "loss": 1.5537,
+      "step": 1323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094704,
+      "loss": 1.4111,
+      "step": 1324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009469999999999999,
+      "loss": 1.4738,
+      "step": 1325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009469600000000001,
+      "loss": 1.5867,
+      "step": 1326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094692,
+      "loss": 1.5109,
+      "step": 1327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009468800000000001,
+      "loss": 1.5109,
+      "step": 1328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094684,
+      "loss": 1.576,
+      "step": 1329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009468,
+      "loss": 1.4157,
+      "step": 1330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094676,
+      "loss": 1.5399,
+      "step": 1331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009467200000000001,
+      "loss": 1.4916,
+      "step": 1332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094668,
+      "loss": 1.5964,
+      "step": 1333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094664,
+      "loss": 1.5396,
+      "step": 1334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009466,
+      "loss": 1.4789,
+      "step": 1335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009465599999999999,
+      "loss": 1.4549,
+      "step": 1336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009465200000000001,
+      "loss": 1.4559,
+      "step": 1337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094648,
+      "loss": 1.5519,
+      "step": 1338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094644,
+      "loss": 1.4742,
+      "step": 1339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009464,
+      "loss": 1.4251,
+      "step": 1340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094636,
+      "loss": 1.5458,
+      "step": 1341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009463200000000001,
+      "loss": 1.4732,
+      "step": 1342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009462800000000001,
+      "loss": 1.5125,
+      "step": 1343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094624,
+      "loss": 1.5492,
+      "step": 1344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009462,
+      "loss": 1.5365,
+      "step": 1345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094616,
+      "loss": 1.4919,
+      "step": 1346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094612,
+      "loss": 1.6372,
+      "step": 1347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009460800000000001,
+      "loss": 1.4416,
+      "step": 1348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094604,
+      "loss": 1.423,
+      "step": 1349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000946,
+      "loss": 1.5726,
+      "step": 1350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094596,
+      "loss": 1.5983,
+      "step": 1351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094592,
+      "loss": 1.5467,
+      "step": 1352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009458800000000001,
+      "loss": 1.4612,
+      "step": 1353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094584,
+      "loss": 1.573,
+      "step": 1354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009458,
+      "loss": 1.4113,
+      "step": 1355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094576,
+      "loss": 1.4837,
+      "step": 1356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094572,
+      "loss": 1.5255,
+      "step": 1357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094568,
+      "loss": 1.4062,
+      "step": 1358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009456400000000001,
+      "loss": 1.4199,
+      "step": 1359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009456,
+      "loss": 1.5698,
+      "step": 1360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094556,
+      "loss": 1.3345,
+      "step": 1361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094552,
+      "loss": 1.5106,
+      "step": 1362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094548,
+      "loss": 1.5418,
+      "step": 1363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094544,
+      "loss": 1.4487,
+      "step": 1364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009454,
+      "loss": 1.577,
+      "step": 1365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094536,
+      "loss": 1.4039,
+      "step": 1366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094532,
+      "loss": 1.4809,
+      "step": 1367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009452800000000001,
+      "loss": 1.4227,
+      "step": 1368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094524,
+      "loss": 1.5047,
+      "step": 1369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009452000000000001,
+      "loss": 1.5219,
+      "step": 1370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094516,
+      "loss": 1.4741,
+      "step": 1371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094512,
+      "loss": 1.3555,
+      "step": 1372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094508,
+      "loss": 1.3605,
+      "step": 1373
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094504,
+      "loss": 1.4525,
+      "step": 1374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000945,
+      "loss": 1.5186,
+      "step": 1375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094496,
+      "loss": 1.3728,
+      "step": 1376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094492,
+      "loss": 1.4857,
+      "step": 1377
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094488,
+      "loss": 1.5216,
+      "step": 1378
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009448400000000001,
+      "loss": 1.441,
+      "step": 1379
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009448,
+      "loss": 1.5648,
+      "step": 1380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009447600000000001,
+      "loss": 1.4977,
+      "step": 1381
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094472,
+      "loss": 1.4278,
+      "step": 1382
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009446799999999999,
+      "loss": 1.4801,
+      "step": 1383
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009446400000000001,
+      "loss": 1.4089,
+      "step": 1384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009446,
+      "loss": 1.4949,
+      "step": 1385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094456,
+      "loss": 1.4536,
+      "step": 1386
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094452,
+      "loss": 1.4548,
+      "step": 1387
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094448,
+      "loss": 1.5052,
+      "step": 1388
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094444,
+      "loss": 1.4852,
+      "step": 1389
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009444000000000001,
+      "loss": 1.5391,
+      "step": 1390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094436,
+      "loss": 1.4966,
+      "step": 1391
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0009443200000000001,
+      "loss": 1.4599,
+      "step": 1392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00094428,
+      "loss": 1.4525,
+      "step": 1393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009442399999999999,
+      "loss": 1.4992,
+      "step": 1394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009442000000000001,
+      "loss": 1.367,
+      "step": 1395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094416,
+      "loss": 1.5604,
+      "step": 1396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094412,
+      "loss": 1.4494,
+      "step": 1397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094408,
+      "loss": 1.526,
+      "step": 1398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094404,
+      "loss": 1.4947,
+      "step": 1399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000944,
+      "loss": 1.475,
+      "step": 1400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009439600000000001,
+      "loss": 1.5108,
+      "step": 1401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094392,
+      "loss": 1.4455,
+      "step": 1402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094388,
+      "loss": 1.5445,
+      "step": 1403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094384,
+      "loss": 1.5048,
+      "step": 1404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009438,
+      "loss": 1.39,
+      "step": 1405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009437600000000001,
+      "loss": 1.4837,
+      "step": 1406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094372,
+      "loss": 1.4744,
+      "step": 1407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094368,
+      "loss": 1.5508,
+      "step": 1408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094364,
+      "loss": 1.5329,
+      "step": 1409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009436000000000001,
+      "loss": 1.5,
+      "step": 1410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094356,
+      "loss": 1.4709,
+      "step": 1411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094352,
+      "loss": 1.5191,
+      "step": 1412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094348,
+      "loss": 1.4685,
+      "step": 1413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009434399999999999,
+      "loss": 1.49,
+      "step": 1414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009434000000000001,
+      "loss": 1.4896,
+      "step": 1415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094336,
+      "loss": 1.556,
+      "step": 1416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009433200000000001,
+      "loss": 1.427,
+      "step": 1417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094328,
+      "loss": 1.4679,
+      "step": 1418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094324,
+      "loss": 1.4792,
+      "step": 1419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009432,
+      "loss": 1.4486,
+      "step": 1420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009431600000000001,
+      "loss": 1.4513,
+      "step": 1421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094312,
+      "loss": 1.5066,
+      "step": 1422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094308,
+      "loss": 1.4879,
+      "step": 1423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094304,
+      "loss": 1.5185,
+      "step": 1424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009429999999999999,
+      "loss": 1.5165,
+      "step": 1425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009429600000000001,
+      "loss": 1.5242,
+      "step": 1426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094292,
+      "loss": 1.5416,
+      "step": 1427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009428800000000001,
+      "loss": 1.5858,
+      "step": 1428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094284,
+      "loss": 1.5864,
+      "step": 1429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009428,
+      "loss": 1.4334,
+      "step": 1430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009427600000000001,
+      "loss": 1.5092,
+      "step": 1431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094272,
+      "loss": 1.4396,
+      "step": 1432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094268,
+      "loss": 1.5486,
+      "step": 1433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094264,
+      "loss": 1.477,
+      "step": 1434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009426,
+      "loss": 1.5607,
+      "step": 1435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094256,
+      "loss": 1.5228,
+      "step": 1436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009425200000000001,
+      "loss": 1.5022,
+      "step": 1437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094248,
+      "loss": 1.4731,
+      "step": 1438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009424400000000001,
+      "loss": 1.4208,
+      "step": 1439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009424,
+      "loss": 1.4636,
+      "step": 1440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009423599999999999,
+      "loss": 1.4634,
+      "step": 1441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009423200000000001,
+      "loss": 1.4623,
+      "step": 1442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094228,
+      "loss": 1.4836,
+      "step": 1443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094224,
+      "loss": 1.5308,
+      "step": 1444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009422,
+      "loss": 1.6092,
+      "step": 1445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094216,
+      "loss": 1.3994,
+      "step": 1446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094212,
+      "loss": 1.5261,
+      "step": 1447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009420800000000001,
+      "loss": 1.5634,
+      "step": 1448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094204,
+      "loss": 1.5469,
+      "step": 1449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000942,
+      "loss": 1.5217,
+      "step": 1450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094196,
+      "loss": 1.4486,
+      "step": 1451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094192,
+      "loss": 1.495,
+      "step": 1452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009418800000000001,
+      "loss": 1.4443,
+      "step": 1453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094184,
+      "loss": 1.4195,
+      "step": 1454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009418,
+      "loss": 1.5045,
+      "step": 1455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094176,
+      "loss": 1.4786,
+      "step": 1456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009417200000000001,
+      "loss": 1.4598,
+      "step": 1457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094168,
+      "loss": 1.4173,
+      "step": 1458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009416400000000001,
+      "loss": 1.5119,
+      "step": 1459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009416,
+      "loss": 1.4517,
+      "step": 1460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009415599999999999,
+      "loss": 1.4898,
+      "step": 1461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094152,
+      "loss": 1.4439,
+      "step": 1462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094148,
+      "loss": 1.4702,
+      "step": 1463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009414400000000001,
+      "loss": 1.3934,
+      "step": 1464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009414,
+      "loss": 1.4679,
+      "step": 1465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094136,
+      "loss": 1.4181,
+      "step": 1466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094132,
+      "loss": 1.4515,
+      "step": 1467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009412800000000001,
+      "loss": 1.4515,
+      "step": 1468
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094124,
+      "loss": 1.4763,
+      "step": 1469
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009412000000000001,
+      "loss": 1.538,
+      "step": 1470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094116,
+      "loss": 1.5301,
+      "step": 1471
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009411199999999999,
+      "loss": 1.4492,
+      "step": 1472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009410800000000001,
+      "loss": 1.5629,
+      "step": 1473
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094104,
+      "loss": 1.474,
+      "step": 1474
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000941,
+      "loss": 1.448,
+      "step": 1475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094096,
+      "loss": 1.5641,
+      "step": 1476
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094092,
+      "loss": 1.5128,
+      "step": 1477
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009408800000000001,
+      "loss": 1.4174,
+      "step": 1478
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009408400000000001,
+      "loss": 1.4809,
+      "step": 1479
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009408,
+      "loss": 1.4529,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094076,
+      "loss": 1.4389,
+      "step": 1481
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094072,
+      "loss": 1.516,
+      "step": 1482
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009406799999999999,
+      "loss": 1.3982,
+      "step": 1483
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009406400000000001,
+      "loss": 1.4797,
+      "step": 1484
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009406,
+      "loss": 1.5238,
+      "step": 1485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094056,
+      "loss": 1.4903,
+      "step": 1486
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094052,
+      "loss": 1.4982,
+      "step": 1487
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094048,
+      "loss": 1.5041,
+      "step": 1488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009404400000000001,
+      "loss": 1.5254,
+      "step": 1489
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009404,
+      "loss": 1.5234,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094036,
+      "loss": 1.5272,
+      "step": 1491
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094032,
+      "loss": 1.4463,
+      "step": 1492
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094028,
+      "loss": 1.539,
+      "step": 1493
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094024,
+      "loss": 1.5147,
+      "step": 1494
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009402000000000001,
+      "loss": 1.4541,
+      "step": 1495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094016,
+      "loss": 1.4718,
+      "step": 1496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094012,
+      "loss": 1.6623,
+      "step": 1497
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094008,
+      "loss": 1.4853,
+      "step": 1498
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009400400000000001,
+      "loss": 1.4985,
+      "step": 1499
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00094,
+      "loss": 1.4524,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093996,
+      "loss": 1.4652,
+      "step": 1501
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093992,
+      "loss": 1.4725,
+      "step": 1502
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093988,
+      "loss": 1.4493,
+      "step": 1503
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009398400000000001,
+      "loss": 1.4388,
+      "step": 1504
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009398,
+      "loss": 1.4106,
+      "step": 1505
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009397600000000001,
+      "loss": 1.504,
+      "step": 1506
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093972,
+      "loss": 1.4489,
+      "step": 1507
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093968,
+      "loss": 1.4787,
+      "step": 1508
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093964,
+      "loss": 1.4942,
+      "step": 1509
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009396,
+      "loss": 1.3949,
+      "step": 1510
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093956,
+      "loss": 1.492,
+      "step": 1511
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093952,
+      "loss": 1.5501,
+      "step": 1512
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093948,
+      "loss": 1.4077,
+      "step": 1513
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093944,
+      "loss": 1.5691,
+      "step": 1514
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009394000000000001,
+      "loss": 1.508,
+      "step": 1515
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093936,
+      "loss": 1.5346,
+      "step": 1516
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009393200000000001,
+      "loss": 1.4132,
+      "step": 1517
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093928,
+      "loss": 1.553,
+      "step": 1518
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009392399999999999,
+      "loss": 1.471,
+      "step": 1519
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009392000000000001,
+      "loss": 1.4326,
+      "step": 1520
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093916,
+      "loss": 1.4561,
+      "step": 1521
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093912,
+      "loss": 1.4913,
+      "step": 1522
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093908,
+      "loss": 1.4924,
+      "step": 1523
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093904,
+      "loss": 1.4683,
+      "step": 1524
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000939,
+      "loss": 1.5429,
+      "step": 1525
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009389600000000001,
+      "loss": 1.435,
+      "step": 1526
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093892,
+      "loss": 1.4893,
+      "step": 1527
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009388800000000001,
+      "loss": 1.5442,
+      "step": 1528
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093884,
+      "loss": 1.4448,
+      "step": 1529
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009387999999999999,
+      "loss": 1.4864,
+      "step": 1530
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009387600000000001,
+      "loss": 1.5733,
+      "step": 1531
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093872,
+      "loss": 1.4367,
+      "step": 1532
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093868,
+      "loss": 1.3816,
+      "step": 1533
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093864,
+      "loss": 1.5212,
+      "step": 1534
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009386,
+      "loss": 1.5657,
+      "step": 1535
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093856,
+      "loss": 1.5516,
+      "step": 1536
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009385200000000001,
+      "loss": 1.4633,
+      "step": 1537
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093848,
+      "loss": 1.5245,
+      "step": 1538
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093844,
+      "loss": 1.4445,
+      "step": 1539
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009384,
+      "loss": 1.4318,
+      "step": 1540
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093836,
+      "loss": 1.526,
+      "step": 1541
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009383200000000001,
+      "loss": 1.4644,
+      "step": 1542
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093828,
+      "loss": 1.5192,
+      "step": 1543
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093824,
+      "loss": 1.4725,
+      "step": 1544
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009382,
+      "loss": 1.4447,
+      "step": 1545
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009381600000000001,
+      "loss": 1.5077,
+      "step": 1546
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093812,
+      "loss": 1.5218,
+      "step": 1547
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009380800000000001,
+      "loss": 1.5817,
+      "step": 1548
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093804,
+      "loss": 1.5184,
+      "step": 1549
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009379999999999999,
+      "loss": 1.4203,
+      "step": 1550
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093796,
+      "loss": 1.5448,
+      "step": 1551
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093792,
+      "loss": 1.5111,
+      "step": 1552
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009378800000000001,
+      "loss": 1.3845,
+      "step": 1553
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093784,
+      "loss": 1.4475,
+      "step": 1554
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009378,
+      "loss": 1.4955,
+      "step": 1555
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093776,
+      "loss": 1.4516,
+      "step": 1556
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009377200000000001,
+      "loss": 1.3918,
+      "step": 1557
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093768,
+      "loss": 1.4841,
+      "step": 1558
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093764,
+      "loss": 1.5009,
+      "step": 1559
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009376,
+      "loss": 1.4624,
+      "step": 1560
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009375599999999999,
+      "loss": 1.4483,
+      "step": 1561
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009375200000000001,
+      "loss": 1.5384,
+      "step": 1562
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093748,
+      "loss": 1.4359,
+      "step": 1563
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009374400000000001,
+      "loss": 1.5306,
+      "step": 1564
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009374,
+      "loss": 1.6409,
+      "step": 1565
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093736,
+      "loss": 1.5001,
+      "step": 1566
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009373200000000001,
+      "loss": 1.5455,
+      "step": 1567
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093728,
+      "loss": 1.4042,
+      "step": 1568
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093724,
+      "loss": 1.3074,
+      "step": 1569
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009372,
+      "loss": 1.5223,
+      "step": 1570
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093716,
+      "loss": 1.4491,
+      "step": 1571
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009371199999999999,
+      "loss": 1.5832,
+      "step": 1572
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009370800000000001,
+      "loss": 1.5085,
+      "step": 1573
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093704,
+      "loss": 1.4583,
+      "step": 1574
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009370000000000001,
+      "loss": 1.5072,
+      "step": 1575
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093696,
+      "loss": 1.407,
+      "step": 1576
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009369199999999999,
+      "loss": 1.356,
+      "step": 1577
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009368800000000001,
+      "loss": 1.5186,
+      "step": 1578
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093684,
+      "loss": 1.4429,
+      "step": 1579
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009368,
+      "loss": 1.5092,
+      "step": 1580
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093676,
+      "loss": 1.4556,
+      "step": 1581
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093672,
+      "loss": 1.4582,
+      "step": 1582
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093668,
+      "loss": 1.3512,
+      "step": 1583
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009366400000000001,
+      "loss": 1.4477,
+      "step": 1584
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009366,
+      "loss": 1.451,
+      "step": 1585
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093656,
+      "loss": 1.4851,
+      "step": 1586
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093652,
+      "loss": 1.5025,
+      "step": 1587
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093648,
+      "loss": 1.4976,
+      "step": 1588
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009364400000000001,
+      "loss": 1.4992,
+      "step": 1589
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009364,
+      "loss": 1.4719,
+      "step": 1590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093636,
+      "loss": 1.5113,
+      "step": 1591
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093632,
+      "loss": 1.4634,
+      "step": 1592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009362800000000001,
+      "loss": 1.4291,
+      "step": 1593
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093624,
+      "loss": 1.3412,
+      "step": 1594
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009362000000000001,
+      "loss": 1.5056,
+      "step": 1595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093616,
+      "loss": 1.5467,
+      "step": 1596
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009361199999999999,
+      "loss": 1.4729,
+      "step": 1597
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093608,
+      "loss": 1.519,
+      "step": 1598
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093604,
+      "loss": 1.3889,
+      "step": 1599
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009360000000000001,
+      "loss": 1.4145,
+      "step": 1600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093596,
+      "loss": 1.5005,
+      "step": 1601
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093592,
+      "loss": 1.4715,
+      "step": 1602
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093588,
+      "loss": 1.495,
+      "step": 1603
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009358400000000001,
+      "loss": 1.464,
+      "step": 1604
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009358,
+      "loss": 1.4457,
+      "step": 1605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009357600000000001,
+      "loss": 1.5112,
+      "step": 1606
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093572,
+      "loss": 1.4427,
+      "step": 1607
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009356799999999999,
+      "loss": 1.3929,
+      "step": 1608
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009356400000000001,
+      "loss": 1.4527,
+      "step": 1609
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009356,
+      "loss": 1.3825,
+      "step": 1610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093556,
+      "loss": 1.5649,
+      "step": 1611
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093552,
+      "loss": 1.4776,
+      "step": 1612
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093548,
+      "loss": 1.4992,
+      "step": 1613
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009354400000000001,
+      "loss": 1.4498,
+      "step": 1614
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009354000000000001,
+      "loss": 1.5285,
+      "step": 1615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093536,
+      "loss": 1.3552,
+      "step": 1616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093532,
+      "loss": 1.4816,
+      "step": 1617
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093528,
+      "loss": 1.6216,
+      "step": 1618
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009352399999999999,
+      "loss": 1.5534,
+      "step": 1619
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009352000000000001,
+      "loss": 1.529,
+      "step": 1620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093516,
+      "loss": 1.5076,
+      "step": 1621
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093512,
+      "loss": 1.4902,
+      "step": 1622
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093508,
+      "loss": 1.4921,
+      "step": 1623
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093504,
+      "loss": 1.4939,
+      "step": 1624
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009350000000000001,
+      "loss": 1.4377,
+      "step": 1625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093496,
+      "loss": 1.3231,
+      "step": 1626
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093492,
+      "loss": 1.4533,
+      "step": 1627
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093488,
+      "loss": 1.4712,
+      "step": 1628
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093484,
+      "loss": 1.483,
+      "step": 1629
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009348,
+      "loss": 1.454,
+      "step": 1630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009347600000000001,
+      "loss": 1.5299,
+      "step": 1631
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093472,
+      "loss": 1.4442,
+      "step": 1632
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093468,
+      "loss": 1.4936,
+      "step": 1633
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093464,
+      "loss": 1.4685,
+      "step": 1634
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009346000000000001,
+      "loss": 1.4183,
+      "step": 1635
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093456,
+      "loss": 1.4402,
+      "step": 1636
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093452,
+      "loss": 1.4275,
+      "step": 1637
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093448,
+      "loss": 1.4938,
+      "step": 1638
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093444,
+      "loss": 1.4346,
+      "step": 1639
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009344,
+      "loss": 1.4178,
+      "step": 1640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093436,
+      "loss": 1.5513,
+      "step": 1641
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009343200000000001,
+      "loss": 1.3569,
+      "step": 1642
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093428,
+      "loss": 1.4952,
+      "step": 1643
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093424,
+      "loss": 1.4295,
+      "step": 1644
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0009342,
+      "loss": 1.5334,
+      "step": 1645
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093416,
+      "loss": 1.5585,
+      "step": 1646
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00093412,
+      "loss": 1.4571,
+      "step": 1647
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093408,
+      "loss": 1.4678,
+      "step": 1648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093404,
+      "loss": 1.5153,
+      "step": 1649
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000934,
+      "loss": 1.5083,
+      "step": 1650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009339600000000001,
+      "loss": 1.4104,
+      "step": 1651
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093392,
+      "loss": 1.445,
+      "step": 1652
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009338800000000001,
+      "loss": 1.416,
+      "step": 1653
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093384,
+      "loss": 1.475,
+      "step": 1654
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009337999999999999,
+      "loss": 1.5507,
+      "step": 1655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009337600000000001,
+      "loss": 1.5793,
+      "step": 1656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093372,
+      "loss": 1.4525,
+      "step": 1657
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093368,
+      "loss": 1.4571,
+      "step": 1658
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093364,
+      "loss": 1.4504,
+      "step": 1659
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009336,
+      "loss": 1.4263,
+      "step": 1660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009335599999999999,
+      "loss": 1.4166,
+      "step": 1661
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009335200000000001,
+      "loss": 1.4678,
+      "step": 1662
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093348,
+      "loss": 1.4821,
+      "step": 1663
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009334400000000001,
+      "loss": 1.5756,
+      "step": 1664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009334,
+      "loss": 1.4393,
+      "step": 1665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009333599999999999,
+      "loss": 1.421,
+      "step": 1666
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009333200000000001,
+      "loss": 1.4723,
+      "step": 1667
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093328,
+      "loss": 1.4758,
+      "step": 1668
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093324,
+      "loss": 1.4866,
+      "step": 1669
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009332,
+      "loss": 1.4899,
+      "step": 1670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093316,
+      "loss": 1.396,
+      "step": 1671
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093312,
+      "loss": 1.4061,
+      "step": 1672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009330800000000001,
+      "loss": 1.4061,
+      "step": 1673
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093304,
+      "loss": 1.4484,
+      "step": 1674
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000933,
+      "loss": 1.4289,
+      "step": 1675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093296,
+      "loss": 1.5288,
+      "step": 1676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093292,
+      "loss": 1.5201,
+      "step": 1677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009328800000000001,
+      "loss": 1.4659,
+      "step": 1678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093284,
+      "loss": 1.3674,
+      "step": 1679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009328,
+      "loss": 1.52,
+      "step": 1680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093276,
+      "loss": 1.4213,
+      "step": 1681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009327200000000001,
+      "loss": 1.5035,
+      "step": 1682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093268,
+      "loss": 1.5646,
+      "step": 1683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009326400000000001,
+      "loss": 1.38,
+      "step": 1684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009326,
+      "loss": 1.5359,
+      "step": 1685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009325599999999999,
+      "loss": 1.4516,
+      "step": 1686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093252,
+      "loss": 1.5208,
+      "step": 1687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093248,
+      "loss": 1.3969,
+      "step": 1688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009324400000000001,
+      "loss": 1.496,
+      "step": 1689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009324,
+      "loss": 1.4625,
+      "step": 1690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093236,
+      "loss": 1.5004,
+      "step": 1691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093232,
+      "loss": 1.4642,
+      "step": 1692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009322800000000001,
+      "loss": 1.4092,
+      "step": 1693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093224,
+      "loss": 1.5002,
+      "step": 1694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009322,
+      "loss": 1.484,
+      "step": 1695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093216,
+      "loss": 1.4588,
+      "step": 1696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009321199999999999,
+      "loss": 1.4638,
+      "step": 1697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009320800000000001,
+      "loss": 1.4824,
+      "step": 1698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093204,
+      "loss": 1.4822,
+      "step": 1699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009320000000000001,
+      "loss": 1.452,
+      "step": 1700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093196,
+      "loss": 1.4678,
+      "step": 1701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093192,
+      "loss": 1.5188,
+      "step": 1702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009318800000000001,
+      "loss": 1.5164,
+      "step": 1703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093184,
+      "loss": 1.4576,
+      "step": 1704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009318,
+      "loss": 1.4584,
+      "step": 1705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093176,
+      "loss": 1.4538,
+      "step": 1706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093172,
+      "loss": 1.4415,
+      "step": 1707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009316799999999999,
+      "loss": 1.432,
+      "step": 1708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009316400000000001,
+      "loss": 1.4064,
+      "step": 1709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009316,
+      "loss": 1.4842,
+      "step": 1710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009315600000000001,
+      "loss": 1.386,
+      "step": 1711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093152,
+      "loss": 1.4635,
+      "step": 1712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093148,
+      "loss": 1.483,
+      "step": 1713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009314400000000001,
+      "loss": 1.5605,
+      "step": 1714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009314,
+      "loss": 1.5523,
+      "step": 1715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093136,
+      "loss": 1.5419,
+      "step": 1716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093132,
+      "loss": 1.4405,
+      "step": 1717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093128,
+      "loss": 1.4805,
+      "step": 1718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093124,
+      "loss": 1.4577,
+      "step": 1719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009312000000000001,
+      "loss": 1.4296,
+      "step": 1720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093116,
+      "loss": 1.3533,
+      "step": 1721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093112,
+      "loss": 1.4887,
+      "step": 1722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093108,
+      "loss": 1.4048,
+      "step": 1723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093104,
+      "loss": 1.4267,
+      "step": 1724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009310000000000001,
+      "loss": 1.3545,
+      "step": 1725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093096,
+      "loss": 1.5132,
+      "step": 1726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093092,
+      "loss": 1.4772,
+      "step": 1727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093088,
+      "loss": 1.4488,
+      "step": 1728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093084,
+      "loss": 1.4809,
+      "step": 1729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009308,
+      "loss": 1.4113,
+      "step": 1730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009307600000000001,
+      "loss": 1.4093,
+      "step": 1731
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093072,
+      "loss": 1.5611,
+      "step": 1732
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009306799999999999,
+      "loss": 1.4239,
+      "step": 1733
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093064,
+      "loss": 1.2969,
+      "step": 1734
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009306,
+      "loss": 1.5053,
+      "step": 1735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009305600000000001,
+      "loss": 1.4562,
+      "step": 1736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093052,
+      "loss": 1.5625,
+      "step": 1737
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093048,
+      "loss": 1.4956,
+      "step": 1738
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093044,
+      "loss": 1.4668,
+      "step": 1739
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009304000000000001,
+      "loss": 1.4042,
+      "step": 1740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093036,
+      "loss": 1.5258,
+      "step": 1741
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009303200000000001,
+      "loss": 1.3615,
+      "step": 1742
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093028,
+      "loss": 1.4482,
+      "step": 1743
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009302399999999999,
+      "loss": 1.4185,
+      "step": 1744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009302000000000001,
+      "loss": 1.4842,
+      "step": 1745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093016,
+      "loss": 1.413,
+      "step": 1746
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093012,
+      "loss": 1.3823,
+      "step": 1747
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093008,
+      "loss": 1.4308,
+      "step": 1748
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093004,
+      "loss": 1.3516,
+      "step": 1749
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00093,
+      "loss": 1.4663,
+      "step": 1750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009299600000000001,
+      "loss": 1.4955,
+      "step": 1751
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092992,
+      "loss": 1.4118,
+      "step": 1752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092988,
+      "loss": 1.6147,
+      "step": 1753
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092984,
+      "loss": 1.3887,
+      "step": 1754
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009297999999999999,
+      "loss": 1.3973,
+      "step": 1755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009297600000000001,
+      "loss": 1.4476,
+      "step": 1756
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092972,
+      "loss": 1.4109,
+      "step": 1757
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092968,
+      "loss": 1.4848,
+      "step": 1758
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092964,
+      "loss": 1.4536,
+      "step": 1759
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009296,
+      "loss": 1.4466,
+      "step": 1760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009295600000000001,
+      "loss": 1.4043,
+      "step": 1761
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009295200000000001,
+      "loss": 1.4422,
+      "step": 1762
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092948,
+      "loss": 1.6308,
+      "step": 1763
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092944,
+      "loss": 1.4161,
+      "step": 1764
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009294,
+      "loss": 1.4373,
+      "step": 1765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092936,
+      "loss": 1.4176,
+      "step": 1766
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009293200000000001,
+      "loss": 1.5588,
+      "step": 1767
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092928,
+      "loss": 1.3956,
+      "step": 1768
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092924,
+      "loss": 1.4933,
+      "step": 1769
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009292,
+      "loss": 1.4004,
+      "step": 1770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009291600000000001,
+      "loss": 1.4701,
+      "step": 1771
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092912,
+      "loss": 1.4337,
+      "step": 1772
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092908,
+      "loss": 1.4168,
+      "step": 1773
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092904,
+      "loss": 1.5134,
+      "step": 1774
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000929,
+      "loss": 1.3879,
+      "step": 1775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092896,
+      "loss": 1.3905,
+      "step": 1776
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092892,
+      "loss": 1.4518,
+      "step": 1777
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009288800000000001,
+      "loss": 1.3674,
+      "step": 1778
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092884,
+      "loss": 1.5378,
+      "step": 1779
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009288,
+      "loss": 1.4421,
+      "step": 1780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092876,
+      "loss": 1.5207,
+      "step": 1781
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092872,
+      "loss": 1.4852,
+      "step": 1782
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092868,
+      "loss": 1.4072,
+      "step": 1783
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092864,
+      "loss": 1.4068,
+      "step": 1784
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009286,
+      "loss": 1.5153,
+      "step": 1785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092856,
+      "loss": 1.4433,
+      "step": 1786
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009285200000000001,
+      "loss": 1.534,
+      "step": 1787
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092848,
+      "loss": 1.4926,
+      "step": 1788
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009284400000000001,
+      "loss": 1.5404,
+      "step": 1789
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009284,
+      "loss": 1.5194,
+      "step": 1790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092836,
+      "loss": 1.3863,
+      "step": 1791
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009283200000000001,
+      "loss": 1.5148,
+      "step": 1792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092828,
+      "loss": 1.4705,
+      "step": 1793
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092824,
+      "loss": 1.4987,
+      "step": 1794
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009282,
+      "loss": 1.4758,
+      "step": 1795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092816,
+      "loss": 1.4617,
+      "step": 1796
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009281199999999999,
+      "loss": 1.5526,
+      "step": 1797
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009280800000000001,
+      "loss": 1.5026,
+      "step": 1798
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092804,
+      "loss": 1.4822,
+      "step": 1799
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009280000000000001,
+      "loss": 1.5725,
+      "step": 1800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092796,
+      "loss": 1.4413,
+      "step": 1801
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009279199999999999,
+      "loss": 1.5061,
+      "step": 1802
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009278800000000001,
+      "loss": 1.4408,
+      "step": 1803
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092784,
+      "loss": 1.454,
+      "step": 1804
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009278,
+      "loss": 1.4315,
+      "step": 1805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092776,
+      "loss": 1.4395,
+      "step": 1806
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092772,
+      "loss": 1.4575,
+      "step": 1807
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092768,
+      "loss": 1.4604,
+      "step": 1808
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009276400000000001,
+      "loss": 1.5218,
+      "step": 1809
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009276,
+      "loss": 1.4561,
+      "step": 1810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092756,
+      "loss": 1.4393,
+      "step": 1811
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092752,
+      "loss": 1.4477,
+      "step": 1812
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092748,
+      "loss": 1.5324,
+      "step": 1813
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009274400000000001,
+      "loss": 1.4827,
+      "step": 1814
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009274,
+      "loss": 1.4063,
+      "step": 1815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092736,
+      "loss": 1.481,
+      "step": 1816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092732,
+      "loss": 1.4005,
+      "step": 1817
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092728,
+      "loss": 1.5436,
+      "step": 1818
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092724,
+      "loss": 1.4284,
+      "step": 1819
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009272000000000001,
+      "loss": 1.4497,
+      "step": 1820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092716,
+      "loss": 1.51,
+      "step": 1821
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092712,
+      "loss": 1.3834,
+      "step": 1822
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092708,
+      "loss": 1.4115,
+      "step": 1823
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092704,
+      "loss": 1.4118,
+      "step": 1824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009270000000000001,
+      "loss": 1.4523,
+      "step": 1825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092696,
+      "loss": 1.452,
+      "step": 1826
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092692,
+      "loss": 1.3896,
+      "step": 1827
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092688,
+      "loss": 1.5422,
+      "step": 1828
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009268400000000001,
+      "loss": 1.3715,
+      "step": 1829
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009268,
+      "loss": 1.4384,
+      "step": 1830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092676,
+      "loss": 1.5219,
+      "step": 1831
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092672,
+      "loss": 1.5165,
+      "step": 1832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009266799999999999,
+      "loss": 1.5408,
+      "step": 1833
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009266400000000001,
+      "loss": 1.4259,
+      "step": 1834
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009266,
+      "loss": 1.4662,
+      "step": 1835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009265600000000001,
+      "loss": 1.3897,
+      "step": 1836
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092652,
+      "loss": 1.5081,
+      "step": 1837
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092648,
+      "loss": 1.5224,
+      "step": 1838
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009264400000000001,
+      "loss": 1.587,
+      "step": 1839
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009264,
+      "loss": 1.4003,
+      "step": 1840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092636,
+      "loss": 1.5339,
+      "step": 1841
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092632,
+      "loss": 1.4425,
+      "step": 1842
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092628,
+      "loss": 1.5208,
+      "step": 1843
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009262399999999999,
+      "loss": 1.3786,
+      "step": 1844
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009262000000000001,
+      "loss": 1.4919,
+      "step": 1845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092616,
+      "loss": 1.3733,
+      "step": 1846
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009261200000000001,
+      "loss": 1.4091,
+      "step": 1847
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092608,
+      "loss": 1.4076,
+      "step": 1848
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092604,
+      "loss": 1.4605,
+      "step": 1849
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009260000000000001,
+      "loss": 1.4756,
+      "step": 1850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092596,
+      "loss": 1.3623,
+      "step": 1851
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092592,
+      "loss": 1.4641,
+      "step": 1852
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092588,
+      "loss": 1.4213,
+      "step": 1853
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092584,
+      "loss": 1.4797,
+      "step": 1854
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009258,
+      "loss": 1.4165,
+      "step": 1855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009257600000000001,
+      "loss": 1.4677,
+      "step": 1856
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092572,
+      "loss": 1.4712,
+      "step": 1857
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092568,
+      "loss": 1.3767,
+      "step": 1858
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092564,
+      "loss": 1.4361,
+      "step": 1859
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009256,
+      "loss": 1.4516,
+      "step": 1860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009255600000000001,
+      "loss": 1.4294,
+      "step": 1861
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092552,
+      "loss": 1.3868,
+      "step": 1862
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092548,
+      "loss": 1.5402,
+      "step": 1863
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092544,
+      "loss": 1.4371,
+      "step": 1864
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009254,
+      "loss": 1.4544,
+      "step": 1865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092536,
+      "loss": 1.4543,
+      "step": 1866
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009253200000000001,
+      "loss": 1.4975,
+      "step": 1867
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092528,
+      "loss": 1.4414,
+      "step": 1868
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009252399999999999,
+      "loss": 1.3622,
+      "step": 1869
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009252,
+      "loss": 1.4957,
+      "step": 1870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092516,
+      "loss": 1.4047,
+      "step": 1871
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009251200000000001,
+      "loss": 1.3914,
+      "step": 1872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092508,
+      "loss": 1.395,
+      "step": 1873
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092504,
+      "loss": 1.3949,
+      "step": 1874
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000925,
+      "loss": 1.5015,
+      "step": 1875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009249600000000001,
+      "loss": 1.478,
+      "step": 1876
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092492,
+      "loss": 1.3631,
+      "step": 1877
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009248800000000001,
+      "loss": 1.3895,
+      "step": 1878
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092484,
+      "loss": 1.4831,
+      "step": 1879
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009247999999999999,
+      "loss": 1.3748,
+      "step": 1880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009247600000000001,
+      "loss": 1.4855,
+      "step": 1881
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092472,
+      "loss": 1.5133,
+      "step": 1882
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092468,
+      "loss": 1.3343,
+      "step": 1883
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092464,
+      "loss": 1.469,
+      "step": 1884
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009246,
+      "loss": 1.4643,
+      "step": 1885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092456,
+      "loss": 1.4279,
+      "step": 1886
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009245200000000001,
+      "loss": 1.4856,
+      "step": 1887
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092448,
+      "loss": 1.3581,
+      "step": 1888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092444,
+      "loss": 1.4425,
+      "step": 1889
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009244,
+      "loss": 1.5237,
+      "step": 1890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009243599999999999,
+      "loss": 1.5304,
+      "step": 1891
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009243200000000001,
+      "loss": 1.4786,
+      "step": 1892
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092428,
+      "loss": 1.3954,
+      "step": 1893
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092424,
+      "loss": 1.3641,
+      "step": 1894
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009242,
+      "loss": 1.4382,
+      "step": 1895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092416,
+      "loss": 1.3538,
+      "step": 1896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009241200000000001,
+      "loss": 1.452,
+      "step": 1897
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0009240800000000001,
+      "loss": 1.4936,
+      "step": 1898
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00092404,
+      "loss": 1.4752,
+      "step": 1899
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000924,
+      "loss": 1.4596,
+      "step": 1900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092396,
+      "loss": 1.3761,
+      "step": 1901
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092392,
+      "loss": 1.4888,
+      "step": 1902
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009238800000000001,
+      "loss": 1.4193,
+      "step": 1903
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092384,
+      "loss": 1.439,
+      "step": 1904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009238,
+      "loss": 1.3728,
+      "step": 1905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092376,
+      "loss": 1.537,
+      "step": 1906
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092372,
+      "loss": 1.3643,
+      "step": 1907
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092368,
+      "loss": 1.5031,
+      "step": 1908
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092364,
+      "loss": 1.4675,
+      "step": 1909
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009236,
+      "loss": 1.4771,
+      "step": 1910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092356,
+      "loss": 1.349,
+      "step": 1911
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092352,
+      "loss": 1.4345,
+      "step": 1912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092348,
+      "loss": 1.4469,
+      "step": 1913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009234400000000001,
+      "loss": 1.4325,
+      "step": 1914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009234,
+      "loss": 1.4332,
+      "step": 1915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092336,
+      "loss": 1.4178,
+      "step": 1916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092332,
+      "loss": 1.4379,
+      "step": 1917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092328,
+      "loss": 1.343,
+      "step": 1918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092324,
+      "loss": 1.4441,
+      "step": 1919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009232,
+      "loss": 1.5095,
+      "step": 1920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092316,
+      "loss": 1.4292,
+      "step": 1921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092312,
+      "loss": 1.5441,
+      "step": 1922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009230800000000001,
+      "loss": 1.4125,
+      "step": 1923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092304,
+      "loss": 1.4588,
+      "step": 1924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009230000000000001,
+      "loss": 1.3778,
+      "step": 1925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092296,
+      "loss": 1.4322,
+      "step": 1926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092292,
+      "loss": 1.4168,
+      "step": 1927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009228800000000001,
+      "loss": 1.3689,
+      "step": 1928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092284,
+      "loss": 1.4125,
+      "step": 1929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009228,
+      "loss": 1.4875,
+      "step": 1930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092276,
+      "loss": 1.4078,
+      "step": 1931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092272,
+      "loss": 1.3731,
+      "step": 1932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009226799999999999,
+      "loss": 1.5505,
+      "step": 1933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009226400000000001,
+      "loss": 1.4907,
+      "step": 1934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009226,
+      "loss": 1.3824,
+      "step": 1935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009225600000000001,
+      "loss": 1.4774,
+      "step": 1936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092252,
+      "loss": 1.4375,
+      "step": 1937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009224799999999999,
+      "loss": 1.3849,
+      "step": 1938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009224400000000001,
+      "loss": 1.5268,
+      "step": 1939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009224,
+      "loss": 1.4841,
+      "step": 1940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092236,
+      "loss": 1.3241,
+      "step": 1941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092232,
+      "loss": 1.5435,
+      "step": 1942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092228,
+      "loss": 1.3887,
+      "step": 1943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092224,
+      "loss": 1.5429,
+      "step": 1944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009222000000000001,
+      "loss": 1.454,
+      "step": 1945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092216,
+      "loss": 1.4926,
+      "step": 1946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009221200000000001,
+      "loss": 1.3941,
+      "step": 1947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092208,
+      "loss": 1.448,
+      "step": 1948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092204,
+      "loss": 1.4052,
+      "step": 1949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009220000000000001,
+      "loss": 1.3609,
+      "step": 1950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092196,
+      "loss": 1.4438,
+      "step": 1951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092192,
+      "loss": 1.4273,
+      "step": 1952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092188,
+      "loss": 1.382,
+      "step": 1953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092184,
+      "loss": 1.3987,
+      "step": 1954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009218,
+      "loss": 1.3915,
+      "step": 1955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009217600000000001,
+      "loss": 1.5337,
+      "step": 1956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092172,
+      "loss": 1.3651,
+      "step": 1957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092168,
+      "loss": 1.4731,
+      "step": 1958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092164,
+      "loss": 1.5,
+      "step": 1959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009216,
+      "loss": 1.5153,
+      "step": 1960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009215600000000001,
+      "loss": 1.4934,
+      "step": 1961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092152,
+      "loss": 1.4067,
+      "step": 1962
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092148,
+      "loss": 1.4493,
+      "step": 1963
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092144,
+      "loss": 1.4209,
+      "step": 1964
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009214000000000001,
+      "loss": 1.4993,
+      "step": 1965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092136,
+      "loss": 1.4133,
+      "step": 1966
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092132,
+      "loss": 1.3629,
+      "step": 1967
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092128,
+      "loss": 1.4071,
+      "step": 1968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009212399999999999,
+      "loss": 1.4242,
+      "step": 1969
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009212000000000001,
+      "loss": 1.527,
+      "step": 1970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092116,
+      "loss": 1.4166,
+      "step": 1971
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009211200000000001,
+      "loss": 1.4253,
+      "step": 1972
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092108,
+      "loss": 1.5168,
+      "step": 1973
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092104,
+      "loss": 1.4605,
+      "step": 1974
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000921,
+      "loss": 1.4807,
+      "step": 1975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009209600000000001,
+      "loss": 1.3138,
+      "step": 1976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092092,
+      "loss": 1.446,
+      "step": 1977
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092088,
+      "loss": 1.3575,
+      "step": 1978
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092084,
+      "loss": 1.3984,
+      "step": 1979
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009207999999999999,
+      "loss": 1.3111,
+      "step": 1980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009207600000000001,
+      "loss": 1.493,
+      "step": 1981
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092072,
+      "loss": 1.4104,
+      "step": 1982
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009206800000000001,
+      "loss": 1.4748,
+      "step": 1983
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092064,
+      "loss": 1.5076,
+      "step": 1984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009206,
+      "loss": 1.4651,
+      "step": 1985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009205600000000001,
+      "loss": 1.4604,
+      "step": 1986
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092052,
+      "loss": 1.5302,
+      "step": 1987
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092048,
+      "loss": 1.4993,
+      "step": 1988
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092044,
+      "loss": 1.4256,
+      "step": 1989
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009204,
+      "loss": 1.4368,
+      "step": 1990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092036,
+      "loss": 1.4142,
+      "step": 1991
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009203200000000001,
+      "loss": 1.3511,
+      "step": 1992
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092028,
+      "loss": 1.3587,
+      "step": 1993
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092024,
+      "loss": 1.4564,
+      "step": 1994
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009202,
+      "loss": 1.4609,
+      "step": 1995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009201599999999999,
+      "loss": 1.4192,
+      "step": 1996
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009201200000000001,
+      "loss": 1.4719,
+      "step": 1997
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092008,
+      "loss": 1.5178,
+      "step": 1998
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092004,
+      "loss": 1.469,
+      "step": 1999
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00092,
+      "loss": 1.4303,
+      "step": 2000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091996,
+      "loss": 1.3028,
+      "step": 2001
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091992,
+      "loss": 1.4264,
+      "step": 2002
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009198800000000001,
+      "loss": 1.3825,
+      "step": 2003
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091984,
+      "loss": 1.4022,
+      "step": 2004
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009198,
+      "loss": 1.3976,
+      "step": 2005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091976,
+      "loss": 1.4437,
+      "step": 2006
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091972,
+      "loss": 1.3908,
+      "step": 2007
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009196800000000001,
+      "loss": 1.4046,
+      "step": 2008
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091964,
+      "loss": 1.4427,
+      "step": 2009
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009196,
+      "loss": 1.42,
+      "step": 2010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091956,
+      "loss": 1.4291,
+      "step": 2011
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009195200000000001,
+      "loss": 1.4983,
+      "step": 2012
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091948,
+      "loss": 1.4763,
+      "step": 2013
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009194400000000001,
+      "loss": 1.3945,
+      "step": 2014
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009194,
+      "loss": 1.4404,
+      "step": 2015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009193599999999999,
+      "loss": 1.4964,
+      "step": 2016
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009193200000000001,
+      "loss": 1.4498,
+      "step": 2017
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091928,
+      "loss": 1.3441,
+      "step": 2018
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091924,
+      "loss": 1.4869,
+      "step": 2019
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009192,
+      "loss": 1.3753,
+      "step": 2020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091916,
+      "loss": 1.5213,
+      "step": 2021
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091912,
+      "loss": 1.5344,
+      "step": 2022
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009190800000000001,
+      "loss": 1.3347,
+      "step": 2023
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091904,
+      "loss": 1.4151,
+      "step": 2024
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009190000000000001,
+      "loss": 1.5077,
+      "step": 2025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091896,
+      "loss": 1.4646,
+      "step": 2026
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009189199999999999,
+      "loss": 1.489,
+      "step": 2027
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009188800000000001,
+      "loss": 1.3593,
+      "step": 2028
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091884,
+      "loss": 1.4941,
+      "step": 2029
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009188,
+      "loss": 1.3348,
+      "step": 2030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091876,
+      "loss": 1.4459,
+      "step": 2031
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091872,
+      "loss": 1.417,
+      "step": 2032
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009186800000000001,
+      "loss": 1.4706,
+      "step": 2033
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009186400000000001,
+      "loss": 1.4158,
+      "step": 2034
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009186,
+      "loss": 1.368,
+      "step": 2035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091856,
+      "loss": 1.4081,
+      "step": 2036
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091852,
+      "loss": 1.4317,
+      "step": 2037
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091848,
+      "loss": 1.3803,
+      "step": 2038
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009184400000000001,
+      "loss": 1.5028,
+      "step": 2039
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009184,
+      "loss": 1.3704,
+      "step": 2040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091836,
+      "loss": 1.4653,
+      "step": 2041
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091832,
+      "loss": 1.432,
+      "step": 2042
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091828,
+      "loss": 1.4356,
+      "step": 2043
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091824,
+      "loss": 1.3133,
+      "step": 2044
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009182,
+      "loss": 1.5548,
+      "step": 2045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091816,
+      "loss": 1.4359,
+      "step": 2046
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091812,
+      "loss": 1.474,
+      "step": 2047
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091808,
+      "loss": 1.3619,
+      "step": 2048
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091804,
+      "loss": 1.3805,
+      "step": 2049
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009180000000000001,
+      "loss": 1.3808,
+      "step": 2050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091796,
+      "loss": 1.4882,
+      "step": 2051
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091792,
+      "loss": 1.4375,
+      "step": 2052
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091788,
+      "loss": 1.4921,
+      "step": 2053
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009178400000000001,
+      "loss": 1.5066,
+      "step": 2054
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009178,
+      "loss": 1.3251,
+      "step": 2055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091776,
+      "loss": 1.4653,
+      "step": 2056
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091772,
+      "loss": 1.367,
+      "step": 2057
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091768,
+      "loss": 1.3525,
+      "step": 2058
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009176400000000001,
+      "loss": 1.4392,
+      "step": 2059
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009176,
+      "loss": 1.4801,
+      "step": 2060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009175600000000001,
+      "loss": 1.468,
+      "step": 2061
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091752,
+      "loss": 1.4623,
+      "step": 2062
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091748,
+      "loss": 1.4065,
+      "step": 2063
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091744,
+      "loss": 1.3904,
+      "step": 2064
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009174,
+      "loss": 1.4304,
+      "step": 2065
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091736,
+      "loss": 1.4946,
+      "step": 2066
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091732,
+      "loss": 1.4477,
+      "step": 2067
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091728,
+      "loss": 1.4748,
+      "step": 2068
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091724,
+      "loss": 1.4538,
+      "step": 2069
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009172000000000001,
+      "loss": 1.4295,
+      "step": 2070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091716,
+      "loss": 1.4094,
+      "step": 2071
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009171200000000001,
+      "loss": 1.4012,
+      "step": 2072
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091708,
+      "loss": 1.437,
+      "step": 2073
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009170399999999999,
+      "loss": 1.4003,
+      "step": 2074
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009170000000000001,
+      "loss": 1.3649,
+      "step": 2075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091696,
+      "loss": 1.4254,
+      "step": 2076
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091692,
+      "loss": 1.429,
+      "step": 2077
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091688,
+      "loss": 1.4911,
+      "step": 2078
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091684,
+      "loss": 1.4599,
+      "step": 2079
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009168,
+      "loss": 1.4671,
+      "step": 2080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009167600000000001,
+      "loss": 1.4578,
+      "step": 2081
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091672,
+      "loss": 1.4193,
+      "step": 2082
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009166800000000001,
+      "loss": 1.3414,
+      "step": 2083
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091664,
+      "loss": 1.3145,
+      "step": 2084
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009165999999999999,
+      "loss": 1.4223,
+      "step": 2085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009165600000000001,
+      "loss": 1.3727,
+      "step": 2086
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091652,
+      "loss": 1.3439,
+      "step": 2087
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091648,
+      "loss": 1.4313,
+      "step": 2088
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091644,
+      "loss": 1.4453,
+      "step": 2089
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009164,
+      "loss": 1.4303,
+      "step": 2090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091636,
+      "loss": 1.375,
+      "step": 2091
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009163200000000001,
+      "loss": 1.4932,
+      "step": 2092
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091628,
+      "loss": 1.4006,
+      "step": 2093
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091624,
+      "loss": 1.4487,
+      "step": 2094
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009162,
+      "loss": 1.404,
+      "step": 2095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091616,
+      "loss": 1.4251,
+      "step": 2096
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009161200000000001,
+      "loss": 1.5113,
+      "step": 2097
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091608,
+      "loss": 1.3545,
+      "step": 2098
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091604,
+      "loss": 1.4648,
+      "step": 2099
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000916,
+      "loss": 1.4861,
+      "step": 2100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009159600000000001,
+      "loss": 1.454,
+      "step": 2101
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091592,
+      "loss": 1.4118,
+      "step": 2102
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091588,
+      "loss": 1.4166,
+      "step": 2103
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091584,
+      "loss": 1.4016,
+      "step": 2104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009157999999999999,
+      "loss": 1.5182,
+      "step": 2105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009157600000000001,
+      "loss": 1.3579,
+      "step": 2106
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091572,
+      "loss": 1.4348,
+      "step": 2107
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009156800000000001,
+      "loss": 1.4082,
+      "step": 2108
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091564,
+      "loss": 1.4581,
+      "step": 2109
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009156,
+      "loss": 1.3677,
+      "step": 2110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091556,
+      "loss": 1.3423,
+      "step": 2111
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009155200000000001,
+      "loss": 1.4297,
+      "step": 2112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091548,
+      "loss": 1.4422,
+      "step": 2113
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091544,
+      "loss": 1.3897,
+      "step": 2114
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009154,
+      "loss": 1.4403,
+      "step": 2115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009153599999999999,
+      "loss": 1.3797,
+      "step": 2116
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009153200000000001,
+      "loss": 1.3729,
+      "step": 2117
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091528,
+      "loss": 1.5307,
+      "step": 2118
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009152400000000001,
+      "loss": 1.4217,
+      "step": 2119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009152,
+      "loss": 1.4301,
+      "step": 2120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091516,
+      "loss": 1.3475,
+      "step": 2121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009151200000000001,
+      "loss": 1.4288,
+      "step": 2122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091508,
+      "loss": 1.4498,
+      "step": 2123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091504,
+      "loss": 1.3702,
+      "step": 2124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000915,
+      "loss": 1.3547,
+      "step": 2125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091496,
+      "loss": 1.4579,
+      "step": 2126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091492,
+      "loss": 1.377,
+      "step": 2127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009148800000000001,
+      "loss": 1.4459,
+      "step": 2128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091484,
+      "loss": 1.3686,
+      "step": 2129
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009148,
+      "loss": 1.4511,
+      "step": 2130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091476,
+      "loss": 1.3288,
+      "step": 2131
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009147199999999999,
+      "loss": 1.4565,
+      "step": 2132
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009146800000000001,
+      "loss": 1.3677,
+      "step": 2133
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091464,
+      "loss": 1.409,
+      "step": 2134
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009146,
+      "loss": 1.3664,
+      "step": 2135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091456,
+      "loss": 1.4438,
+      "step": 2136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091452,
+      "loss": 1.3216,
+      "step": 2137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091448,
+      "loss": 1.3828,
+      "step": 2138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009144400000000001,
+      "loss": 1.3994,
+      "step": 2139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009144,
+      "loss": 1.4865,
+      "step": 2140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091436,
+      "loss": 1.4814,
+      "step": 2141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091432,
+      "loss": 1.3826,
+      "step": 2142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091428,
+      "loss": 1.2808,
+      "step": 2143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009142400000000001,
+      "loss": 1.3974,
+      "step": 2144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009142,
+      "loss": 1.405,
+      "step": 2145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091416,
+      "loss": 1.4882,
+      "step": 2146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091412,
+      "loss": 1.3944,
+      "step": 2147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009140800000000001,
+      "loss": 1.4568,
+      "step": 2148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091404,
+      "loss": 1.3385,
+      "step": 2149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009140000000000001,
+      "loss": 1.3859,
+      "step": 2150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091396,
+      "loss": 1.3718,
+      "step": 2151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0009139199999999999,
+      "loss": 1.4816,
+      "step": 2152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00091388,
+      "loss": 1.3679,
+      "step": 2153
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091384,
+      "loss": 1.4757,
+      "step": 2154
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009138,
+      "loss": 1.4377,
+      "step": 2155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091376,
+      "loss": 1.4253,
+      "step": 2156
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091372,
+      "loss": 1.3892,
+      "step": 2157
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091368,
+      "loss": 1.3515,
+      "step": 2158
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009136400000000001,
+      "loss": 1.2599,
+      "step": 2159
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009136,
+      "loss": 1.3213,
+      "step": 2160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009135600000000001,
+      "loss": 1.3524,
+      "step": 2161
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091352,
+      "loss": 1.3893,
+      "step": 2162
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009134799999999999,
+      "loss": 1.4094,
+      "step": 2163
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009134400000000001,
+      "loss": 1.3771,
+      "step": 2164
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009134,
+      "loss": 1.4681,
+      "step": 2165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091336,
+      "loss": 1.482,
+      "step": 2166
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091332,
+      "loss": 1.4351,
+      "step": 2167
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091328,
+      "loss": 1.3874,
+      "step": 2168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009132400000000001,
+      "loss": 1.4966,
+      "step": 2169
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009132000000000001,
+      "loss": 1.4021,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091316,
+      "loss": 1.4686,
+      "step": 2171
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091312,
+      "loss": 1.3838,
+      "step": 2172
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091308,
+      "loss": 1.4481,
+      "step": 2173
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009130399999999999,
+      "loss": 1.4134,
+      "step": 2174
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009130000000000001,
+      "loss": 1.4614,
+      "step": 2175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091296,
+      "loss": 1.3984,
+      "step": 2176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091292,
+      "loss": 1.4633,
+      "step": 2177
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091288,
+      "loss": 1.3754,
+      "step": 2178
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091284,
+      "loss": 1.4217,
+      "step": 2179
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009128,
+      "loss": 1.3871,
+      "step": 2180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091276,
+      "loss": 1.3634,
+      "step": 2181
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091272,
+      "loss": 1.4896,
+      "step": 2182
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091268,
+      "loss": 1.5189,
+      "step": 2183
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091264,
+      "loss": 1.4187,
+      "step": 2184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009126,
+      "loss": 1.4161,
+      "step": 2185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009125600000000001,
+      "loss": 1.3524,
+      "step": 2186
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091252,
+      "loss": 1.3942,
+      "step": 2187
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091248,
+      "loss": 1.4216,
+      "step": 2188
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091244,
+      "loss": 1.4966,
+      "step": 2189
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009124000000000001,
+      "loss": 1.3812,
+      "step": 2190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091236,
+      "loss": 1.4219,
+      "step": 2191
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091232,
+      "loss": 1.3949,
+      "step": 2192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091228,
+      "loss": 1.4087,
+      "step": 2193
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091224,
+      "loss": 1.4488,
+      "step": 2194
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009122000000000001,
+      "loss": 1.395,
+      "step": 2195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091216,
+      "loss": 1.3463,
+      "step": 2196
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009121200000000001,
+      "loss": 1.4165,
+      "step": 2197
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091208,
+      "loss": 1.4571,
+      "step": 2198
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091204,
+      "loss": 1.4263,
+      "step": 2199
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000912,
+      "loss": 1.3413,
+      "step": 2200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091196,
+      "loss": 1.4269,
+      "step": 2201
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091192,
+      "loss": 1.3406,
+      "step": 2202
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091188,
+      "loss": 1.4292,
+      "step": 2203
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091184,
+      "loss": 1.4138,
+      "step": 2204
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009118,
+      "loss": 1.4234,
+      "step": 2205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009117600000000001,
+      "loss": 1.5033,
+      "step": 2206
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091172,
+      "loss": 1.4237,
+      "step": 2207
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009116800000000001,
+      "loss": 1.3936,
+      "step": 2208
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091164,
+      "loss": 1.3919,
+      "step": 2209
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009115999999999999,
+      "loss": 1.4401,
+      "step": 2210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009115600000000001,
+      "loss": 1.3904,
+      "step": 2211
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091152,
+      "loss": 1.4752,
+      "step": 2212
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091148,
+      "loss": 1.4066,
+      "step": 2213
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091144,
+      "loss": 1.4674,
+      "step": 2214
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009114,
+      "loss": 1.4259,
+      "step": 2215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091136,
+      "loss": 1.3239,
+      "step": 2216
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009113200000000001,
+      "loss": 1.336,
+      "step": 2217
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091128,
+      "loss": 1.3283,
+      "step": 2218
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009112400000000001,
+      "loss": 1.4088,
+      "step": 2219
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009112,
+      "loss": 1.4024,
+      "step": 2220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009111599999999999,
+      "loss": 1.4692,
+      "step": 2221
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009111200000000001,
+      "loss": 1.3469,
+      "step": 2222
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091108,
+      "loss": 1.4938,
+      "step": 2223
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091104,
+      "loss": 1.3376,
+      "step": 2224
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000911,
+      "loss": 1.4815,
+      "step": 2225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091096,
+      "loss": 1.4023,
+      "step": 2226
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091092,
+      "loss": 1.4983,
+      "step": 2227
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009108800000000001,
+      "loss": 1.3284,
+      "step": 2228
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091084,
+      "loss": 1.3921,
+      "step": 2229
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009108,
+      "loss": 1.4119,
+      "step": 2230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091076,
+      "loss": 1.4804,
+      "step": 2231
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091072,
+      "loss": 1.414,
+      "step": 2232
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009106800000000001,
+      "loss": 1.4147,
+      "step": 2233
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091064,
+      "loss": 1.4275,
+      "step": 2234
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009106,
+      "loss": 1.2782,
+      "step": 2235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091056,
+      "loss": 1.5062,
+      "step": 2236
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009105200000000001,
+      "loss": 1.4691,
+      "step": 2237
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091048,
+      "loss": 1.3331,
+      "step": 2238
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009104400000000001,
+      "loss": 1.4946,
+      "step": 2239
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009104,
+      "loss": 1.5141,
+      "step": 2240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009103599999999999,
+      "loss": 1.3072,
+      "step": 2241
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091032,
+      "loss": 1.4313,
+      "step": 2242
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091028,
+      "loss": 1.4701,
+      "step": 2243
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009102400000000001,
+      "loss": 1.4055,
+      "step": 2244
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009102,
+      "loss": 1.4169,
+      "step": 2245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091016,
+      "loss": 1.4074,
+      "step": 2246
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091012,
+      "loss": 1.3671,
+      "step": 2247
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009100800000000001,
+      "loss": 1.4668,
+      "step": 2248
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091004,
+      "loss": 1.5311,
+      "step": 2249
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00091,
+      "loss": 1.4069,
+      "step": 2250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090996,
+      "loss": 1.3247,
+      "step": 2251
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009099199999999999,
+      "loss": 1.3602,
+      "step": 2252
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009098800000000001,
+      "loss": 1.4386,
+      "step": 2253
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090984,
+      "loss": 1.3698,
+      "step": 2254
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009098000000000001,
+      "loss": 1.3712,
+      "step": 2255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090976,
+      "loss": 1.3684,
+      "step": 2256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090972,
+      "loss": 1.426,
+      "step": 2257
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009096800000000001,
+      "loss": 1.4865,
+      "step": 2258
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090964,
+      "loss": 1.3435,
+      "step": 2259
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009096,
+      "loss": 1.3707,
+      "step": 2260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090956,
+      "loss": 1.4502,
+      "step": 2261
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090952,
+      "loss": 1.4942,
+      "step": 2262
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009094799999999999,
+      "loss": 1.4154,
+      "step": 2263
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009094400000000001,
+      "loss": 1.454,
+      "step": 2264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009094,
+      "loss": 1.6158,
+      "step": 2265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090936,
+      "loss": 1.3367,
+      "step": 2266
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090932,
+      "loss": 1.3204,
+      "step": 2267
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090928,
+      "loss": 1.3923,
+      "step": 2268
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009092400000000001,
+      "loss": 1.3457,
+      "step": 2269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009092,
+      "loss": 1.3861,
+      "step": 2270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090916,
+      "loss": 1.3903,
+      "step": 2271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090912,
+      "loss": 1.3557,
+      "step": 2272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090908,
+      "loss": 1.4553,
+      "step": 2273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090904,
+      "loss": 1.385,
+      "step": 2274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009090000000000001,
+      "loss": 1.3772,
+      "step": 2275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090896,
+      "loss": 1.3577,
+      "step": 2276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090892,
+      "loss": 1.3379,
+      "step": 2277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090888,
+      "loss": 1.494,
+      "step": 2278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090884,
+      "loss": 1.3898,
+      "step": 2279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009088000000000001,
+      "loss": 1.3822,
+      "step": 2280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090876,
+      "loss": 1.4052,
+      "step": 2281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090872,
+      "loss": 1.335,
+      "step": 2282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090868,
+      "loss": 1.4758,
+      "step": 2283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009086400000000001,
+      "loss": 1.4454,
+      "step": 2284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009086,
+      "loss": 1.4547,
+      "step": 2285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009085600000000001,
+      "loss": 1.4464,
+      "step": 2286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090852,
+      "loss": 1.3575,
+      "step": 2287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009084799999999999,
+      "loss": 1.3331,
+      "step": 2288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090844,
+      "loss": 1.4667,
+      "step": 2289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009084,
+      "loss": 1.3754,
+      "step": 2290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090836,
+      "loss": 1.3817,
+      "step": 2291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090832,
+      "loss": 1.439,
+      "step": 2292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090828,
+      "loss": 1.4085,
+      "step": 2293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090824,
+      "loss": 1.3845,
+      "step": 2294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009082000000000001,
+      "loss": 1.5131,
+      "step": 2295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090816,
+      "loss": 1.469,
+      "step": 2296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009081200000000001,
+      "loss": 1.449,
+      "step": 2297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090808,
+      "loss": 1.4211,
+      "step": 2298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009080399999999999,
+      "loss": 1.4692,
+      "step": 2299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009080000000000001,
+      "loss": 1.2622,
+      "step": 2300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090796,
+      "loss": 1.3433,
+      "step": 2301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090792,
+      "loss": 1.38,
+      "step": 2302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090788,
+      "loss": 1.3743,
+      "step": 2303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090784,
+      "loss": 1.3223,
+      "step": 2304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009078000000000001,
+      "loss": 1.3836,
+      "step": 2305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009077600000000001,
+      "loss": 1.4501,
+      "step": 2306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090772,
+      "loss": 1.3703,
+      "step": 2307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090768,
+      "loss": 1.4001,
+      "step": 2308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090764,
+      "loss": 1.4366,
+      "step": 2309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009075999999999999,
+      "loss": 1.3064,
+      "step": 2310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009075600000000001,
+      "loss": 1.3452,
+      "step": 2311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090752,
+      "loss": 1.4205,
+      "step": 2312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090748,
+      "loss": 1.3589,
+      "step": 2313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090744,
+      "loss": 1.3394,
+      "step": 2314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009074,
+      "loss": 1.3783,
+      "step": 2315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009073600000000001,
+      "loss": 1.4155,
+      "step": 2316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009073200000000001,
+      "loss": 1.3503,
+      "step": 2317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090728,
+      "loss": 1.3684,
+      "step": 2318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090724,
+      "loss": 1.3935,
+      "step": 2319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009072,
+      "loss": 1.4597,
+      "step": 2320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090716,
+      "loss": 1.4166,
+      "step": 2321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009071200000000001,
+      "loss": 1.4227,
+      "step": 2322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090708,
+      "loss": 1.4067,
+      "step": 2323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090704,
+      "loss": 1.356,
+      "step": 2324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000907,
+      "loss": 1.4034,
+      "step": 2325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009069600000000001,
+      "loss": 1.3235,
+      "step": 2326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090692,
+      "loss": 1.4427,
+      "step": 2327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090688,
+      "loss": 1.4268,
+      "step": 2328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090684,
+      "loss": 1.351,
+      "step": 2329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009068,
+      "loss": 1.3499,
+      "step": 2330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090676,
+      "loss": 1.4633,
+      "step": 2331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090672,
+      "loss": 1.4338,
+      "step": 2332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009066800000000001,
+      "loss": 1.4697,
+      "step": 2333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090664,
+      "loss": 1.5296,
+      "step": 2334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009066,
+      "loss": 1.3864,
+      "step": 2335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090656,
+      "loss": 1.4628,
+      "step": 2336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090652,
+      "loss": 1.382,
+      "step": 2337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090648,
+      "loss": 1.4595,
+      "step": 2338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090644,
+      "loss": 1.4265,
+      "step": 2339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009064,
+      "loss": 1.3933,
+      "step": 2340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090636,
+      "loss": 1.448,
+      "step": 2341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009063200000000001,
+      "loss": 1.4561,
+      "step": 2342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090628,
+      "loss": 1.4707,
+      "step": 2343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009062400000000001,
+      "loss": 1.44,
+      "step": 2344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009062,
+      "loss": 1.347,
+      "step": 2345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090616,
+      "loss": 1.4297,
+      "step": 2346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009061200000000001,
+      "loss": 1.3737,
+      "step": 2347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090608,
+      "loss": 1.4496,
+      "step": 2348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090604,
+      "loss": 1.3686,
+      "step": 2349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000906,
+      "loss": 1.3303,
+      "step": 2350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090596,
+      "loss": 1.464,
+      "step": 2351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090592,
+      "loss": 1.3918,
+      "step": 2352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009058800000000001,
+      "loss": 1.4027,
+      "step": 2353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090584,
+      "loss": 1.441,
+      "step": 2354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009058000000000001,
+      "loss": 1.2615,
+      "step": 2355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090576,
+      "loss": 1.3981,
+      "step": 2356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009057199999999999,
+      "loss": 1.5002,
+      "step": 2357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009056800000000001,
+      "loss": 1.3339,
+      "step": 2358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090564,
+      "loss": 1.3191,
+      "step": 2359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009056,
+      "loss": 1.395,
+      "step": 2360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090556,
+      "loss": 1.4879,
+      "step": 2361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090552,
+      "loss": 1.3627,
+      "step": 2362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090548,
+      "loss": 1.3706,
+      "step": 2363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009054400000000001,
+      "loss": 1.4129,
+      "step": 2364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009054,
+      "loss": 1.4296,
+      "step": 2365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090536,
+      "loss": 1.3924,
+      "step": 2366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090532,
+      "loss": 1.3869,
+      "step": 2367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090528,
+      "loss": 1.5071,
+      "step": 2368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009052400000000001,
+      "loss": 1.3814,
+      "step": 2369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009052,
+      "loss": 1.3985,
+      "step": 2370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090516,
+      "loss": 1.3602,
+      "step": 2371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090512,
+      "loss": 1.358,
+      "step": 2372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009050800000000001,
+      "loss": 1.3691,
+      "step": 2373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090504,
+      "loss": 1.3557,
+      "step": 2374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009050000000000001,
+      "loss": 1.419,
+      "step": 2375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090496,
+      "loss": 1.3079,
+      "step": 2376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009049199999999999,
+      "loss": 1.4179,
+      "step": 2377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090488,
+      "loss": 1.4481,
+      "step": 2378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090484,
+      "loss": 1.4672,
+      "step": 2379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009048000000000001,
+      "loss": 1.4086,
+      "step": 2380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090476,
+      "loss": 1.5124,
+      "step": 2381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090472,
+      "loss": 1.452,
+      "step": 2382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090468,
+      "loss": 1.2976,
+      "step": 2383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009046400000000001,
+      "loss": 1.3854,
+      "step": 2384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009046,
+      "loss": 1.3579,
+      "step": 2385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090456,
+      "loss": 1.4149,
+      "step": 2386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090452,
+      "loss": 1.4404,
+      "step": 2387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009044799999999999,
+      "loss": 1.4098,
+      "step": 2388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009044400000000001,
+      "loss": 1.4535,
+      "step": 2389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009044,
+      "loss": 1.4565,
+      "step": 2390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009043600000000001,
+      "loss": 1.4856,
+      "step": 2391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090432,
+      "loss": 1.3615,
+      "step": 2392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090428,
+      "loss": 1.4039,
+      "step": 2393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009042400000000001,
+      "loss": 1.4146,
+      "step": 2394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009042,
+      "loss": 1.3432,
+      "step": 2395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090416,
+      "loss": 1.3789,
+      "step": 2396
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090412,
+      "loss": 1.382,
+      "step": 2397
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090408,
+      "loss": 1.3308,
+      "step": 2398
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009040399999999999,
+      "loss": 1.4371,
+      "step": 2399
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009040000000000001,
+      "loss": 1.4033,
+      "step": 2400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090396,
+      "loss": 1.3774,
+      "step": 2401
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090392,
+      "loss": 1.4655,
+      "step": 2402
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090388,
+      "loss": 1.3117,
+      "step": 2403
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090384,
+      "loss": 1.4403,
+      "step": 2404
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0009038000000000001,
+      "loss": 1.391,
+      "step": 2405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090376,
+      "loss": 1.3802,
+      "step": 2406
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00090372,
+      "loss": 1.359,
+      "step": 2407
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090368,
+      "loss": 1.3552,
+      "step": 2408
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090364,
+      "loss": 1.3138,
+      "step": 2409
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009036,
+      "loss": 1.4411,
+      "step": 2410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009035600000000001,
+      "loss": 1.3491,
+      "step": 2411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090352,
+      "loss": 1.3913,
+      "step": 2412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090348,
+      "loss": 1.371,
+      "step": 2413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090344,
+      "loss": 1.4471,
+      "step": 2414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009034,
+      "loss": 1.4664,
+      "step": 2415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009033600000000001,
+      "loss": 1.3354,
+      "step": 2416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090332,
+      "loss": 1.3579,
+      "step": 2417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090328,
+      "loss": 1.4683,
+      "step": 2418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090324,
+      "loss": 1.4453,
+      "step": 2419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009032,
+      "loss": 1.3815,
+      "step": 2420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090316,
+      "loss": 1.3975,
+      "step": 2421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009031200000000001,
+      "loss": 1.4397,
+      "step": 2422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090308,
+      "loss": 1.3165,
+      "step": 2423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009030399999999999,
+      "loss": 1.3656,
+      "step": 2424
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000903,
+      "loss": 1.3612,
+      "step": 2425
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090296,
+      "loss": 1.4482,
+      "step": 2426
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090292,
+      "loss": 1.3646,
+      "step": 2427
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090288,
+      "loss": 1.3778,
+      "step": 2428
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090284,
+      "loss": 1.4246,
+      "step": 2429
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009028,
+      "loss": 1.3048,
+      "step": 2430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009027600000000001,
+      "loss": 1.3926,
+      "step": 2431
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090272,
+      "loss": 1.3531,
+      "step": 2432
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009026800000000001,
+      "loss": 1.3996,
+      "step": 2433
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090264,
+      "loss": 1.3839,
+      "step": 2434
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009025999999999999,
+      "loss": 1.4263,
+      "step": 2435
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009025600000000001,
+      "loss": 1.3462,
+      "step": 2436
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090252,
+      "loss": 1.4429,
+      "step": 2437
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090248,
+      "loss": 1.3551,
+      "step": 2438
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090244,
+      "loss": 1.3175,
+      "step": 2439
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009024,
+      "loss": 1.4173,
+      "step": 2440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009023600000000001,
+      "loss": 1.3635,
+      "step": 2441
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009023200000000001,
+      "loss": 1.4135,
+      "step": 2442
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090228,
+      "loss": 1.4108,
+      "step": 2443
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090224,
+      "loss": 1.3146,
+      "step": 2444
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009022,
+      "loss": 1.4049,
+      "step": 2445
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009021599999999999,
+      "loss": 1.3602,
+      "step": 2446
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009021200000000001,
+      "loss": 1.4457,
+      "step": 2447
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090208,
+      "loss": 1.3745,
+      "step": 2448
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090204,
+      "loss": 1.3922,
+      "step": 2449
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000902,
+      "loss": 1.4318,
+      "step": 2450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090196,
+      "loss": 1.3361,
+      "step": 2451
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009019200000000001,
+      "loss": 1.3785,
+      "step": 2452
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009018800000000001,
+      "loss": 1.361,
+      "step": 2453
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090184,
+      "loss": 1.3344,
+      "step": 2454
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009018,
+      "loss": 1.4164,
+      "step": 2455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090176,
+      "loss": 1.3129,
+      "step": 2456
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090172,
+      "loss": 1.4104,
+      "step": 2457
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009016800000000001,
+      "loss": 1.4407,
+      "step": 2458
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090164,
+      "loss": 1.4142,
+      "step": 2459
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009016,
+      "loss": 1.3822,
+      "step": 2460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090156,
+      "loss": 1.2709,
+      "step": 2461
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009015200000000001,
+      "loss": 1.3681,
+      "step": 2462
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090148,
+      "loss": 1.442,
+      "step": 2463
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090144,
+      "loss": 1.4453,
+      "step": 2464
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009014,
+      "loss": 1.3808,
+      "step": 2465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090136,
+      "loss": 1.3743,
+      "step": 2466
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090132,
+      "loss": 1.3321,
+      "step": 2467
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090128,
+      "loss": 1.352,
+      "step": 2468
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009012400000000001,
+      "loss": 1.3953,
+      "step": 2469
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009012,
+      "loss": 1.3907,
+      "step": 2470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090116,
+      "loss": 1.3437,
+      "step": 2471
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090112,
+      "loss": 1.4397,
+      "step": 2472
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090108,
+      "loss": 1.4384,
+      "step": 2473
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090104,
+      "loss": 1.3782,
+      "step": 2474
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000901,
+      "loss": 1.2893,
+      "step": 2475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090096,
+      "loss": 1.4081,
+      "step": 2476
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090092,
+      "loss": 1.3215,
+      "step": 2477
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009008800000000001,
+      "loss": 1.3392,
+      "step": 2478
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090084,
+      "loss": 1.4475,
+      "step": 2479
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009008000000000001,
+      "loss": 1.4134,
+      "step": 2480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090076,
+      "loss": 1.2957,
+      "step": 2481
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090072,
+      "loss": 1.3878,
+      "step": 2482
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009006800000000001,
+      "loss": 1.4861,
+      "step": 2483
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090064,
+      "loss": 1.347,
+      "step": 2484
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009006,
+      "loss": 1.5297,
+      "step": 2485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090056,
+      "loss": 1.4421,
+      "step": 2486
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090052,
+      "loss": 1.263,
+      "step": 2487
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009004799999999999,
+      "loss": 1.2433,
+      "step": 2488
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009004400000000001,
+      "loss": 1.3211,
+      "step": 2489
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009004,
+      "loss": 1.3736,
+      "step": 2490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009003600000000001,
+      "loss": 1.3204,
+      "step": 2491
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090032,
+      "loss": 1.4121,
+      "step": 2492
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009002799999999999,
+      "loss": 1.4057,
+      "step": 2493
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009002400000000001,
+      "loss": 1.3442,
+      "step": 2494
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009002,
+      "loss": 1.3073,
+      "step": 2495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090016,
+      "loss": 1.4359,
+      "step": 2496
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090012,
+      "loss": 1.3693,
+      "step": 2497
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090008,
+      "loss": 1.259,
+      "step": 2498
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00090004,
+      "loss": 1.3405,
+      "step": 2499
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0009000000000000001,
+      "loss": 1.356,
+      "step": 2500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089996,
+      "loss": 1.3718,
+      "step": 2501
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008999200000000001,
+      "loss": 1.4311,
+      "step": 2502
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089988,
+      "loss": 1.393,
+      "step": 2503
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089984,
+      "loss": 1.4583,
+      "step": 2504
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008998000000000001,
+      "loss": 1.3205,
+      "step": 2505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089976,
+      "loss": 1.4281,
+      "step": 2506
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089972,
+      "loss": 1.3769,
+      "step": 2507
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089968,
+      "loss": 1.3589,
+      "step": 2508
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089964,
+      "loss": 1.5276,
+      "step": 2509
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008996,
+      "loss": 1.4625,
+      "step": 2510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008995600000000001,
+      "loss": 1.337,
+      "step": 2511
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089952,
+      "loss": 1.3689,
+      "step": 2512
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008994799999999999,
+      "loss": 1.3939,
+      "step": 2513
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089944,
+      "loss": 1.4258,
+      "step": 2514
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008994,
+      "loss": 1.4763,
+      "step": 2515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008993600000000001,
+      "loss": 1.4751,
+      "step": 2516
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089932,
+      "loss": 1.3282,
+      "step": 2517
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089928,
+      "loss": 1.3517,
+      "step": 2518
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089924,
+      "loss": 1.3856,
+      "step": 2519
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008992000000000001,
+      "loss": 1.3557,
+      "step": 2520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089916,
+      "loss": 1.3533,
+      "step": 2521
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089912,
+      "loss": 1.3251,
+      "step": 2522
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089908,
+      "loss": 1.3252,
+      "step": 2523
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008990399999999999,
+      "loss": 1.4456,
+      "step": 2524
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008990000000000001,
+      "loss": 1.4087,
+      "step": 2525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089896,
+      "loss": 1.4083,
+      "step": 2526
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008989200000000001,
+      "loss": 1.3406,
+      "step": 2527
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089888,
+      "loss": 1.3794,
+      "step": 2528
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089884,
+      "loss": 1.3621,
+      "step": 2529
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008988000000000001,
+      "loss": 1.4076,
+      "step": 2530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008987600000000001,
+      "loss": 1.4119,
+      "step": 2531
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089872,
+      "loss": 1.2932,
+      "step": 2532
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089868,
+      "loss": 1.4531,
+      "step": 2533
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089864,
+      "loss": 1.3849,
+      "step": 2534
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008985999999999999,
+      "loss": 1.4242,
+      "step": 2535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008985600000000001,
+      "loss": 1.4205,
+      "step": 2536
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089852,
+      "loss": 1.3348,
+      "step": 2537
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089848,
+      "loss": 1.3339,
+      "step": 2538
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089844,
+      "loss": 1.3767,
+      "step": 2539
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008984,
+      "loss": 1.3393,
+      "step": 2540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008983600000000001,
+      "loss": 1.3976,
+      "step": 2541
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089832,
+      "loss": 1.3523,
+      "step": 2542
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089828,
+      "loss": 1.4927,
+      "step": 2543
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089824,
+      "loss": 1.3096,
+      "step": 2544
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008982,
+      "loss": 1.3755,
+      "step": 2545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089816,
+      "loss": 1.3465,
+      "step": 2546
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008981200000000001,
+      "loss": 1.42,
+      "step": 2547
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089808,
+      "loss": 1.4187,
+      "step": 2548
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089804,
+      "loss": 1.4228,
+      "step": 2549
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000898,
+      "loss": 1.3768,
+      "step": 2550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089796,
+      "loss": 1.377,
+      "step": 2551
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008979200000000001,
+      "loss": 1.333,
+      "step": 2552
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089788,
+      "loss": 1.3692,
+      "step": 2553
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089784,
+      "loss": 1.368,
+      "step": 2554
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008978,
+      "loss": 1.3788,
+      "step": 2555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089776,
+      "loss": 1.4307,
+      "step": 2556
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089772,
+      "loss": 1.4283,
+      "step": 2557
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008976800000000001,
+      "loss": 1.3971,
+      "step": 2558
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089764,
+      "loss": 1.4227,
+      "step": 2559
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008976,
+      "loss": 1.3834,
+      "step": 2560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089756,
+      "loss": 1.3264,
+      "step": 2561
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089752,
+      "loss": 1.3621,
+      "step": 2562
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008974800000000001,
+      "loss": 1.2429,
+      "step": 2563
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089744,
+      "loss": 1.4004,
+      "step": 2564
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008974,
+      "loss": 1.4595,
+      "step": 2565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089736,
+      "loss": 1.3948,
+      "step": 2566
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008973200000000001,
+      "loss": 1.3659,
+      "step": 2567
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089728,
+      "loss": 1.4118,
+      "step": 2568
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008972400000000001,
+      "loss": 1.3864,
+      "step": 2569
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008972,
+      "loss": 1.3282,
+      "step": 2570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008971599999999999,
+      "loss": 1.3271,
+      "step": 2571
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008971200000000001,
+      "loss": 1.3696,
+      "step": 2572
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089708,
+      "loss": 1.3928,
+      "step": 2573
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089704,
+      "loss": 1.3752,
+      "step": 2574
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000897,
+      "loss": 1.3963,
+      "step": 2575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089696,
+      "loss": 1.4096,
+      "step": 2576
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089692,
+      "loss": 1.2147,
+      "step": 2577
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008968800000000001,
+      "loss": 1.3495,
+      "step": 2578
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089684,
+      "loss": 1.3897,
+      "step": 2579
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008968000000000001,
+      "loss": 1.4615,
+      "step": 2580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089676,
+      "loss": 1.4533,
+      "step": 2581
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008967199999999999,
+      "loss": 1.4303,
+      "step": 2582
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008966800000000001,
+      "loss": 1.3472,
+      "step": 2583
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089664,
+      "loss": 1.3914,
+      "step": 2584
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008966,
+      "loss": 1.3344,
+      "step": 2585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089656,
+      "loss": 1.4108,
+      "step": 2586
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089652,
+      "loss": 1.4234,
+      "step": 2587
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008964800000000001,
+      "loss": 1.3438,
+      "step": 2588
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008964400000000001,
+      "loss": 1.3674,
+      "step": 2589
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008964,
+      "loss": 1.3742,
+      "step": 2590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089636,
+      "loss": 1.3698,
+      "step": 2591
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089632,
+      "loss": 1.3667,
+      "step": 2592
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089628,
+      "loss": 1.3712,
+      "step": 2593
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008962400000000001,
+      "loss": 1.3383,
+      "step": 2594
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008962,
+      "loss": 1.3796,
+      "step": 2595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089616,
+      "loss": 1.4221,
+      "step": 2596
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089612,
+      "loss": 1.385,
+      "step": 2597
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089608,
+      "loss": 1.3836,
+      "step": 2598
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089604,
+      "loss": 1.3742,
+      "step": 2599
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000896,
+      "loss": 1.2941,
+      "step": 2600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089596,
+      "loss": 1.3322,
+      "step": 2601
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089592,
+      "loss": 1.345,
+      "step": 2602
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089588,
+      "loss": 1.3248,
+      "step": 2603
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089584,
+      "loss": 1.2837,
+      "step": 2604
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008958000000000001,
+      "loss": 1.329,
+      "step": 2605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089576,
+      "loss": 1.3516,
+      "step": 2606
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089572,
+      "loss": 1.3352,
+      "step": 2607
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089568,
+      "loss": 1.348,
+      "step": 2608
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089564,
+      "loss": 1.3748,
+      "step": 2609
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008956,
+      "loss": 1.5087,
+      "step": 2610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089556,
+      "loss": 1.2847,
+      "step": 2611
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089552,
+      "loss": 1.3817,
+      "step": 2612
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089548,
+      "loss": 1.3084,
+      "step": 2613
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008954400000000001,
+      "loss": 1.4012,
+      "step": 2614
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008954,
+      "loss": 1.3746,
+      "step": 2615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008953600000000001,
+      "loss": 1.3825,
+      "step": 2616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089532,
+      "loss": 1.4736,
+      "step": 2617
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089528,
+      "loss": 1.3731,
+      "step": 2618
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008952400000000001,
+      "loss": 1.3602,
+      "step": 2619
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008952,
+      "loss": 1.367,
+      "step": 2620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089516,
+      "loss": 1.3971,
+      "step": 2621
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089512,
+      "loss": 1.3692,
+      "step": 2622
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089508,
+      "loss": 1.2941,
+      "step": 2623
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008950399999999999,
+      "loss": 1.5489,
+      "step": 2624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008950000000000001,
+      "loss": 1.4024,
+      "step": 2625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089496,
+      "loss": 1.3096,
+      "step": 2626
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008949200000000001,
+      "loss": 1.4465,
+      "step": 2627
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089488,
+      "loss": 1.2628,
+      "step": 2628
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008948399999999999,
+      "loss": 1.3057,
+      "step": 2629
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008948000000000001,
+      "loss": 1.3607,
+      "step": 2630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089476,
+      "loss": 1.4399,
+      "step": 2631
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089472,
+      "loss": 1.3764,
+      "step": 2632
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089468,
+      "loss": 1.4052,
+      "step": 2633
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089464,
+      "loss": 1.4319,
+      "step": 2634
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008946,
+      "loss": 1.2974,
+      "step": 2635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008945600000000001,
+      "loss": 1.3987,
+      "step": 2636
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089452,
+      "loss": 1.3124,
+      "step": 2637
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008944800000000001,
+      "loss": 1.4245,
+      "step": 2638
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089444,
+      "loss": 1.2907,
+      "step": 2639
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008944,
+      "loss": 1.339,
+      "step": 2640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008943600000000001,
+      "loss": 1.3255,
+      "step": 2641
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089432,
+      "loss": 1.3686,
+      "step": 2642
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089428,
+      "loss": 1.4203,
+      "step": 2643
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089424,
+      "loss": 1.4102,
+      "step": 2644
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008942,
+      "loss": 1.2542,
+      "step": 2645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089416,
+      "loss": 1.3222,
+      "step": 2646
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008941200000000001,
+      "loss": 1.4176,
+      "step": 2647
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089408,
+      "loss": 1.406,
+      "step": 2648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008940399999999999,
+      "loss": 1.3911,
+      "step": 2649
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000894,
+      "loss": 1.3359,
+      "step": 2650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089396,
+      "loss": 1.4215,
+      "step": 2651
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008939200000000001,
+      "loss": 1.3449,
+      "step": 2652
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089388,
+      "loss": 1.2487,
+      "step": 2653
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089384,
+      "loss": 1.3529,
+      "step": 2654
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008938,
+      "loss": 1.3617,
+      "step": 2655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008937600000000001,
+      "loss": 1.3763,
+      "step": 2656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089372,
+      "loss": 1.3875,
+      "step": 2657
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089368,
+      "loss": 1.3042,
+      "step": 2658
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00089364,
+      "loss": 1.3108,
+      "step": 2659
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0008935999999999999,
+      "loss": 1.3738,
+      "step": 2660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008935600000000001,
+      "loss": 1.3223,
+      "step": 2661
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089352,
+      "loss": 1.3336,
+      "step": 2662
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008934800000000001,
+      "loss": 1.3698,
+      "step": 2663
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089344,
+      "loss": 1.3119,
+      "step": 2664
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008934,
+      "loss": 1.3659,
+      "step": 2665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089336,
+      "loss": 1.3204,
+      "step": 2666
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008933200000000001,
+      "loss": 1.3088,
+      "step": 2667
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089328,
+      "loss": 1.3794,
+      "step": 2668
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089324,
+      "loss": 1.376,
+      "step": 2669
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008932,
+      "loss": 1.4137,
+      "step": 2670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008931599999999999,
+      "loss": 1.3033,
+      "step": 2671
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008931200000000001,
+      "loss": 1.3942,
+      "step": 2672
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089308,
+      "loss": 1.388,
+      "step": 2673
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089304,
+      "loss": 1.3762,
+      "step": 2674
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000893,
+      "loss": 1.4051,
+      "step": 2675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089296,
+      "loss": 1.319,
+      "step": 2676
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008929200000000001,
+      "loss": 1.3987,
+      "step": 2677
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089288,
+      "loss": 1.4306,
+      "step": 2678
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089284,
+      "loss": 1.4425,
+      "step": 2679
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008928,
+      "loss": 1.3176,
+      "step": 2680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089276,
+      "loss": 1.4261,
+      "step": 2681
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089272,
+      "loss": 1.313,
+      "step": 2682
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008926800000000001,
+      "loss": 1.3601,
+      "step": 2683
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089264,
+      "loss": 1.434,
+      "step": 2684
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008926,
+      "loss": 1.3524,
+      "step": 2685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089256,
+      "loss": 1.3816,
+      "step": 2686
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008925199999999999,
+      "loss": 1.46,
+      "step": 2687
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008924800000000001,
+      "loss": 1.3919,
+      "step": 2688
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089244,
+      "loss": 1.34,
+      "step": 2689
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008924,
+      "loss": 1.4076,
+      "step": 2690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089236,
+      "loss": 1.352,
+      "step": 2691
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089232,
+      "loss": 1.3713,
+      "step": 2692
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089228,
+      "loss": 1.3035,
+      "step": 2693
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008922400000000001,
+      "loss": 1.252,
+      "step": 2694
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008922,
+      "loss": 1.2916,
+      "step": 2695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089216,
+      "loss": 1.4131,
+      "step": 2696
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089212,
+      "loss": 1.3737,
+      "step": 2697
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089208,
+      "loss": 1.4336,
+      "step": 2698
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008920400000000001,
+      "loss": 1.4569,
+      "step": 2699
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000892,
+      "loss": 1.4333,
+      "step": 2700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089196,
+      "loss": 1.4553,
+      "step": 2701
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089192,
+      "loss": 1.3959,
+      "step": 2702
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008918800000000001,
+      "loss": 1.2165,
+      "step": 2703
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089184,
+      "loss": 1.3498,
+      "step": 2704
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008918000000000001,
+      "loss": 1.3457,
+      "step": 2705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089176,
+      "loss": 1.4153,
+      "step": 2706
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008917199999999999,
+      "loss": 1.3586,
+      "step": 2707
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008916800000000001,
+      "loss": 1.4096,
+      "step": 2708
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089164,
+      "loss": 1.375,
+      "step": 2709
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008916,
+      "loss": 1.4436,
+      "step": 2710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089156,
+      "loss": 1.2852,
+      "step": 2711
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089152,
+      "loss": 1.3099,
+      "step": 2712
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089148,
+      "loss": 1.3833,
+      "step": 2713
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008914400000000001,
+      "loss": 1.3479,
+      "step": 2714
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008914,
+      "loss": 1.3373,
+      "step": 2715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008913600000000001,
+      "loss": 1.4056,
+      "step": 2716
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089132,
+      "loss": 1.4041,
+      "step": 2717
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008912799999999999,
+      "loss": 1.32,
+      "step": 2718
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008912400000000001,
+      "loss": 1.3627,
+      "step": 2719
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008912,
+      "loss": 1.3493,
+      "step": 2720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089116,
+      "loss": 1.3138,
+      "step": 2721
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089112,
+      "loss": 1.3297,
+      "step": 2722
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089108,
+      "loss": 1.319,
+      "step": 2723
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008910400000000001,
+      "loss": 1.2723,
+      "step": 2724
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008910000000000001,
+      "loss": 1.372,
+      "step": 2725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089096,
+      "loss": 1.365,
+      "step": 2726
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089092,
+      "loss": 1.4517,
+      "step": 2727
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089088,
+      "loss": 1.3757,
+      "step": 2728
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089084,
+      "loss": 1.3108,
+      "step": 2729
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008908000000000001,
+      "loss": 1.3243,
+      "step": 2730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089076,
+      "loss": 1.3599,
+      "step": 2731
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089072,
+      "loss": 1.3806,
+      "step": 2732
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089068,
+      "loss": 1.4277,
+      "step": 2733
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089064,
+      "loss": 1.3208,
+      "step": 2734
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008906,
+      "loss": 1.4613,
+      "step": 2735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089056,
+      "loss": 1.388,
+      "step": 2736
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089052,
+      "loss": 1.3171,
+      "step": 2737
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089048,
+      "loss": 1.3921,
+      "step": 2738
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089044,
+      "loss": 1.3897,
+      "step": 2739
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008904,
+      "loss": 1.3986,
+      "step": 2740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008903600000000001,
+      "loss": 1.2815,
+      "step": 2741
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089032,
+      "loss": 1.4467,
+      "step": 2742
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089028,
+      "loss": 1.3837,
+      "step": 2743
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089024,
+      "loss": 1.3658,
+      "step": 2744
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008902000000000001,
+      "loss": 1.3466,
+      "step": 2745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089016,
+      "loss": 1.3626,
+      "step": 2746
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089012,
+      "loss": 1.2704,
+      "step": 2747
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089008,
+      "loss": 1.3356,
+      "step": 2748
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00089004,
+      "loss": 1.3887,
+      "step": 2749
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008900000000000001,
+      "loss": 1.3908,
+      "step": 2750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088996,
+      "loss": 1.3891,
+      "step": 2751
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008899200000000001,
+      "loss": 1.4215,
+      "step": 2752
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088988,
+      "loss": 1.3253,
+      "step": 2753
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088984,
+      "loss": 1.347,
+      "step": 2754
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008898,
+      "loss": 1.3762,
+      "step": 2755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088976,
+      "loss": 1.4131,
+      "step": 2756
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088972,
+      "loss": 1.3577,
+      "step": 2757
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088968,
+      "loss": 1.3602,
+      "step": 2758
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088964,
+      "loss": 1.5643,
+      "step": 2759
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008895999999999999,
+      "loss": 1.3423,
+      "step": 2760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008895600000000001,
+      "loss": 1.3305,
+      "step": 2761
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088952,
+      "loss": 1.2789,
+      "step": 2762
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008894800000000001,
+      "loss": 1.4254,
+      "step": 2763
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088944,
+      "loss": 1.4148,
+      "step": 2764
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008893999999999999,
+      "loss": 1.4249,
+      "step": 2765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008893600000000001,
+      "loss": 1.3798,
+      "step": 2766
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088932,
+      "loss": 1.3581,
+      "step": 2767
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088928,
+      "loss": 1.3929,
+      "step": 2768
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088924,
+      "loss": 1.3911,
+      "step": 2769
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008892,
+      "loss": 1.4563,
+      "step": 2770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088916,
+      "loss": 1.2926,
+      "step": 2771
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008891200000000001,
+      "loss": 1.4078,
+      "step": 2772
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088908,
+      "loss": 1.3372,
+      "step": 2773
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008890400000000001,
+      "loss": 1.3949,
+      "step": 2774
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000889,
+      "loss": 1.3177,
+      "step": 2775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008889599999999999,
+      "loss": 1.3335,
+      "step": 2776
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008889200000000001,
+      "loss": 1.3358,
+      "step": 2777
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088888,
+      "loss": 1.3743,
+      "step": 2778
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088884,
+      "loss": 1.4143,
+      "step": 2779
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008888,
+      "loss": 1.3127,
+      "step": 2780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088876,
+      "loss": 1.3737,
+      "step": 2781
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088872,
+      "loss": 1.2645,
+      "step": 2782
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008886800000000001,
+      "loss": 1.3045,
+      "step": 2783
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088864,
+      "loss": 1.3585,
+      "step": 2784
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008885999999999999,
+      "loss": 1.4615,
+      "step": 2785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088856,
+      "loss": 1.359,
+      "step": 2786
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088852,
+      "loss": 1.3037,
+      "step": 2787
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008884800000000001,
+      "loss": 1.3313,
+      "step": 2788
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088844,
+      "loss": 1.3415,
+      "step": 2789
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008884,
+      "loss": 1.3764,
+      "step": 2790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088836,
+      "loss": 1.3379,
+      "step": 2791
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008883200000000001,
+      "loss": 1.373,
+      "step": 2792
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088828,
+      "loss": 1.465,
+      "step": 2793
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008882400000000001,
+      "loss": 1.3529,
+      "step": 2794
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008882,
+      "loss": 1.3286,
+      "step": 2795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008881599999999999,
+      "loss": 1.3089,
+      "step": 2796
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008881200000000001,
+      "loss": 1.3876,
+      "step": 2797
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088808,
+      "loss": 1.3017,
+      "step": 2798
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008880400000000001,
+      "loss": 1.3472,
+      "step": 2799
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000888,
+      "loss": 1.3289,
+      "step": 2800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088796,
+      "loss": 1.3708,
+      "step": 2801
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088792,
+      "loss": 1.3094,
+      "step": 2802
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008878800000000001,
+      "loss": 1.316,
+      "step": 2803
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088784,
+      "loss": 1.3402,
+      "step": 2804
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008878,
+      "loss": 1.3849,
+      "step": 2805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088776,
+      "loss": 1.3686,
+      "step": 2806
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008877199999999999,
+      "loss": 1.393,
+      "step": 2807
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008876800000000001,
+      "loss": 1.4126,
+      "step": 2808
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088764,
+      "loss": 1.401,
+      "step": 2809
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008876,
+      "loss": 1.3366,
+      "step": 2810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088756,
+      "loss": 1.3635,
+      "step": 2811
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088752,
+      "loss": 1.3059,
+      "step": 2812
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008874800000000001,
+      "loss": 1.3533,
+      "step": 2813
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088744,
+      "loss": 1.4028,
+      "step": 2814
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008874,
+      "loss": 1.3777,
+      "step": 2815
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088736,
+      "loss": 1.4062,
+      "step": 2816
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088732,
+      "loss": 1.3496,
+      "step": 2817
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088728,
+      "loss": 1.3324,
+      "step": 2818
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008872400000000001,
+      "loss": 1.3254,
+      "step": 2819
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008872,
+      "loss": 1.3643,
+      "step": 2820
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088716,
+      "loss": 1.3809,
+      "step": 2821
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088712,
+      "loss": 1.3697,
+      "step": 2822
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088708,
+      "loss": 1.3571,
+      "step": 2823
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008870400000000001,
+      "loss": 1.3488,
+      "step": 2824
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000887,
+      "loss": 1.4446,
+      "step": 2825
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088696,
+      "loss": 1.4309,
+      "step": 2826
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088692,
+      "loss": 1.3301,
+      "step": 2827
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088688,
+      "loss": 1.4168,
+      "step": 2828
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088684,
+      "loss": 1.2484,
+      "step": 2829
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008868000000000001,
+      "loss": 1.2678,
+      "step": 2830
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088676,
+      "loss": 1.3136,
+      "step": 2831
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088672,
+      "loss": 1.4059,
+      "step": 2832
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088668,
+      "loss": 1.3589,
+      "step": 2833
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088664,
+      "loss": 1.2902,
+      "step": 2834
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008866000000000001,
+      "loss": 1.2723,
+      "step": 2835
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088656,
+      "loss": 1.2505,
+      "step": 2836
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088652,
+      "loss": 1.325,
+      "step": 2837
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088648,
+      "loss": 1.2999,
+      "step": 2838
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008864400000000001,
+      "loss": 1.3542,
+      "step": 2839
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008864,
+      "loss": 1.3467,
+      "step": 2840
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008863600000000001,
+      "loss": 1.3515,
+      "step": 2841
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088632,
+      "loss": 1.4755,
+      "step": 2842
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008862799999999999,
+      "loss": 1.3664,
+      "step": 2843
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088624,
+      "loss": 1.3863,
+      "step": 2844
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008862,
+      "loss": 1.2671,
+      "step": 2845
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088616,
+      "loss": 1.3249,
+      "step": 2846
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088612,
+      "loss": 1.3428,
+      "step": 2847
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088608,
+      "loss": 1.3871,
+      "step": 2848
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088604,
+      "loss": 1.3687,
+      "step": 2849
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008860000000000001,
+      "loss": 1.3247,
+      "step": 2850
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088596,
+      "loss": 1.4036,
+      "step": 2851
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008859200000000001,
+      "loss": 1.2532,
+      "step": 2852
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088588,
+      "loss": 1.4435,
+      "step": 2853
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008858399999999999,
+      "loss": 1.4177,
+      "step": 2854
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008858000000000001,
+      "loss": 1.412,
+      "step": 2855
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088576,
+      "loss": 1.4611,
+      "step": 2856
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088572,
+      "loss": 1.3354,
+      "step": 2857
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088568,
+      "loss": 1.3075,
+      "step": 2858
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088564,
+      "loss": 1.327,
+      "step": 2859
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008856000000000001,
+      "loss": 1.3695,
+      "step": 2860
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008855600000000001,
+      "loss": 1.4198,
+      "step": 2861
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088552,
+      "loss": 1.3349,
+      "step": 2862
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088548,
+      "loss": 1.4232,
+      "step": 2863
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088544,
+      "loss": 1.3619,
+      "step": 2864
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008853999999999999,
+      "loss": 1.4202,
+      "step": 2865
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008853600000000001,
+      "loss": 1.4533,
+      "step": 2866
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088532,
+      "loss": 1.2886,
+      "step": 2867
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088528,
+      "loss": 1.4779,
+      "step": 2868
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088524,
+      "loss": 1.3075,
+      "step": 2869
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008852,
+      "loss": 1.3701,
+      "step": 2870
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088516,
+      "loss": 1.3603,
+      "step": 2871
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088512,
+      "loss": 1.354,
+      "step": 2872
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088508,
+      "loss": 1.2124,
+      "step": 2873
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088504,
+      "loss": 1.3876,
+      "step": 2874
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000885,
+      "loss": 1.3159,
+      "step": 2875
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088496,
+      "loss": 1.3534,
+      "step": 2876
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008849200000000001,
+      "loss": 1.4771,
+      "step": 2877
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088488,
+      "loss": 1.2748,
+      "step": 2878
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088484,
+      "loss": 1.3049,
+      "step": 2879
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008848,
+      "loss": 1.3033,
+      "step": 2880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008847600000000001,
+      "loss": 1.4566,
+      "step": 2881
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088472,
+      "loss": 1.3789,
+      "step": 2882
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088468,
+      "loss": 1.3505,
+      "step": 2883
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088464,
+      "loss": 1.4166,
+      "step": 2884
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008846,
+      "loss": 1.3178,
+      "step": 2885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008845600000000001,
+      "loss": 1.3383,
+      "step": 2886
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088452,
+      "loss": 1.3765,
+      "step": 2887
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008844800000000001,
+      "loss": 1.3254,
+      "step": 2888
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088444,
+      "loss": 1.395,
+      "step": 2889
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008844,
+      "loss": 1.4589,
+      "step": 2890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088436,
+      "loss": 1.363,
+      "step": 2891
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088432,
+      "loss": 1.3283,
+      "step": 2892
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088428,
+      "loss": 1.3247,
+      "step": 2893
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088424,
+      "loss": 1.3598,
+      "step": 2894
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008842,
+      "loss": 1.3302,
+      "step": 2895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008841599999999999,
+      "loss": 1.4718,
+      "step": 2896
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008841200000000001,
+      "loss": 1.3982,
+      "step": 2897
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088408,
+      "loss": 1.3533,
+      "step": 2898
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008840400000000001,
+      "loss": 1.3059,
+      "step": 2899
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000884,
+      "loss": 1.4912,
+      "step": 2900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008839599999999999,
+      "loss": 1.3863,
+      "step": 2901
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008839200000000001,
+      "loss": 1.3852,
+      "step": 2902
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088388,
+      "loss": 1.2962,
+      "step": 2903
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088384,
+      "loss": 1.3743,
+      "step": 2904
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008838,
+      "loss": 1.3446,
+      "step": 2905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088376,
+      "loss": 1.3853,
+      "step": 2906
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088372,
+      "loss": 1.3392,
+      "step": 2907
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008836800000000001,
+      "loss": 1.3223,
+      "step": 2908
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088364,
+      "loss": 1.3001,
+      "step": 2909
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008836000000000001,
+      "loss": 1.3811,
+      "step": 2910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088356,
+      "loss": 1.3267,
+      "step": 2911
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008835199999999999,
+      "loss": 1.313,
+      "step": 2912
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0008834800000000001,
+      "loss": 1.4028,
+      "step": 2913
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00088344,
+      "loss": 1.3979,
+      "step": 2914
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008834,
+      "loss": 1.2684,
+      "step": 2915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088336,
+      "loss": 1.294,
+      "step": 2916
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088332,
+      "loss": 1.3349,
+      "step": 2917
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088328,
+      "loss": 1.4084,
+      "step": 2918
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008832400000000001,
+      "loss": 1.425,
+      "step": 2919
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008832,
+      "loss": 1.3591,
+      "step": 2920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008831599999999999,
+      "loss": 1.4645,
+      "step": 2921
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088312,
+      "loss": 1.3259,
+      "step": 2922
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088308,
+      "loss": 1.379,
+      "step": 2923
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008830400000000001,
+      "loss": 1.2853,
+      "step": 2924
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000883,
+      "loss": 1.3051,
+      "step": 2925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088296,
+      "loss": 1.3396,
+      "step": 2926
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088292,
+      "loss": 1.3217,
+      "step": 2927
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008828800000000001,
+      "loss": 1.4087,
+      "step": 2928
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088284,
+      "loss": 1.3771,
+      "step": 2929
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008828000000000001,
+      "loss": 1.4249,
+      "step": 2930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088276,
+      "loss": 1.3293,
+      "step": 2931
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008827199999999999,
+      "loss": 1.3863,
+      "step": 2932
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088268,
+      "loss": 1.3447,
+      "step": 2933
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088264,
+      "loss": 1.3178,
+      "step": 2934
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008826000000000001,
+      "loss": 1.382,
+      "step": 2935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088256,
+      "loss": 1.3417,
+      "step": 2936
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088252,
+      "loss": 1.3778,
+      "step": 2937
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088248,
+      "loss": 1.3651,
+      "step": 2938
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008824400000000001,
+      "loss": 1.3574,
+      "step": 2939
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008824,
+      "loss": 1.3599,
+      "step": 2940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088236,
+      "loss": 1.2867,
+      "step": 2941
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088232,
+      "loss": 1.2822,
+      "step": 2942
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008822799999999999,
+      "loss": 1.2609,
+      "step": 2943
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008822400000000001,
+      "loss": 1.3836,
+      "step": 2944
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008822,
+      "loss": 1.431,
+      "step": 2945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008821600000000001,
+      "loss": 1.3575,
+      "step": 2946
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088212,
+      "loss": 1.4363,
+      "step": 2947
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088208,
+      "loss": 1.3842,
+      "step": 2948
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008820400000000001,
+      "loss": 1.3915,
+      "step": 2949
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000882,
+      "loss": 1.2702,
+      "step": 2950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088196,
+      "loss": 1.3426,
+      "step": 2951
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088192,
+      "loss": 1.2397,
+      "step": 2952
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088188,
+      "loss": 1.3684,
+      "step": 2953
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088184,
+      "loss": 1.3032,
+      "step": 2954
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008818000000000001,
+      "loss": 1.4369,
+      "step": 2955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088176,
+      "loss": 1.3324,
+      "step": 2956
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088172,
+      "loss": 1.4169,
+      "step": 2957
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088168,
+      "loss": 1.4237,
+      "step": 2958
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088164,
+      "loss": 1.2877,
+      "step": 2959
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008816000000000001,
+      "loss": 1.3505,
+      "step": 2960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088156,
+      "loss": 1.3943,
+      "step": 2961
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088152,
+      "loss": 1.3188,
+      "step": 2962
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088148,
+      "loss": 1.3542,
+      "step": 2963
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088144,
+      "loss": 1.335,
+      "step": 2964
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008814,
+      "loss": 1.3331,
+      "step": 2965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008813600000000001,
+      "loss": 1.3064,
+      "step": 2966
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088132,
+      "loss": 1.3281,
+      "step": 2967
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088128,
+      "loss": 1.2722,
+      "step": 2968
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088124,
+      "loss": 1.3446,
+      "step": 2969
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008812,
+      "loss": 1.3123,
+      "step": 2970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008811600000000001,
+      "loss": 1.4169,
+      "step": 2971
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088112,
+      "loss": 1.4293,
+      "step": 2972
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088108,
+      "loss": 1.3802,
+      "step": 2973
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088104,
+      "loss": 1.3146,
+      "step": 2974
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008810000000000001,
+      "loss": 1.338,
+      "step": 2975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088096,
+      "loss": 1.3096,
+      "step": 2976
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008809200000000001,
+      "loss": 1.2712,
+      "step": 2977
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088088,
+      "loss": 1.3563,
+      "step": 2978
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008808399999999999,
+      "loss": 1.373,
+      "step": 2979
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008808,
+      "loss": 1.4376,
+      "step": 2980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088076,
+      "loss": 1.4348,
+      "step": 2981
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088072,
+      "loss": 1.3924,
+      "step": 2982
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088068,
+      "loss": 1.3864,
+      "step": 2983
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088064,
+      "loss": 1.3675,
+      "step": 2984
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008806,
+      "loss": 1.4377,
+      "step": 2985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008805600000000001,
+      "loss": 1.3498,
+      "step": 2986
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088052,
+      "loss": 1.3683,
+      "step": 2987
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008804800000000001,
+      "loss": 1.3279,
+      "step": 2988
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088044,
+      "loss": 1.3176,
+      "step": 2989
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008803999999999999,
+      "loss": 1.4243,
+      "step": 2990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008803600000000001,
+      "loss": 1.3522,
+      "step": 2991
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088032,
+      "loss": 1.2691,
+      "step": 2992
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088028,
+      "loss": 1.3149,
+      "step": 2993
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088024,
+      "loss": 1.3224,
+      "step": 2994
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008802,
+      "loss": 1.342,
+      "step": 2995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008801600000000001,
+      "loss": 1.2697,
+      "step": 2996
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008801200000000001,
+      "loss": 1.2461,
+      "step": 2997
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088008,
+      "loss": 1.3939,
+      "step": 2998
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088004,
+      "loss": 1.2759,
+      "step": 2999
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00088,
+      "loss": 1.2825,
+      "step": 3000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008799599999999999,
+      "loss": 1.3698,
+      "step": 3001
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008799200000000001,
+      "loss": 1.321,
+      "step": 3002
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087988,
+      "loss": 1.2936,
+      "step": 3003
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087984,
+      "loss": 1.3607,
+      "step": 3004
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008798,
+      "loss": 1.2377,
+      "step": 3005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087976,
+      "loss": 1.3975,
+      "step": 3006
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087972,
+      "loss": 1.2998,
+      "step": 3007
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008796800000000001,
+      "loss": 1.3348,
+      "step": 3008
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087964,
+      "loss": 1.3927,
+      "step": 3009
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008796,
+      "loss": 1.3454,
+      "step": 3010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087956,
+      "loss": 1.4394,
+      "step": 3011
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087952,
+      "loss": 1.4552,
+      "step": 3012
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008794800000000001,
+      "loss": 1.3238,
+      "step": 3013
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087944,
+      "loss": 1.3173,
+      "step": 3014
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008794,
+      "loss": 1.433,
+      "step": 3015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087936,
+      "loss": 1.3011,
+      "step": 3016
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008793200000000001,
+      "loss": 1.3506,
+      "step": 3017
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087928,
+      "loss": 1.325,
+      "step": 3018
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087924,
+      "loss": 1.3105,
+      "step": 3019
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008792,
+      "loss": 1.2967,
+      "step": 3020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087916,
+      "loss": 1.3717,
+      "step": 3021
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087912,
+      "loss": 1.36,
+      "step": 3022
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087908,
+      "loss": 1.434,
+      "step": 3023
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008790400000000001,
+      "loss": 1.3042,
+      "step": 3024
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000879,
+      "loss": 1.3105,
+      "step": 3025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087896,
+      "loss": 1.3521,
+      "step": 3026
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087892,
+      "loss": 1.3228,
+      "step": 3027
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087888,
+      "loss": 1.291,
+      "step": 3028
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087884,
+      "loss": 1.3711,
+      "step": 3029
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008788,
+      "loss": 1.2525,
+      "step": 3030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087876,
+      "loss": 1.2574,
+      "step": 3031
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008787199999999999,
+      "loss": 1.3958,
+      "step": 3032
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008786800000000001,
+      "loss": 1.3166,
+      "step": 3033
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087864,
+      "loss": 1.3829,
+      "step": 3034
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008786000000000001,
+      "loss": 1.3639,
+      "step": 3035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087856,
+      "loss": 1.3166,
+      "step": 3036
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087852,
+      "loss": 1.3137,
+      "step": 3037
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008784800000000001,
+      "loss": 1.3751,
+      "step": 3038
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087844,
+      "loss": 1.3111,
+      "step": 3039
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008784,
+      "loss": 1.4471,
+      "step": 3040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087836,
+      "loss": 1.3997,
+      "step": 3041
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087832,
+      "loss": 1.3682,
+      "step": 3042
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087828,
+      "loss": 1.3942,
+      "step": 3043
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008782400000000001,
+      "loss": 1.3171,
+      "step": 3044
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008782,
+      "loss": 1.3731,
+      "step": 3045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008781600000000001,
+      "loss": 1.3779,
+      "step": 3046
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087812,
+      "loss": 1.3488,
+      "step": 3047
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008780799999999999,
+      "loss": 1.3281,
+      "step": 3048
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008780400000000001,
+      "loss": 1.36,
+      "step": 3049
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000878,
+      "loss": 1.384,
+      "step": 3050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087796,
+      "loss": 1.3295,
+      "step": 3051
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087792,
+      "loss": 1.2422,
+      "step": 3052
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087788,
+      "loss": 1.274,
+      "step": 3053
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087784,
+      "loss": 1.33,
+      "step": 3054
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008778000000000001,
+      "loss": 1.36,
+      "step": 3055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087776,
+      "loss": 1.33,
+      "step": 3056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008777199999999999,
+      "loss": 1.1895,
+      "step": 3057
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087768,
+      "loss": 1.3859,
+      "step": 3058
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087764,
+      "loss": 1.3091,
+      "step": 3059
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008776000000000001,
+      "loss": 1.393,
+      "step": 3060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087756,
+      "loss": 1.3081,
+      "step": 3061
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087752,
+      "loss": 1.4844,
+      "step": 3062
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087748,
+      "loss": 1.2836,
+      "step": 3063
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008774400000000001,
+      "loss": 1.3681,
+      "step": 3064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008774,
+      "loss": 1.2138,
+      "step": 3065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008773600000000001,
+      "loss": 1.3354,
+      "step": 3066
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087732,
+      "loss": 1.3242,
+      "step": 3067
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008772799999999999,
+      "loss": 1.2584,
+      "step": 3068
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087724,
+      "loss": 1.3344,
+      "step": 3069
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008772,
+      "loss": 1.3042,
+      "step": 3070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008771600000000001,
+      "loss": 1.337,
+      "step": 3071
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087712,
+      "loss": 1.3794,
+      "step": 3072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087708,
+      "loss": 1.3693,
+      "step": 3073
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087704,
+      "loss": 1.3989,
+      "step": 3074
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008770000000000001,
+      "loss": 1.3289,
+      "step": 3075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087696,
+      "loss": 1.4064,
+      "step": 3076
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087692,
+      "loss": 1.3981,
+      "step": 3077
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087688,
+      "loss": 1.2877,
+      "step": 3078
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008768399999999999,
+      "loss": 1.3876,
+      "step": 3079
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008768000000000001,
+      "loss": 1.3904,
+      "step": 3080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087676,
+      "loss": 1.3485,
+      "step": 3081
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008767200000000001,
+      "loss": 1.2825,
+      "step": 3082
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087668,
+      "loss": 1.3445,
+      "step": 3083
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087664,
+      "loss": 1.2862,
+      "step": 3084
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008766000000000001,
+      "loss": 1.3662,
+      "step": 3085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008765600000000001,
+      "loss": 1.3597,
+      "step": 3086
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087652,
+      "loss": 1.4342,
+      "step": 3087
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087648,
+      "loss": 1.3908,
+      "step": 3088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087644,
+      "loss": 1.3791,
+      "step": 3089
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008763999999999999,
+      "loss": 1.3811,
+      "step": 3090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008763600000000001,
+      "loss": 1.3489,
+      "step": 3091
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087632,
+      "loss": 1.3045,
+      "step": 3092
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087628,
+      "loss": 1.2715,
+      "step": 3093
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087624,
+      "loss": 1.2673,
+      "step": 3094
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008762,
+      "loss": 1.2808,
+      "step": 3095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008761600000000001,
+      "loss": 1.3108,
+      "step": 3096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087612,
+      "loss": 1.2294,
+      "step": 3097
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087608,
+      "loss": 1.3632,
+      "step": 3098
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087604,
+      "loss": 1.3821,
+      "step": 3099
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000876,
+      "loss": 1.3384,
+      "step": 3100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087596,
+      "loss": 1.3697,
+      "step": 3101
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008759200000000001,
+      "loss": 1.3291,
+      "step": 3102
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087588,
+      "loss": 1.2056,
+      "step": 3103
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087584,
+      "loss": 1.3848,
+      "step": 3104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008758,
+      "loss": 1.4035,
+      "step": 3105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087576,
+      "loss": 1.223,
+      "step": 3106
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008757200000000001,
+      "loss": 1.3099,
+      "step": 3107
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087568,
+      "loss": 1.3965,
+      "step": 3108
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087564,
+      "loss": 1.286,
+      "step": 3109
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008756,
+      "loss": 1.3237,
+      "step": 3110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087556,
+      "loss": 1.2098,
+      "step": 3111
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087552,
+      "loss": 1.3094,
+      "step": 3112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008754800000000001,
+      "loss": 1.3312,
+      "step": 3113
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087544,
+      "loss": 1.3028,
+      "step": 3114
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008754,
+      "loss": 1.3409,
+      "step": 3115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087536,
+      "loss": 1.3132,
+      "step": 3116
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087532,
+      "loss": 1.2689,
+      "step": 3117
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087528,
+      "loss": 1.3922,
+      "step": 3118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087524,
+      "loss": 1.2903,
+      "step": 3119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008752,
+      "loss": 1.2642,
+      "step": 3120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087516,
+      "loss": 1.4449,
+      "step": 3121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008751200000000001,
+      "loss": 1.3332,
+      "step": 3122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087508,
+      "loss": 1.3442,
+      "step": 3123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008750400000000001,
+      "loss": 1.3454,
+      "step": 3124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000875,
+      "loss": 1.3343,
+      "step": 3125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008749599999999999,
+      "loss": 1.2755,
+      "step": 3126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008749200000000001,
+      "loss": 1.3525,
+      "step": 3127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087488,
+      "loss": 1.3509,
+      "step": 3128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087484,
+      "loss": 1.305,
+      "step": 3129
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008748,
+      "loss": 1.2715,
+      "step": 3130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087476,
+      "loss": 1.4225,
+      "step": 3131
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008747200000000001,
+      "loss": 1.2433,
+      "step": 3132
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008746800000000001,
+      "loss": 1.3679,
+      "step": 3133
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087464,
+      "loss": 1.2881,
+      "step": 3134
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008746,
+      "loss": 1.3068,
+      "step": 3135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087456,
+      "loss": 1.339,
+      "step": 3136
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008745199999999999,
+      "loss": 1.2903,
+      "step": 3137
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008744800000000001,
+      "loss": 1.3808,
+      "step": 3138
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087444,
+      "loss": 1.3081,
+      "step": 3139
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008744,
+      "loss": 1.3318,
+      "step": 3140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087436,
+      "loss": 1.3644,
+      "step": 3141
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087432,
+      "loss": 1.3508,
+      "step": 3142
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087428,
+      "loss": 1.368,
+      "step": 3143
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008742400000000001,
+      "loss": 1.3445,
+      "step": 3144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008742,
+      "loss": 1.4261,
+      "step": 3145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087416,
+      "loss": 1.2845,
+      "step": 3146
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087412,
+      "loss": 1.4031,
+      "step": 3147
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087408,
+      "loss": 1.3089,
+      "step": 3148
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008740400000000001,
+      "loss": 1.4025,
+      "step": 3149
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000874,
+      "loss": 1.3257,
+      "step": 3150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087396,
+      "loss": 1.3481,
+      "step": 3151
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087392,
+      "loss": 1.362,
+      "step": 3152
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008738800000000001,
+      "loss": 1.4066,
+      "step": 3153
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087384,
+      "loss": 1.3643,
+      "step": 3154
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008738,
+      "loss": 1.3349,
+      "step": 3155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087376,
+      "loss": 1.2486,
+      "step": 3156
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087372,
+      "loss": 1.283,
+      "step": 3157
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087368,
+      "loss": 1.3914,
+      "step": 3158
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087364,
+      "loss": 1.351,
+      "step": 3159
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008736000000000001,
+      "loss": 1.253,
+      "step": 3160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087356,
+      "loss": 1.3206,
+      "step": 3161
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087352,
+      "loss": 1.386,
+      "step": 3162
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087348,
+      "loss": 1.325,
+      "step": 3163
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087344,
+      "loss": 1.3572,
+      "step": 3164
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0008734,
+      "loss": 1.3245,
+      "step": 3165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087336,
+      "loss": 1.2842,
+      "step": 3166
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00087332,
+      "loss": 1.3025,
+      "step": 3167
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008732799999999999,
+      "loss": 1.2882,
+      "step": 3168
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008732400000000001,
+      "loss": 1.2717,
+      "step": 3169
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008732,
+      "loss": 1.2706,
+      "step": 3170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008731600000000001,
+      "loss": 1.3061,
+      "step": 3171
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087312,
+      "loss": 1.3036,
+      "step": 3172
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087308,
+      "loss": 1.3142,
+      "step": 3173
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008730400000000001,
+      "loss": 1.3118,
+      "step": 3174
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000873,
+      "loss": 1.3871,
+      "step": 3175
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087296,
+      "loss": 1.3488,
+      "step": 3176
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087292,
+      "loss": 1.358,
+      "step": 3177
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087288,
+      "loss": 1.302,
+      "step": 3178
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008728399999999999,
+      "loss": 1.3817,
+      "step": 3179
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008728000000000001,
+      "loss": 1.395,
+      "step": 3180
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087276,
+      "loss": 1.3504,
+      "step": 3181
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008727200000000001,
+      "loss": 1.32,
+      "step": 3182
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087268,
+      "loss": 1.2874,
+      "step": 3183
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008726399999999999,
+      "loss": 1.3382,
+      "step": 3184
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008726000000000001,
+      "loss": 1.3251,
+      "step": 3185
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087256,
+      "loss": 1.3797,
+      "step": 3186
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087252,
+      "loss": 1.2961,
+      "step": 3187
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087248,
+      "loss": 1.3257,
+      "step": 3188
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087244,
+      "loss": 1.3642,
+      "step": 3189
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008724,
+      "loss": 1.4851,
+      "step": 3190
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008723600000000001,
+      "loss": 1.2824,
+      "step": 3191
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087232,
+      "loss": 1.3845,
+      "step": 3192
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008722800000000001,
+      "loss": 1.3163,
+      "step": 3193
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087224,
+      "loss": 1.3511,
+      "step": 3194
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008722,
+      "loss": 1.403,
+      "step": 3195
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008721600000000001,
+      "loss": 1.4209,
+      "step": 3196
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087212,
+      "loss": 1.3525,
+      "step": 3197
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087208,
+      "loss": 1.2974,
+      "step": 3198
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087204,
+      "loss": 1.383,
+      "step": 3199
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000872,
+      "loss": 1.3853,
+      "step": 3200
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087196,
+      "loss": 1.3144,
+      "step": 3201
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008719200000000001,
+      "loss": 1.2853,
+      "step": 3202
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087188,
+      "loss": 1.3241,
+      "step": 3203
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008718399999999999,
+      "loss": 1.3938,
+      "step": 3204
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008718,
+      "loss": 1.3733,
+      "step": 3205
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087176,
+      "loss": 1.2404,
+      "step": 3206
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008717200000000001,
+      "loss": 1.2872,
+      "step": 3207
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087168,
+      "loss": 1.292,
+      "step": 3208
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087164,
+      "loss": 1.2877,
+      "step": 3209
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008716,
+      "loss": 1.323,
+      "step": 3210
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008715600000000001,
+      "loss": 1.2382,
+      "step": 3211
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087152,
+      "loss": 1.3261,
+      "step": 3212
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087148,
+      "loss": 1.3806,
+      "step": 3213
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087144,
+      "loss": 1.3224,
+      "step": 3214
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008713999999999999,
+      "loss": 1.3198,
+      "step": 3215
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008713600000000001,
+      "loss": 1.342,
+      "step": 3216
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087132,
+      "loss": 1.3367,
+      "step": 3217
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008712800000000001,
+      "loss": 1.3062,
+      "step": 3218
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087124,
+      "loss": 1.2792,
+      "step": 3219
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008712,
+      "loss": 1.37,
+      "step": 3220
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008711600000000001,
+      "loss": 1.2628,
+      "step": 3221
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008711200000000001,
+      "loss": 1.3178,
+      "step": 3222
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087108,
+      "loss": 1.2403,
+      "step": 3223
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087104,
+      "loss": 1.3015,
+      "step": 3224
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000871,
+      "loss": 1.3078,
+      "step": 3225
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008709599999999999,
+      "loss": 1.3604,
+      "step": 3226
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008709200000000001,
+      "loss": 1.3511,
+      "step": 3227
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087088,
+      "loss": 1.3919,
+      "step": 3228
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087084,
+      "loss": 1.41,
+      "step": 3229
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008708,
+      "loss": 1.3149,
+      "step": 3230
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087076,
+      "loss": 1.291,
+      "step": 3231
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008707200000000001,
+      "loss": 1.2578,
+      "step": 3232
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087068,
+      "loss": 1.4075,
+      "step": 3233
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087064,
+      "loss": 1.352,
+      "step": 3234
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008706,
+      "loss": 1.3051,
+      "step": 3235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087056,
+      "loss": 1.3545,
+      "step": 3236
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087052,
+      "loss": 1.3415,
+      "step": 3237
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008704800000000001,
+      "loss": 1.3815,
+      "step": 3238
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087044,
+      "loss": 1.3738,
+      "step": 3239
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008704,
+      "loss": 1.422,
+      "step": 3240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087036,
+      "loss": 1.3692,
+      "step": 3241
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087032,
+      "loss": 1.2649,
+      "step": 3242
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008702800000000001,
+      "loss": 1.3059,
+      "step": 3243
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087024,
+      "loss": 1.3763,
+      "step": 3244
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008702,
+      "loss": 1.2948,
+      "step": 3245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087016,
+      "loss": 1.3448,
+      "step": 3246
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087012,
+      "loss": 1.3813,
+      "step": 3247
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087008,
+      "loss": 1.3627,
+      "step": 3248
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008700400000000001,
+      "loss": 1.4509,
+      "step": 3249
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00087,
+      "loss": 1.3264,
+      "step": 3250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086996,
+      "loss": 1.4076,
+      "step": 3251
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086992,
+      "loss": 1.3351,
+      "step": 3252
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086988,
+      "loss": 1.2858,
+      "step": 3253
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086984,
+      "loss": 1.3749,
+      "step": 3254
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008698,
+      "loss": 1.3504,
+      "step": 3255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086976,
+      "loss": 1.347,
+      "step": 3256
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086972,
+      "loss": 1.3873,
+      "step": 3257
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008696800000000001,
+      "loss": 1.3004,
+      "step": 3258
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086964,
+      "loss": 1.3582,
+      "step": 3259
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008696000000000001,
+      "loss": 1.2818,
+      "step": 3260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086956,
+      "loss": 1.3041,
+      "step": 3261
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008695199999999999,
+      "loss": 1.3174,
+      "step": 3262
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008694800000000001,
+      "loss": 1.4047,
+      "step": 3263
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086944,
+      "loss": 1.3432,
+      "step": 3264
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008694,
+      "loss": 1.2807,
+      "step": 3265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086936,
+      "loss": 1.3998,
+      "step": 3266
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086932,
+      "loss": 1.3677,
+      "step": 3267
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086928,
+      "loss": 1.3156,
+      "step": 3268
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008692400000000001,
+      "loss": 1.379,
+      "step": 3269
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008692,
+      "loss": 1.2945,
+      "step": 3270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008691600000000001,
+      "loss": 1.3549,
+      "step": 3271
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086912,
+      "loss": 1.3576,
+      "step": 3272
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008690799999999999,
+      "loss": 1.4184,
+      "step": 3273
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008690400000000001,
+      "loss": 1.3351,
+      "step": 3274
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000869,
+      "loss": 1.3036,
+      "step": 3275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086896,
+      "loss": 1.2876,
+      "step": 3276
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086892,
+      "loss": 1.3675,
+      "step": 3277
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086888,
+      "loss": 1.3328,
+      "step": 3278
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086884,
+      "loss": 1.3721,
+      "step": 3279
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008688000000000001,
+      "loss": 1.3564,
+      "step": 3280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086876,
+      "loss": 1.3862,
+      "step": 3281
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086872,
+      "loss": 1.3399,
+      "step": 3282
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086868,
+      "loss": 1.4168,
+      "step": 3283
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086864,
+      "loss": 1.3069,
+      "step": 3284
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008686000000000001,
+      "loss": 1.4308,
+      "step": 3285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086856,
+      "loss": 1.3068,
+      "step": 3286
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086852,
+      "loss": 1.3967,
+      "step": 3287
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086848,
+      "loss": 1.3341,
+      "step": 3288
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086844,
+      "loss": 1.2686,
+      "step": 3289
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008684,
+      "loss": 1.3036,
+      "step": 3290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086836,
+      "loss": 1.3163,
+      "step": 3291
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086832,
+      "loss": 1.3551,
+      "step": 3292
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086828,
+      "loss": 1.2504,
+      "step": 3293
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086824,
+      "loss": 1.326,
+      "step": 3294
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008682,
+      "loss": 1.3762,
+      "step": 3295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008681600000000001,
+      "loss": 1.3433,
+      "step": 3296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086812,
+      "loss": 1.3823,
+      "step": 3297
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086808,
+      "loss": 1.2888,
+      "step": 3298
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086804,
+      "loss": 1.4278,
+      "step": 3299
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008680000000000001,
+      "loss": 1.425,
+      "step": 3300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086796,
+      "loss": 1.354,
+      "step": 3301
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086792,
+      "loss": 1.3345,
+      "step": 3302
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086788,
+      "loss": 1.3253,
+      "step": 3303
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008678399999999999,
+      "loss": 1.3575,
+      "step": 3304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008678000000000001,
+      "loss": 1.2524,
+      "step": 3305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086776,
+      "loss": 1.3161,
+      "step": 3306
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008677200000000001,
+      "loss": 1.3625,
+      "step": 3307
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086768,
+      "loss": 1.3595,
+      "step": 3308
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086764,
+      "loss": 1.3594,
+      "step": 3309
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008676000000000001,
+      "loss": 1.2987,
+      "step": 3310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086756,
+      "loss": 1.3036,
+      "step": 3311
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086752,
+      "loss": 1.3524,
+      "step": 3312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086748,
+      "loss": 1.2381,
+      "step": 3313
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086744,
+      "loss": 1.2772,
+      "step": 3314
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008673999999999999,
+      "loss": 1.2917,
+      "step": 3315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008673600000000001,
+      "loss": 1.3822,
+      "step": 3316
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086732,
+      "loss": 1.2283,
+      "step": 3317
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008672800000000001,
+      "loss": 1.3607,
+      "step": 3318
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086724,
+      "loss": 1.3069,
+      "step": 3319
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008671999999999999,
+      "loss": 1.2921,
+      "step": 3320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008671600000000001,
+      "loss": 1.3751,
+      "step": 3321
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086712,
+      "loss": 1.3071,
+      "step": 3322
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086708,
+      "loss": 1.3346,
+      "step": 3323
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086704,
+      "loss": 1.3287,
+      "step": 3324
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000867,
+      "loss": 1.363,
+      "step": 3325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086696,
+      "loss": 1.2618,
+      "step": 3326
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008669200000000001,
+      "loss": 1.4424,
+      "step": 3327
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086688,
+      "loss": 1.3709,
+      "step": 3328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008668400000000001,
+      "loss": 1.2929,
+      "step": 3329
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008668,
+      "loss": 1.3286,
+      "step": 3330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086676,
+      "loss": 1.4059,
+      "step": 3331
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008667200000000001,
+      "loss": 1.4009,
+      "step": 3332
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086668,
+      "loss": 1.3645,
+      "step": 3333
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086664,
+      "loss": 1.3448,
+      "step": 3334
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008666,
+      "loss": 1.3876,
+      "step": 3335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086656,
+      "loss": 1.3003,
+      "step": 3336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086652,
+      "loss": 1.3247,
+      "step": 3337
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008664800000000001,
+      "loss": 1.2831,
+      "step": 3338
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086644,
+      "loss": 1.3795,
+      "step": 3339
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008663999999999999,
+      "loss": 1.2529,
+      "step": 3340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086636,
+      "loss": 1.2782,
+      "step": 3341
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086632,
+      "loss": 1.3209,
+      "step": 3342
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008662800000000001,
+      "loss": 1.3183,
+      "step": 3343
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086624,
+      "loss": 1.2737,
+      "step": 3344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008662,
+      "loss": 1.4091,
+      "step": 3345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086616,
+      "loss": 1.3662,
+      "step": 3346
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008661200000000001,
+      "loss": 1.2902,
+      "step": 3347
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086608,
+      "loss": 1.2825,
+      "step": 3348
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008660400000000001,
+      "loss": 1.2689,
+      "step": 3349
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000866,
+      "loss": 1.3294,
+      "step": 3350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008659599999999999,
+      "loss": 1.4068,
+      "step": 3351
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008659200000000001,
+      "loss": 1.4037,
+      "step": 3352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086588,
+      "loss": 1.3647,
+      "step": 3353
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008658400000000001,
+      "loss": 1.3468,
+      "step": 3354
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008658,
+      "loss": 1.317,
+      "step": 3355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086576,
+      "loss": 1.3344,
+      "step": 3356
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086572,
+      "loss": 1.3517,
+      "step": 3357
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008656800000000001,
+      "loss": 1.3602,
+      "step": 3358
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086564,
+      "loss": 1.4157,
+      "step": 3359
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008656,
+      "loss": 1.3042,
+      "step": 3360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086556,
+      "loss": 1.3925,
+      "step": 3361
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008655199999999999,
+      "loss": 1.2234,
+      "step": 3362
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008654800000000001,
+      "loss": 1.3685,
+      "step": 3363
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086544,
+      "loss": 1.3274,
+      "step": 3364
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008654,
+      "loss": 1.2936,
+      "step": 3365
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086536,
+      "loss": 1.2779,
+      "step": 3366
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086532,
+      "loss": 1.3129,
+      "step": 3367
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008652800000000001,
+      "loss": 1.2295,
+      "step": 3368
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086524,
+      "loss": 1.2789,
+      "step": 3369
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008652,
+      "loss": 1.3121,
+      "step": 3370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086516,
+      "loss": 1.3846,
+      "step": 3371
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086512,
+      "loss": 1.3609,
+      "step": 3372
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086508,
+      "loss": 1.2471,
+      "step": 3373
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008650400000000001,
+      "loss": 1.3935,
+      "step": 3374
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000865,
+      "loss": 1.2703,
+      "step": 3375
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086496,
+      "loss": 1.3442,
+      "step": 3376
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086492,
+      "loss": 1.3403,
+      "step": 3377
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086488,
+      "loss": 1.3558,
+      "step": 3378
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008648400000000001,
+      "loss": 1.3486,
+      "step": 3379
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008648,
+      "loss": 1.2986,
+      "step": 3380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086476,
+      "loss": 1.3029,
+      "step": 3381
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086472,
+      "loss": 1.3346,
+      "step": 3382
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086468,
+      "loss": 1.3596,
+      "step": 3383
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086464,
+      "loss": 1.3242,
+      "step": 3384
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008646000000000001,
+      "loss": 1.3651,
+      "step": 3385
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086456,
+      "loss": 1.3607,
+      "step": 3386
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086452,
+      "loss": 1.3088,
+      "step": 3387
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086448,
+      "loss": 1.3145,
+      "step": 3388
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086444,
+      "loss": 1.3473,
+      "step": 3389
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008644,
+      "loss": 1.2643,
+      "step": 3390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086436,
+      "loss": 1.3258,
+      "step": 3391
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086432,
+      "loss": 1.2657,
+      "step": 3392
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086428,
+      "loss": 1.3812,
+      "step": 3393
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008642400000000001,
+      "loss": 1.3788,
+      "step": 3394
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008642,
+      "loss": 1.2818,
+      "step": 3395
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008641600000000001,
+      "loss": 1.282,
+      "step": 3396
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086412,
+      "loss": 1.332,
+      "step": 3397
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008640799999999999,
+      "loss": 1.324,
+      "step": 3398
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008640400000000001,
+      "loss": 1.2176,
+      "step": 3399
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000864,
+      "loss": 1.3299,
+      "step": 3400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086396,
+      "loss": 1.2944,
+      "step": 3401
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086392,
+      "loss": 1.3111,
+      "step": 3402
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086388,
+      "loss": 1.404,
+      "step": 3403
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086384,
+      "loss": 1.2922,
+      "step": 3404
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008638000000000001,
+      "loss": 1.3434,
+      "step": 3405
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086376,
+      "loss": 1.3307,
+      "step": 3406
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008637200000000001,
+      "loss": 1.2732,
+      "step": 3407
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086368,
+      "loss": 1.275,
+      "step": 3408
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008636399999999999,
+      "loss": 1.3163,
+      "step": 3409
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008636000000000001,
+      "loss": 1.2618,
+      "step": 3410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086356,
+      "loss": 1.2465,
+      "step": 3411
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086352,
+      "loss": 1.2127,
+      "step": 3412
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086348,
+      "loss": 1.3181,
+      "step": 3413
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086344,
+      "loss": 1.3957,
+      "step": 3414
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008634,
+      "loss": 1.2925,
+      "step": 3415
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008633600000000001,
+      "loss": 1.3725,
+      "step": 3416
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086332,
+      "loss": 1.2933,
+      "step": 3417
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086328,
+      "loss": 1.303,
+      "step": 3418
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00086324,
+      "loss": 1.3311,
+      "step": 3419
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0008632,
+      "loss": 1.4049,
+      "step": 3420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008631600000000001,
+      "loss": 1.397,
+      "step": 3421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086312,
+      "loss": 1.3772,
+      "step": 3422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086308,
+      "loss": 1.3348,
+      "step": 3423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086304,
+      "loss": 1.3634,
+      "step": 3424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000863,
+      "loss": 1.2499,
+      "step": 3425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086296,
+      "loss": 1.3574,
+      "step": 3426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086292,
+      "loss": 1.3287,
+      "step": 3427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086288,
+      "loss": 1.3021,
+      "step": 3428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086284,
+      "loss": 1.4192,
+      "step": 3429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008628,
+      "loss": 1.3206,
+      "step": 3430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086276,
+      "loss": 1.3406,
+      "step": 3431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008627200000000001,
+      "loss": 1.3481,
+      "step": 3432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086268,
+      "loss": 1.2293,
+      "step": 3433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086264,
+      "loss": 1.2693,
+      "step": 3434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008626,
+      "loss": 1.2086,
+      "step": 3435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008625600000000001,
+      "loss": 1.2919,
+      "step": 3436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086252,
+      "loss": 1.3351,
+      "step": 3437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086248,
+      "loss": 1.2066,
+      "step": 3438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086244,
+      "loss": 1.3662,
+      "step": 3439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008624,
+      "loss": 1.307,
+      "step": 3440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008623600000000001,
+      "loss": 1.2911,
+      "step": 3441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086232,
+      "loss": 1.2813,
+      "step": 3442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008622800000000001,
+      "loss": 1.3321,
+      "step": 3443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086224,
+      "loss": 1.3225,
+      "step": 3444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008622,
+      "loss": 1.3251,
+      "step": 3445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086216,
+      "loss": 1.307,
+      "step": 3446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086212,
+      "loss": 1.3343,
+      "step": 3447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086208,
+      "loss": 1.3158,
+      "step": 3448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086204,
+      "loss": 1.364,
+      "step": 3449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000862,
+      "loss": 1.3508,
+      "step": 3450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008619599999999999,
+      "loss": 1.339,
+      "step": 3451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008619200000000001,
+      "loss": 1.2637,
+      "step": 3452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086188,
+      "loss": 1.3191,
+      "step": 3453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008618400000000001,
+      "loss": 1.3637,
+      "step": 3454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008618,
+      "loss": 1.3173,
+      "step": 3455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008617599999999999,
+      "loss": 1.37,
+      "step": 3456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008617200000000001,
+      "loss": 1.3395,
+      "step": 3457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086168,
+      "loss": 1.326,
+      "step": 3458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086164,
+      "loss": 1.2774,
+      "step": 3459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008616,
+      "loss": 1.2586,
+      "step": 3460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086156,
+      "loss": 1.3277,
+      "step": 3461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086152,
+      "loss": 1.283,
+      "step": 3462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008614800000000001,
+      "loss": 1.4082,
+      "step": 3463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086144,
+      "loss": 1.2696,
+      "step": 3464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008614000000000001,
+      "loss": 1.3268,
+      "step": 3465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086136,
+      "loss": 1.2822,
+      "step": 3466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086132,
+      "loss": 1.349,
+      "step": 3467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008612800000000001,
+      "loss": 1.3121,
+      "step": 3468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086124,
+      "loss": 1.3722,
+      "step": 3469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008612,
+      "loss": 1.3517,
+      "step": 3470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086116,
+      "loss": 1.4042,
+      "step": 3471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086112,
+      "loss": 1.3201,
+      "step": 3472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086108,
+      "loss": 1.302,
+      "step": 3473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008610400000000001,
+      "loss": 1.3304,
+      "step": 3474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000861,
+      "loss": 1.3476,
+      "step": 3475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008609599999999999,
+      "loss": 1.2769,
+      "step": 3476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086092,
+      "loss": 1.3378,
+      "step": 3477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086088,
+      "loss": 1.3427,
+      "step": 3478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008608400000000001,
+      "loss": 1.4124,
+      "step": 3479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008608,
+      "loss": 1.3126,
+      "step": 3480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086076,
+      "loss": 1.2925,
+      "step": 3481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086072,
+      "loss": 1.3127,
+      "step": 3482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008606800000000001,
+      "loss": 1.339,
+      "step": 3483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086064,
+      "loss": 1.3309,
+      "step": 3484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008606000000000001,
+      "loss": 1.2509,
+      "step": 3485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086056,
+      "loss": 1.3232,
+      "step": 3486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008605199999999999,
+      "loss": 1.3135,
+      "step": 3487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008604800000000001,
+      "loss": 1.4261,
+      "step": 3488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086044,
+      "loss": 1.3387,
+      "step": 3489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008604000000000001,
+      "loss": 1.3566,
+      "step": 3490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086036,
+      "loss": 1.3215,
+      "step": 3491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086032,
+      "loss": 1.2722,
+      "step": 3492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086028,
+      "loss": 1.3504,
+      "step": 3493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008602400000000001,
+      "loss": 1.3238,
+      "step": 3494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008602,
+      "loss": 1.3273,
+      "step": 3495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086016,
+      "loss": 1.3284,
+      "step": 3496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086012,
+      "loss": 1.276,
+      "step": 3497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008600799999999999,
+      "loss": 1.3602,
+      "step": 3498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008600400000000001,
+      "loss": 1.3262,
+      "step": 3499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00086,
+      "loss": 1.3068,
+      "step": 3500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085996,
+      "loss": 1.3297,
+      "step": 3501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085992,
+      "loss": 1.2607,
+      "step": 3502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085988,
+      "loss": 1.3389,
+      "step": 3503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008598400000000001,
+      "loss": 1.2655,
+      "step": 3504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008598,
+      "loss": 1.3137,
+      "step": 3505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085976,
+      "loss": 1.3325,
+      "step": 3506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085972,
+      "loss": 1.2913,
+      "step": 3507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085968,
+      "loss": 1.2919,
+      "step": 3508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085964,
+      "loss": 1.2664,
+      "step": 3509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008596000000000001,
+      "loss": 1.3306,
+      "step": 3510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085956,
+      "loss": 1.3215,
+      "step": 3511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085952,
+      "loss": 1.4088,
+      "step": 3512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085948,
+      "loss": 1.2274,
+      "step": 3513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085944,
+      "loss": 1.4181,
+      "step": 3514
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008594000000000001,
+      "loss": 1.3119,
+      "step": 3515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085936,
+      "loss": 1.3175,
+      "step": 3516
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085932,
+      "loss": 1.3967,
+      "step": 3517
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085928,
+      "loss": 1.2886,
+      "step": 3518
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085924,
+      "loss": 1.2913,
+      "step": 3519
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008592,
+      "loss": 1.3512,
+      "step": 3520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008591600000000001,
+      "loss": 1.368,
+      "step": 3521
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085912,
+      "loss": 1.3524,
+      "step": 3522
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085908,
+      "loss": 1.3067,
+      "step": 3523
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085904,
+      "loss": 1.3443,
+      "step": 3524
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000859,
+      "loss": 1.3398,
+      "step": 3525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085896,
+      "loss": 1.3468,
+      "step": 3526
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085892,
+      "loss": 1.3113,
+      "step": 3527
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085888,
+      "loss": 1.3403,
+      "step": 3528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085884,
+      "loss": 1.4238,
+      "step": 3529
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008588000000000001,
+      "loss": 1.2256,
+      "step": 3530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085876,
+      "loss": 1.3384,
+      "step": 3531
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008587200000000001,
+      "loss": 1.239,
+      "step": 3532
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085868,
+      "loss": 1.3168,
+      "step": 3533
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008586399999999999,
+      "loss": 1.3865,
+      "step": 3534
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008586,
+      "loss": 1.2682,
+      "step": 3535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085856,
+      "loss": 1.2701,
+      "step": 3536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085852,
+      "loss": 1.3689,
+      "step": 3537
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085848,
+      "loss": 1.3002,
+      "step": 3538
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085844,
+      "loss": 1.4118,
+      "step": 3539
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008584,
+      "loss": 1.3941,
+      "step": 3540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008583600000000001,
+      "loss": 1.34,
+      "step": 3541
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085832,
+      "loss": 1.3056,
+      "step": 3542
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008582800000000001,
+      "loss": 1.3644,
+      "step": 3543
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085824,
+      "loss": 1.2942,
+      "step": 3544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008581999999999999,
+      "loss": 1.3545,
+      "step": 3545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008581600000000001,
+      "loss": 1.4165,
+      "step": 3546
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085812,
+      "loss": 1.2486,
+      "step": 3547
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085808,
+      "loss": 1.3148,
+      "step": 3548
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085804,
+      "loss": 1.4861,
+      "step": 3549
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000858,
+      "loss": 1.3507,
+      "step": 3550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085796,
+      "loss": 1.2645,
+      "step": 3551
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008579200000000001,
+      "loss": 1.3092,
+      "step": 3552
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085788,
+      "loss": 1.3217,
+      "step": 3553
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085784,
+      "loss": 1.2784,
+      "step": 3554
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008578,
+      "loss": 1.2167,
+      "step": 3555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085776,
+      "loss": 1.3276,
+      "step": 3556
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008577200000000001,
+      "loss": 1.4145,
+      "step": 3557
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085768,
+      "loss": 1.3046,
+      "step": 3558
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085764,
+      "loss": 1.3204,
+      "step": 3559
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008576,
+      "loss": 1.3984,
+      "step": 3560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085756,
+      "loss": 1.2993,
+      "step": 3561
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085752,
+      "loss": 1.2318,
+      "step": 3562
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008574800000000001,
+      "loss": 1.2641,
+      "step": 3563
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085744,
+      "loss": 1.3006,
+      "step": 3564
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008574,
+      "loss": 1.3516,
+      "step": 3565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085736,
+      "loss": 1.3159,
+      "step": 3566
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085732,
+      "loss": 1.3234,
+      "step": 3567
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008572800000000001,
+      "loss": 1.3473,
+      "step": 3568
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085724,
+      "loss": 1.397,
+      "step": 3569
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008572,
+      "loss": 1.3332,
+      "step": 3570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085716,
+      "loss": 1.3245,
+      "step": 3571
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008571200000000001,
+      "loss": 1.4709,
+      "step": 3572
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085708,
+      "loss": 1.3946,
+      "step": 3573
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085704,
+      "loss": 1.3444,
+      "step": 3574
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000857,
+      "loss": 1.2205,
+      "step": 3575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085696,
+      "loss": 1.3405,
+      "step": 3576
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008569200000000001,
+      "loss": 1.3277,
+      "step": 3577
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085688,
+      "loss": 1.328,
+      "step": 3578
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008568400000000001,
+      "loss": 1.2443,
+      "step": 3579
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008568,
+      "loss": 1.3089,
+      "step": 3580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085676,
+      "loss": 1.2177,
+      "step": 3581
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085672,
+      "loss": 1.3481,
+      "step": 3582
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085668,
+      "loss": 1.3786,
+      "step": 3583
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085664,
+      "loss": 1.3239,
+      "step": 3584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008566,
+      "loss": 1.3264,
+      "step": 3585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085656,
+      "loss": 1.3769,
+      "step": 3586
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008565199999999999,
+      "loss": 1.2719,
+      "step": 3587
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008564800000000001,
+      "loss": 1.2771,
+      "step": 3588
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085644,
+      "loss": 1.3678,
+      "step": 3589
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008564000000000001,
+      "loss": 1.3428,
+      "step": 3590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085636,
+      "loss": 1.3297,
+      "step": 3591
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085632,
+      "loss": 1.2284,
+      "step": 3592
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008562800000000001,
+      "loss": 1.3483,
+      "step": 3593
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085624,
+      "loss": 1.3508,
+      "step": 3594
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008562,
+      "loss": 1.3232,
+      "step": 3595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085616,
+      "loss": 1.3496,
+      "step": 3596
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085612,
+      "loss": 1.2923,
+      "step": 3597
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085608,
+      "loss": 1.3732,
+      "step": 3598
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008560400000000001,
+      "loss": 1.3033,
+      "step": 3599
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000856,
+      "loss": 1.2823,
+      "step": 3600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008559600000000001,
+      "loss": 1.3049,
+      "step": 3601
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085592,
+      "loss": 1.3573,
+      "step": 3602
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008558799999999999,
+      "loss": 1.311,
+      "step": 3603
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008558400000000001,
+      "loss": 1.2976,
+      "step": 3604
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008558,
+      "loss": 1.3592,
+      "step": 3605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085576,
+      "loss": 1.285,
+      "step": 3606
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085572,
+      "loss": 1.3394,
+      "step": 3607
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085568,
+      "loss": 1.2688,
+      "step": 3608
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085564,
+      "loss": 1.3328,
+      "step": 3609
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008556000000000001,
+      "loss": 1.351,
+      "step": 3610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085556,
+      "loss": 1.2264,
+      "step": 3611
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008555199999999999,
+      "loss": 1.3238,
+      "step": 3612
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085548,
+      "loss": 1.1833,
+      "step": 3613
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085544,
+      "loss": 1.2552,
+      "step": 3614
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008554000000000001,
+      "loss": 1.3422,
+      "step": 3615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085536,
+      "loss": 1.3536,
+      "step": 3616
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085532,
+      "loss": 1.3697,
+      "step": 3617
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085528,
+      "loss": 1.38,
+      "step": 3618
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008552400000000001,
+      "loss": 1.1822,
+      "step": 3619
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008552,
+      "loss": 1.297,
+      "step": 3620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008551600000000001,
+      "loss": 1.3245,
+      "step": 3621
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085512,
+      "loss": 1.2935,
+      "step": 3622
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008550799999999999,
+      "loss": 1.3434,
+      "step": 3623
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085504,
+      "loss": 1.2851,
+      "step": 3624
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000855,
+      "loss": 1.313,
+      "step": 3625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008549600000000001,
+      "loss": 1.2721,
+      "step": 3626
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085492,
+      "loss": 1.3653,
+      "step": 3627
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085488,
+      "loss": 1.2954,
+      "step": 3628
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085484,
+      "loss": 1.2734,
+      "step": 3629
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008548000000000001,
+      "loss": 1.2429,
+      "step": 3630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085476,
+      "loss": 1.2746,
+      "step": 3631
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085472,
+      "loss": 1.2285,
+      "step": 3632
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085468,
+      "loss": 1.3122,
+      "step": 3633
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008546399999999999,
+      "loss": 1.3558,
+      "step": 3634
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008546000000000001,
+      "loss": 1.3161,
+      "step": 3635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085456,
+      "loss": 1.2952,
+      "step": 3636
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085452,
+      "loss": 1.3049,
+      "step": 3637
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085448,
+      "loss": 1.3718,
+      "step": 3638
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085444,
+      "loss": 1.2866,
+      "step": 3639
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008544000000000001,
+      "loss": 1.2846,
+      "step": 3640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008543600000000001,
+      "loss": 1.2315,
+      "step": 3641
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085432,
+      "loss": 1.377,
+      "step": 3642
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085428,
+      "loss": 1.2279,
+      "step": 3643
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085424,
+      "loss": 1.1876,
+      "step": 3644
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008542,
+      "loss": 1.2413,
+      "step": 3645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008541600000000001,
+      "loss": 1.2942,
+      "step": 3646
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085412,
+      "loss": 1.2168,
+      "step": 3647
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085408,
+      "loss": 1.2748,
+      "step": 3648
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085404,
+      "loss": 1.3347,
+      "step": 3649
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000854,
+      "loss": 1.3027,
+      "step": 3650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008539600000000001,
+      "loss": 1.3056,
+      "step": 3651
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085392,
+      "loss": 1.3264,
+      "step": 3652
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085388,
+      "loss": 1.3094,
+      "step": 3653
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085384,
+      "loss": 1.2604,
+      "step": 3654
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008538,
+      "loss": 1.2875,
+      "step": 3655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085376,
+      "loss": 1.3018,
+      "step": 3656
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008537200000000001,
+      "loss": 1.2913,
+      "step": 3657
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085368,
+      "loss": 1.3241,
+      "step": 3658
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085364,
+      "loss": 1.2717,
+      "step": 3659
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008536,
+      "loss": 1.252,
+      "step": 3660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085356,
+      "loss": 1.2762,
+      "step": 3661
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085352,
+      "loss": 1.4988,
+      "step": 3662
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085348,
+      "loss": 1.3532,
+      "step": 3663
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085344,
+      "loss": 1.2902,
+      "step": 3664
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008534,
+      "loss": 1.3119,
+      "step": 3665
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008533600000000001,
+      "loss": 1.3612,
+      "step": 3666
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085332,
+      "loss": 1.284,
+      "step": 3667
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008532800000000001,
+      "loss": 1.3302,
+      "step": 3668
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085324,
+      "loss": 1.2497,
+      "step": 3669
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0008531999999999999,
+      "loss": 1.2694,
+      "step": 3670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085316,
+      "loss": 1.3261,
+      "step": 3671
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085312,
+      "loss": 1.3284,
+      "step": 3672
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085308,
+      "loss": 1.3042,
+      "step": 3673
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00085304,
+      "loss": 1.3785,
+      "step": 3674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000853,
+      "loss": 1.3182,
+      "step": 3675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085296,
+      "loss": 1.3656,
+      "step": 3676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008529200000000001,
+      "loss": 1.4174,
+      "step": 3677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085288,
+      "loss": 1.4273,
+      "step": 3678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008528400000000001,
+      "loss": 1.3546,
+      "step": 3679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008528,
+      "loss": 1.3629,
+      "step": 3680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008527599999999999,
+      "loss": 1.251,
+      "step": 3681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008527200000000001,
+      "loss": 1.3675,
+      "step": 3682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085268,
+      "loss": 1.3473,
+      "step": 3683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085264,
+      "loss": 1.2577,
+      "step": 3684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008526,
+      "loss": 1.3039,
+      "step": 3685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085256,
+      "loss": 1.3209,
+      "step": 3686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085252,
+      "loss": 1.2949,
+      "step": 3687
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008524800000000001,
+      "loss": 1.3483,
+      "step": 3688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085244,
+      "loss": 1.2425,
+      "step": 3689
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008524,
+      "loss": 1.3496,
+      "step": 3690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085236,
+      "loss": 1.3332,
+      "step": 3691
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008523199999999999,
+      "loss": 1.3903,
+      "step": 3692
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008522800000000001,
+      "loss": 1.2791,
+      "step": 3693
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085224,
+      "loss": 1.2674,
+      "step": 3694
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008522,
+      "loss": 1.331,
+      "step": 3695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085216,
+      "loss": 1.3504,
+      "step": 3696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085212,
+      "loss": 1.3272,
+      "step": 3697
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085208,
+      "loss": 1.3377,
+      "step": 3698
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008520400000000001,
+      "loss": 1.2432,
+      "step": 3699
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000852,
+      "loss": 1.3448,
+      "step": 3700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085196,
+      "loss": 1.3685,
+      "step": 3701
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085192,
+      "loss": 1.2034,
+      "step": 3702
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085188,
+      "loss": 1.3208,
+      "step": 3703
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008518400000000001,
+      "loss": 1.2465,
+      "step": 3704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008518,
+      "loss": 1.3621,
+      "step": 3705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085176,
+      "loss": 1.3039,
+      "step": 3706
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085172,
+      "loss": 1.1913,
+      "step": 3707
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008516800000000001,
+      "loss": 1.321,
+      "step": 3708
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085164,
+      "loss": 1.317,
+      "step": 3709
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008516,
+      "loss": 1.4295,
+      "step": 3710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085156,
+      "loss": 1.296,
+      "step": 3711
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085152,
+      "loss": 1.2784,
+      "step": 3712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085148,
+      "loss": 1.2641,
+      "step": 3713
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085144,
+      "loss": 1.3349,
+      "step": 3714
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008514000000000001,
+      "loss": 1.3689,
+      "step": 3715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085136,
+      "loss": 1.3245,
+      "step": 3716
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085132,
+      "loss": 1.357,
+      "step": 3717
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085128,
+      "loss": 1.2654,
+      "step": 3718
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085124,
+      "loss": 1.2873,
+      "step": 3719
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008512,
+      "loss": 1.3285,
+      "step": 3720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085116,
+      "loss": 1.202,
+      "step": 3721
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085112,
+      "loss": 1.2383,
+      "step": 3722
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008510799999999999,
+      "loss": 1.3921,
+      "step": 3723
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008510400000000001,
+      "loss": 1.2732,
+      "step": 3724
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000851,
+      "loss": 1.3857,
+      "step": 3725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008509600000000001,
+      "loss": 1.2917,
+      "step": 3726
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085092,
+      "loss": 1.3325,
+      "step": 3727
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085088,
+      "loss": 1.2613,
+      "step": 3728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008508400000000001,
+      "loss": 1.3183,
+      "step": 3729
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008508,
+      "loss": 1.3219,
+      "step": 3730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085076,
+      "loss": 1.3592,
+      "step": 3731
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085072,
+      "loss": 1.3025,
+      "step": 3732
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085068,
+      "loss": 1.242,
+      "step": 3733
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085064,
+      "loss": 1.3374,
+      "step": 3734
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008506000000000001,
+      "loss": 1.412,
+      "step": 3735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085056,
+      "loss": 1.1977,
+      "step": 3736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008505200000000001,
+      "loss": 1.2512,
+      "step": 3737
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085048,
+      "loss": 1.279,
+      "step": 3738
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008504399999999999,
+      "loss": 1.2879,
+      "step": 3739
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008504000000000001,
+      "loss": 1.2532,
+      "step": 3740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085036,
+      "loss": 1.2859,
+      "step": 3741
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085032,
+      "loss": 1.3114,
+      "step": 3742
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085028,
+      "loss": 1.3732,
+      "step": 3743
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085024,
+      "loss": 1.3533,
+      "step": 3744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008502,
+      "loss": 1.2278,
+      "step": 3745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008501600000000001,
+      "loss": 1.3491,
+      "step": 3746
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085012,
+      "loss": 1.2954,
+      "step": 3747
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008500799999999999,
+      "loss": 1.3167,
+      "step": 3748
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085004,
+      "loss": 1.4205,
+      "step": 3749
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00085,
+      "loss": 1.242,
+      "step": 3750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008499600000000001,
+      "loss": 1.3203,
+      "step": 3751
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084992,
+      "loss": 1.2637,
+      "step": 3752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084988,
+      "loss": 1.3641,
+      "step": 3753
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084984,
+      "loss": 1.3325,
+      "step": 3754
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008498000000000001,
+      "loss": 1.3895,
+      "step": 3755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084976,
+      "loss": 1.2469,
+      "step": 3756
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008497200000000001,
+      "loss": 1.3621,
+      "step": 3757
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084968,
+      "loss": 1.3147,
+      "step": 3758
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008496399999999999,
+      "loss": 1.3557,
+      "step": 3759
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008496,
+      "loss": 1.2942,
+      "step": 3760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084956,
+      "loss": 1.2845,
+      "step": 3761
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008495200000000001,
+      "loss": 1.3216,
+      "step": 3762
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084948,
+      "loss": 1.3031,
+      "step": 3763
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084944,
+      "loss": 1.2931,
+      "step": 3764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008494,
+      "loss": 1.3224,
+      "step": 3765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008493600000000001,
+      "loss": 1.3271,
+      "step": 3766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084932,
+      "loss": 1.348,
+      "step": 3767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084928,
+      "loss": 1.2932,
+      "step": 3768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084924,
+      "loss": 1.2691,
+      "step": 3769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008491999999999999,
+      "loss": 1.2274,
+      "step": 3770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008491600000000001,
+      "loss": 1.3135,
+      "step": 3771
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084912,
+      "loss": 1.2509,
+      "step": 3772
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084908,
+      "loss": 1.3274,
+      "step": 3773
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084904,
+      "loss": 1.3743,
+      "step": 3774
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000849,
+      "loss": 1.3652,
+      "step": 3775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008489600000000001,
+      "loss": 1.3317,
+      "step": 3776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008489200000000001,
+      "loss": 1.2335,
+      "step": 3777
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084888,
+      "loss": 1.2842,
+      "step": 3778
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084884,
+      "loss": 1.2462,
+      "step": 3779
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008488,
+      "loss": 1.2297,
+      "step": 3780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008487599999999999,
+      "loss": 1.33,
+      "step": 3781
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008487200000000001,
+      "loss": 1.1973,
+      "step": 3782
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084868,
+      "loss": 1.2904,
+      "step": 3783
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084864,
+      "loss": 1.3637,
+      "step": 3784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008486,
+      "loss": 1.2262,
+      "step": 3785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084856,
+      "loss": 1.3201,
+      "step": 3786
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008485200000000001,
+      "loss": 1.2287,
+      "step": 3787
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084848,
+      "loss": 1.3276,
+      "step": 3788
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084844,
+      "loss": 1.3383,
+      "step": 3789
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008484,
+      "loss": 1.2789,
+      "step": 3790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084836,
+      "loss": 1.2678,
+      "step": 3791
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084832,
+      "loss": 1.3935,
+      "step": 3792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008482800000000001,
+      "loss": 1.3052,
+      "step": 3793
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084824,
+      "loss": 1.3726,
+      "step": 3794
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008482,
+      "loss": 1.2338,
+      "step": 3795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084816,
+      "loss": 1.3153,
+      "step": 3796
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084812,
+      "loss": 1.4097,
+      "step": 3797
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084808,
+      "loss": 1.2922,
+      "step": 3798
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084804,
+      "loss": 1.3096,
+      "step": 3799
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000848,
+      "loss": 1.2772,
+      "step": 3800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084796,
+      "loss": 1.2743,
+      "step": 3801
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084792,
+      "loss": 1.3236,
+      "step": 3802
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084788,
+      "loss": 1.2656,
+      "step": 3803
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008478400000000001,
+      "loss": 1.2513,
+      "step": 3804
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008478,
+      "loss": 1.3492,
+      "step": 3805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084776,
+      "loss": 1.3039,
+      "step": 3806
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084772,
+      "loss": 1.3432,
+      "step": 3807
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084768,
+      "loss": 1.2586,
+      "step": 3808
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084764,
+      "loss": 1.267,
+      "step": 3809
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008476,
+      "loss": 1.3789,
+      "step": 3810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084756,
+      "loss": 1.2839,
+      "step": 3811
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084752,
+      "loss": 1.2827,
+      "step": 3812
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008474800000000001,
+      "loss": 1.3735,
+      "step": 3813
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084744,
+      "loss": 1.34,
+      "step": 3814
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008474000000000001,
+      "loss": 1.3831,
+      "step": 3815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084736,
+      "loss": 1.3393,
+      "step": 3816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008473199999999999,
+      "loss": 1.1841,
+      "step": 3817
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008472800000000001,
+      "loss": 1.2877,
+      "step": 3818
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084724,
+      "loss": 1.3502,
+      "step": 3819
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008472,
+      "loss": 1.3511,
+      "step": 3820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084716,
+      "loss": 1.3029,
+      "step": 3821
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084712,
+      "loss": 1.3346,
+      "step": 3822
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008470800000000001,
+      "loss": 1.2687,
+      "step": 3823
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008470400000000001,
+      "loss": 1.3245,
+      "step": 3824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000847,
+      "loss": 1.3035,
+      "step": 3825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008469600000000001,
+      "loss": 1.2869,
+      "step": 3826
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084692,
+      "loss": 1.3201,
+      "step": 3827
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008468799999999999,
+      "loss": 1.2735,
+      "step": 3828
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008468400000000001,
+      "loss": 1.2026,
+      "step": 3829
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008468,
+      "loss": 1.2644,
+      "step": 3830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084676,
+      "loss": 1.2956,
+      "step": 3831
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084672,
+      "loss": 1.3263,
+      "step": 3832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084668,
+      "loss": 1.2909,
+      "step": 3833
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084664,
+      "loss": 1.2494,
+      "step": 3834
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008466000000000001,
+      "loss": 1.3351,
+      "step": 3835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084656,
+      "loss": 1.304,
+      "step": 3836
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084652,
+      "loss": 1.2331,
+      "step": 3837
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084648,
+      "loss": 1.3192,
+      "step": 3838
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084644,
+      "loss": 1.2535,
+      "step": 3839
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008464000000000001,
+      "loss": 1.3092,
+      "step": 3840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084636,
+      "loss": 1.3415,
+      "step": 3841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084632,
+      "loss": 1.2411,
+      "step": 3842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084628,
+      "loss": 1.2406,
+      "step": 3843
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008462400000000001,
+      "loss": 1.2599,
+      "step": 3844
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008462,
+      "loss": 1.2992,
+      "step": 3845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084616,
+      "loss": 1.3348,
+      "step": 3846
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084612,
+      "loss": 1.3394,
+      "step": 3847
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084608,
+      "loss": 1.2817,
+      "step": 3848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084604,
+      "loss": 1.4033,
+      "step": 3849
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000846,
+      "loss": 1.3917,
+      "step": 3850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008459600000000001,
+      "loss": 1.3294,
+      "step": 3851
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084592,
+      "loss": 1.3233,
+      "step": 3852
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084588,
+      "loss": 1.2691,
+      "step": 3853
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084584,
+      "loss": 1.2721,
+      "step": 3854
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008458000000000001,
+      "loss": 1.2907,
+      "step": 3855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084576,
+      "loss": 1.2162,
+      "step": 3856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084572,
+      "loss": 1.2648,
+      "step": 3857
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084568,
+      "loss": 1.301,
+      "step": 3858
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008456399999999999,
+      "loss": 1.2159,
+      "step": 3859
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008456000000000001,
+      "loss": 1.2893,
+      "step": 3860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084556,
+      "loss": 1.3147,
+      "step": 3861
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008455200000000001,
+      "loss": 1.2645,
+      "step": 3862
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084548,
+      "loss": 1.2825,
+      "step": 3863
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084544,
+      "loss": 1.2547,
+      "step": 3864
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008454000000000001,
+      "loss": 1.3038,
+      "step": 3865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084536,
+      "loss": 1.3,
+      "step": 3866
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084532,
+      "loss": 1.3225,
+      "step": 3867
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084528,
+      "loss": 1.268,
+      "step": 3868
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084524,
+      "loss": 1.3191,
+      "step": 3869
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008451999999999999,
+      "loss": 1.2055,
+      "step": 3870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008451600000000001,
+      "loss": 1.3225,
+      "step": 3871
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084512,
+      "loss": 1.4227,
+      "step": 3872
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008450800000000001,
+      "loss": 1.2915,
+      "step": 3873
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084504,
+      "loss": 1.2655,
+      "step": 3874
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008449999999999999,
+      "loss": 1.2822,
+      "step": 3875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008449600000000001,
+      "loss": 1.1953,
+      "step": 3876
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084492,
+      "loss": 1.3582,
+      "step": 3877
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084488,
+      "loss": 1.2935,
+      "step": 3878
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084484,
+      "loss": 1.3173,
+      "step": 3879
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008448,
+      "loss": 1.3504,
+      "step": 3880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084476,
+      "loss": 1.3334,
+      "step": 3881
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008447200000000001,
+      "loss": 1.3436,
+      "step": 3882
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084468,
+      "loss": 1.3146,
+      "step": 3883
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084464,
+      "loss": 1.2481,
+      "step": 3884
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008446,
+      "loss": 1.3655,
+      "step": 3885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084456,
+      "loss": 1.4029,
+      "step": 3886
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008445200000000001,
+      "loss": 1.3496,
+      "step": 3887
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084448,
+      "loss": 1.2828,
+      "step": 3888
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084444,
+      "loss": 1.2577,
+      "step": 3889
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008444,
+      "loss": 1.3048,
+      "step": 3890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084436,
+      "loss": 1.3012,
+      "step": 3891
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084432,
+      "loss": 1.3437,
+      "step": 3892
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008442800000000001,
+      "loss": 1.298,
+      "step": 3893
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084424,
+      "loss": 1.3573,
+      "step": 3894
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008441999999999999,
+      "loss": 1.232,
+      "step": 3895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084416,
+      "loss": 1.2827,
+      "step": 3896
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084412,
+      "loss": 1.2553,
+      "step": 3897
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008440800000000001,
+      "loss": 1.251,
+      "step": 3898
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084404,
+      "loss": 1.3726,
+      "step": 3899
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000844,
+      "loss": 1.3774,
+      "step": 3900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084396,
+      "loss": 1.2657,
+      "step": 3901
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008439200000000001,
+      "loss": 1.2908,
+      "step": 3902
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084388,
+      "loss": 1.3631,
+      "step": 3903
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084384,
+      "loss": 1.3296,
+      "step": 3904
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008438,
+      "loss": 1.2959,
+      "step": 3905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008437599999999999,
+      "loss": 1.3497,
+      "step": 3906
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008437200000000001,
+      "loss": 1.2218,
+      "step": 3907
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084368,
+      "loss": 1.3138,
+      "step": 3908
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084364,
+      "loss": 1.2582,
+      "step": 3909
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008436,
+      "loss": 1.2835,
+      "step": 3910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084356,
+      "loss": 1.2271,
+      "step": 3911
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008435200000000001,
+      "loss": 1.2853,
+      "step": 3912
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008434800000000001,
+      "loss": 1.3682,
+      "step": 3913
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084344,
+      "loss": 1.3289,
+      "step": 3914
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008434,
+      "loss": 1.2833,
+      "step": 3915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084336,
+      "loss": 1.2687,
+      "step": 3916
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008433199999999999,
+      "loss": 1.2206,
+      "step": 3917
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008432800000000001,
+      "loss": 1.2727,
+      "step": 3918
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084324,
+      "loss": 1.3041,
+      "step": 3919
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008432,
+      "loss": 1.2959,
+      "step": 3920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084316,
+      "loss": 1.2975,
+      "step": 3921
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084312,
+      "loss": 1.2353,
+      "step": 3922
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0008430800000000001,
+      "loss": 1.3435,
+      "step": 3923
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084304,
+      "loss": 1.256,
+      "step": 3924
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000843,
+      "loss": 1.2738,
+      "step": 3925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084296,
+      "loss": 1.3133,
+      "step": 3926
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00084292,
+      "loss": 1.3017,
+      "step": 3927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084288,
+      "loss": 1.3651,
+      "step": 3928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008428400000000001,
+      "loss": 1.2904,
+      "step": 3929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008428,
+      "loss": 1.2478,
+      "step": 3930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084276,
+      "loss": 1.2992,
+      "step": 3931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084272,
+      "loss": 1.2205,
+      "step": 3932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084268,
+      "loss": 1.2658,
+      "step": 3933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084264,
+      "loss": 1.3483,
+      "step": 3934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008426,
+      "loss": 1.228,
+      "step": 3935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084256,
+      "loss": 1.295,
+      "step": 3936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084252,
+      "loss": 1.2882,
+      "step": 3937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084248,
+      "loss": 1.3422,
+      "step": 3938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084244,
+      "loss": 1.2092,
+      "step": 3939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008424000000000001,
+      "loss": 1.3356,
+      "step": 3940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084236,
+      "loss": 1.2731,
+      "step": 3941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084232,
+      "loss": 1.2949,
+      "step": 3942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084228,
+      "loss": 1.3476,
+      "step": 3943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084224,
+      "loss": 1.1381,
+      "step": 3944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008422,
+      "loss": 1.3474,
+      "step": 3945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084216,
+      "loss": 1.2985,
+      "step": 3946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084212,
+      "loss": 1.306,
+      "step": 3947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084208,
+      "loss": 1.3021,
+      "step": 3948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008420400000000001,
+      "loss": 1.2213,
+      "step": 3949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000842,
+      "loss": 1.3057,
+      "step": 3950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008419600000000001,
+      "loss": 1.2628,
+      "step": 3951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084192,
+      "loss": 1.2016,
+      "step": 3952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008418799999999999,
+      "loss": 1.3106,
+      "step": 3953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008418400000000001,
+      "loss": 1.3222,
+      "step": 3954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008418,
+      "loss": 1.2714,
+      "step": 3955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084176,
+      "loss": 1.2363,
+      "step": 3956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084172,
+      "loss": 1.2943,
+      "step": 3957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084168,
+      "loss": 1.3433,
+      "step": 3958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084164,
+      "loss": 1.2797,
+      "step": 3959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008416000000000001,
+      "loss": 1.2951,
+      "step": 3960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084156,
+      "loss": 1.3249,
+      "step": 3961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008415200000000001,
+      "loss": 1.3113,
+      "step": 3962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084148,
+      "loss": 1.2891,
+      "step": 3963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008414399999999999,
+      "loss": 1.3171,
+      "step": 3964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008414000000000001,
+      "loss": 1.3308,
+      "step": 3965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084136,
+      "loss": 1.3175,
+      "step": 3966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084132,
+      "loss": 1.1719,
+      "step": 3967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084128,
+      "loss": 1.2379,
+      "step": 3968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084124,
+      "loss": 1.283,
+      "step": 3969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008412,
+      "loss": 1.3997,
+      "step": 3970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008411600000000001,
+      "loss": 1.3094,
+      "step": 3971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084112,
+      "loss": 1.3382,
+      "step": 3972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084108,
+      "loss": 1.449,
+      "step": 3973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084104,
+      "loss": 1.2755,
+      "step": 3974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000841,
+      "loss": 1.3233,
+      "step": 3975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008409600000000001,
+      "loss": 1.2203,
+      "step": 3976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084092,
+      "loss": 1.3125,
+      "step": 3977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084088,
+      "loss": 1.311,
+      "step": 3978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084084,
+      "loss": 1.2774,
+      "step": 3979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008408000000000001,
+      "loss": 1.3579,
+      "step": 3980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084076,
+      "loss": 1.2784,
+      "step": 3981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084072,
+      "loss": 1.2943,
+      "step": 3982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084068,
+      "loss": 1.2378,
+      "step": 3983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084064,
+      "loss": 1.3241,
+      "step": 3984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008406,
+      "loss": 1.2959,
+      "step": 3985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084056,
+      "loss": 1.3186,
+      "step": 3986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008405200000000001,
+      "loss": 1.255,
+      "step": 3987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084048,
+      "loss": 1.2767,
+      "step": 3988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084044,
+      "loss": 1.303,
+      "step": 3989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008404,
+      "loss": 1.3187,
+      "step": 3990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008403600000000001,
+      "loss": 1.2496,
+      "step": 3991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084032,
+      "loss": 1.2877,
+      "step": 3992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084028,
+      "loss": 1.3824,
+      "step": 3993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084024,
+      "loss": 1.2747,
+      "step": 3994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008401999999999999,
+      "loss": 1.2299,
+      "step": 3995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008401600000000001,
+      "loss": 1.4037,
+      "step": 3996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084012,
+      "loss": 1.2596,
+      "step": 3997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008400800000000001,
+      "loss": 1.2432,
+      "step": 3998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084004,
+      "loss": 1.3156,
+      "step": 3999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00084,
+      "loss": 1.2934,
+      "step": 4000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008399600000000001,
+      "loss": 1.3199,
+      "step": 4001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083992,
+      "loss": 1.313,
+      "step": 4002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083988,
+      "loss": 1.3207,
+      "step": 4003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083984,
+      "loss": 1.2717,
+      "step": 4004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008398,
+      "loss": 1.3465,
+      "step": 4005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008397599999999999,
+      "loss": 1.2839,
+      "step": 4006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008397200000000001,
+      "loss": 1.2884,
+      "step": 4007
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083968,
+      "loss": 1.2435,
+      "step": 4008
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008396400000000001,
+      "loss": 1.2656,
+      "step": 4009
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008396,
+      "loss": 1.3003,
+      "step": 4010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008395599999999999,
+      "loss": 1.229,
+      "step": 4011
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008395200000000001,
+      "loss": 1.3012,
+      "step": 4012
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083948,
+      "loss": 1.2271,
+      "step": 4013
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083944,
+      "loss": 1.1923,
+      "step": 4014
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008394,
+      "loss": 1.2064,
+      "step": 4015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083936,
+      "loss": 1.3029,
+      "step": 4016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083932,
+      "loss": 1.2409,
+      "step": 4017
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008392800000000001,
+      "loss": 1.3412,
+      "step": 4018
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083924,
+      "loss": 1.1929,
+      "step": 4019
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008392,
+      "loss": 1.2975,
+      "step": 4020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083916,
+      "loss": 1.3669,
+      "step": 4021
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083912,
+      "loss": 1.3221,
+      "step": 4022
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008390800000000001,
+      "loss": 1.2972,
+      "step": 4023
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083904,
+      "loss": 1.32,
+      "step": 4024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000839,
+      "loss": 1.2738,
+      "step": 4025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083896,
+      "loss": 1.3303,
+      "step": 4026
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083892,
+      "loss": 1.1786,
+      "step": 4027
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083888,
+      "loss": 1.3314,
+      "step": 4028
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008388400000000001,
+      "loss": 1.2892,
+      "step": 4029
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008388,
+      "loss": 1.295,
+      "step": 4030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008387599999999999,
+      "loss": 1.1873,
+      "step": 4031
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083872,
+      "loss": 1.3152,
+      "step": 4032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083868,
+      "loss": 1.342,
+      "step": 4033
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008386400000000001,
+      "loss": 1.3163,
+      "step": 4034
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008386,
+      "loss": 1.3166,
+      "step": 4035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083856,
+      "loss": 1.2897,
+      "step": 4036
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083852,
+      "loss": 1.262,
+      "step": 4037
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008384800000000001,
+      "loss": 1.2937,
+      "step": 4038
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083844,
+      "loss": 1.3319,
+      "step": 4039
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008384000000000001,
+      "loss": 1.2942,
+      "step": 4040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083836,
+      "loss": 1.2629,
+      "step": 4041
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008383199999999999,
+      "loss": 1.2599,
+      "step": 4042
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008382800000000001,
+      "loss": 1.3198,
+      "step": 4043
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083824,
+      "loss": 1.3425,
+      "step": 4044
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008382,
+      "loss": 1.3271,
+      "step": 4045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083816,
+      "loss": 1.2215,
+      "step": 4046
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083812,
+      "loss": 1.2548,
+      "step": 4047
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083808,
+      "loss": 1.1592,
+      "step": 4048
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008380400000000001,
+      "loss": 1.2916,
+      "step": 4049
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000838,
+      "loss": 1.2516,
+      "step": 4050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083796,
+      "loss": 1.3347,
+      "step": 4051
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083792,
+      "loss": 1.2866,
+      "step": 4052
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008378799999999999,
+      "loss": 1.3486,
+      "step": 4053
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008378400000000001,
+      "loss": 1.3506,
+      "step": 4054
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008378,
+      "loss": 1.2046,
+      "step": 4055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083776,
+      "loss": 1.2152,
+      "step": 4056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083772,
+      "loss": 1.3562,
+      "step": 4057
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083768,
+      "loss": 1.3985,
+      "step": 4058
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008376400000000001,
+      "loss": 1.2193,
+      "step": 4059
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008376,
+      "loss": 1.3222,
+      "step": 4060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083756,
+      "loss": 1.2262,
+      "step": 4061
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083752,
+      "loss": 1.3172,
+      "step": 4062
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083748,
+      "loss": 1.1802,
+      "step": 4063
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083744,
+      "loss": 1.2662,
+      "step": 4064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008374000000000001,
+      "loss": 1.2191,
+      "step": 4065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083736,
+      "loss": 1.2464,
+      "step": 4066
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083732,
+      "loss": 1.1784,
+      "step": 4067
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083728,
+      "loss": 1.3371,
+      "step": 4068
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008372400000000001,
+      "loss": 1.2889,
+      "step": 4069
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008372000000000001,
+      "loss": 1.2503,
+      "step": 4070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083716,
+      "loss": 1.2811,
+      "step": 4071
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083712,
+      "loss": 1.2364,
+      "step": 4072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083708,
+      "loss": 1.2149,
+      "step": 4073
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083704,
+      "loss": 1.2668,
+      "step": 4074
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000837,
+      "loss": 1.3036,
+      "step": 4075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008369600000000001,
+      "loss": 1.2838,
+      "step": 4076
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083692,
+      "loss": 1.2784,
+      "step": 4077
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083688,
+      "loss": 1.3025,
+      "step": 4078
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083684,
+      "loss": 1.3243,
+      "step": 4079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008368,
+      "loss": 1.3124,
+      "step": 4080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083676,
+      "loss": 1.3495,
+      "step": 4081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083672,
+      "loss": 1.2728,
+      "step": 4082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083668,
+      "loss": 1.2452,
+      "step": 4083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083664,
+      "loss": 1.3212,
+      "step": 4084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008366000000000001,
+      "loss": 1.3082,
+      "step": 4085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083656,
+      "loss": 1.2921,
+      "step": 4086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008365200000000001,
+      "loss": 1.2231,
+      "step": 4087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083648,
+      "loss": 1.3064,
+      "step": 4088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008364399999999999,
+      "loss": 1.2272,
+      "step": 4089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008364000000000001,
+      "loss": 1.2966,
+      "step": 4090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083636,
+      "loss": 1.3422,
+      "step": 4091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083632,
+      "loss": 1.3266,
+      "step": 4092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083628,
+      "loss": 1.3083,
+      "step": 4093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083624,
+      "loss": 1.3915,
+      "step": 4094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008362,
+      "loss": 1.2426,
+      "step": 4095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008361600000000001,
+      "loss": 1.326,
+      "step": 4096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083612,
+      "loss": 1.3472,
+      "step": 4097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008360800000000001,
+      "loss": 1.2868,
+      "step": 4098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083604,
+      "loss": 1.2023,
+      "step": 4099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008359999999999999,
+      "loss": 1.2486,
+      "step": 4100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008359600000000001,
+      "loss": 1.3387,
+      "step": 4101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083592,
+      "loss": 1.2891,
+      "step": 4102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083588,
+      "loss": 1.2683,
+      "step": 4103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083584,
+      "loss": 1.248,
+      "step": 4104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008358,
+      "loss": 1.3523,
+      "step": 4105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083576,
+      "loss": 1.3208,
+      "step": 4106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008357200000000001,
+      "loss": 1.2392,
+      "step": 4107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083568,
+      "loss": 1.2684,
+      "step": 4108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083564,
+      "loss": 1.2243,
+      "step": 4109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008356,
+      "loss": 1.3342,
+      "step": 4110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083556,
+      "loss": 1.2391,
+      "step": 4111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008355200000000001,
+      "loss": 1.2689,
+      "step": 4112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083548,
+      "loss": 1.2357,
+      "step": 4113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083544,
+      "loss": 1.2422,
+      "step": 4114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008354,
+      "loss": 1.2576,
+      "step": 4115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083536,
+      "loss": 1.3635,
+      "step": 4116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083532,
+      "loss": 1.1979,
+      "step": 4117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008352800000000001,
+      "loss": 1.294,
+      "step": 4118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083524,
+      "loss": 1.3503,
+      "step": 4119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008352,
+      "loss": 1.228,
+      "step": 4120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083516,
+      "loss": 1.2935,
+      "step": 4121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083512,
+      "loss": 1.2384,
+      "step": 4122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008350800000000001,
+      "loss": 1.3708,
+      "step": 4123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083504,
+      "loss": 1.2167,
+      "step": 4124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000835,
+      "loss": 1.2097,
+      "step": 4125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083496,
+      "loss": 1.2812,
+      "step": 4126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008349200000000001,
+      "loss": 1.2536,
+      "step": 4127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083488,
+      "loss": 1.2749,
+      "step": 4128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083484,
+      "loss": 1.2397,
+      "step": 4129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008348,
+      "loss": 1.2613,
+      "step": 4130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008347599999999999,
+      "loss": 1.2296,
+      "step": 4131
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008347200000000001,
+      "loss": 1.2117,
+      "step": 4132
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083468,
+      "loss": 1.2726,
+      "step": 4133
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008346400000000001,
+      "loss": 1.2482,
+      "step": 4134
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008346,
+      "loss": 1.3021,
+      "step": 4135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083456,
+      "loss": 1.3064,
+      "step": 4136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083452,
+      "loss": 1.2663,
+      "step": 4137
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083448,
+      "loss": 1.3598,
+      "step": 4138
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083444,
+      "loss": 1.2389,
+      "step": 4139
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008344,
+      "loss": 1.2894,
+      "step": 4140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083436,
+      "loss": 1.3074,
+      "step": 4141
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008343199999999999,
+      "loss": 1.2607,
+      "step": 4142
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008342800000000001,
+      "loss": 1.2986,
+      "step": 4143
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083424,
+      "loss": 1.3499,
+      "step": 4144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008342000000000001,
+      "loss": 1.2104,
+      "step": 4145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083416,
+      "loss": 1.3438,
+      "step": 4146
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083412,
+      "loss": 1.2554,
+      "step": 4147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008340800000000001,
+      "loss": 1.2249,
+      "step": 4148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083404,
+      "loss": 1.2833,
+      "step": 4149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000834,
+      "loss": 1.239,
+      "step": 4150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083396,
+      "loss": 1.1996,
+      "step": 4151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083392,
+      "loss": 1.2983,
+      "step": 4152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083388,
+      "loss": 1.3339,
+      "step": 4153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008338400000000001,
+      "loss": 1.2934,
+      "step": 4154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008338,
+      "loss": 1.2233,
+      "step": 4155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083376,
+      "loss": 1.2728,
+      "step": 4156
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083372,
+      "loss": 1.3204,
+      "step": 4157
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083368,
+      "loss": 1.2712,
+      "step": 4158
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008336400000000001,
+      "loss": 1.3177,
+      "step": 4159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008336,
+      "loss": 1.2119,
+      "step": 4160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083356,
+      "loss": 1.3011,
+      "step": 4161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083352,
+      "loss": 1.201,
+      "step": 4162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083348,
+      "loss": 1.3568,
+      "step": 4163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083344,
+      "loss": 1.2188,
+      "step": 4164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008334000000000001,
+      "loss": 1.2321,
+      "step": 4165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083336,
+      "loss": 1.2568,
+      "step": 4166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008333199999999999,
+      "loss": 1.2974,
+      "step": 4167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083328,
+      "loss": 1.2409,
+      "step": 4168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083324,
+      "loss": 1.226,
+      "step": 4169
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008332000000000001,
+      "loss": 1.2127,
+      "step": 4170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083316,
+      "loss": 1.1999,
+      "step": 4171
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083312,
+      "loss": 1.2165,
+      "step": 4172
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083308,
+      "loss": 1.2999,
+      "step": 4173
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008330400000000001,
+      "loss": 1.3072,
+      "step": 4174
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000833,
+      "loss": 1.3204,
+      "step": 4175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008329600000000001,
+      "loss": 1.2946,
+      "step": 4176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083292,
+      "loss": 1.2556,
+      "step": 4177
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008328799999999999,
+      "loss": 1.3338,
+      "step": 4178
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008328400000000001,
+      "loss": 1.3107,
+      "step": 4179
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0008328,
+      "loss": 1.231,
+      "step": 4180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00083276,
+      "loss": 1.3408,
+      "step": 4181
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083272,
+      "loss": 1.2851,
+      "step": 4182
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083268,
+      "loss": 1.2843,
+      "step": 4183
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083264,
+      "loss": 1.3441,
+      "step": 4184
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008326000000000001,
+      "loss": 1.2731,
+      "step": 4185
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083256,
+      "loss": 1.2719,
+      "step": 4186
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083252,
+      "loss": 1.279,
+      "step": 4187
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083248,
+      "loss": 1.3004,
+      "step": 4188
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008324399999999999,
+      "loss": 1.193,
+      "step": 4189
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008324000000000001,
+      "loss": 1.2883,
+      "step": 4190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083236,
+      "loss": 1.3225,
+      "step": 4191
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083232,
+      "loss": 1.256,
+      "step": 4192
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083228,
+      "loss": 1.3633,
+      "step": 4193
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083224,
+      "loss": 1.3098,
+      "step": 4194
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008322000000000001,
+      "loss": 1.2338,
+      "step": 4195
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083216,
+      "loss": 1.226,
+      "step": 4196
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083212,
+      "loss": 1.2973,
+      "step": 4197
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083208,
+      "loss": 1.2363,
+      "step": 4198
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083204,
+      "loss": 1.2204,
+      "step": 4199
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000832,
+      "loss": 1.3633,
+      "step": 4200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008319600000000001,
+      "loss": 1.355,
+      "step": 4201
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083192,
+      "loss": 1.2121,
+      "step": 4202
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083188,
+      "loss": 1.2561,
+      "step": 4203
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083184,
+      "loss": 1.3369,
+      "step": 4204
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008318,
+      "loss": 1.2522,
+      "step": 4205
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008317600000000001,
+      "loss": 1.3395,
+      "step": 4206
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083172,
+      "loss": 1.2682,
+      "step": 4207
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083168,
+      "loss": 1.2843,
+      "step": 4208
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083164,
+      "loss": 1.2234,
+      "step": 4209
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008316,
+      "loss": 1.3015,
+      "step": 4210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083156,
+      "loss": 1.2845,
+      "step": 4211
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008315200000000001,
+      "loss": 1.3494,
+      "step": 4212
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083148,
+      "loss": 1.2975,
+      "step": 4213
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083144,
+      "loss": 1.2323,
+      "step": 4214
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008314,
+      "loss": 1.2882,
+      "step": 4215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083136,
+      "loss": 1.3045,
+      "step": 4216
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083132,
+      "loss": 1.17,
+      "step": 4217
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083128,
+      "loss": 1.2745,
+      "step": 4218
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083124,
+      "loss": 1.2765,
+      "step": 4219
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008312,
+      "loss": 1.2401,
+      "step": 4220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008311600000000001,
+      "loss": 1.379,
+      "step": 4221
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083112,
+      "loss": 1.2451,
+      "step": 4222
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008310800000000001,
+      "loss": 1.2834,
+      "step": 4223
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083104,
+      "loss": 1.3236,
+      "step": 4224
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008309999999999999,
+      "loss": 1.2536,
+      "step": 4225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083096,
+      "loss": 1.3297,
+      "step": 4226
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083092,
+      "loss": 1.2906,
+      "step": 4227
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083088,
+      "loss": 1.2591,
+      "step": 4228
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083084,
+      "loss": 1.2016,
+      "step": 4229
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008308,
+      "loss": 1.2317,
+      "step": 4230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083076,
+      "loss": 1.2159,
+      "step": 4231
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008307200000000001,
+      "loss": 1.2957,
+      "step": 4232
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083068,
+      "loss": 1.3032,
+      "step": 4233
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008306400000000001,
+      "loss": 1.2607,
+      "step": 4234
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008306,
+      "loss": 1.2669,
+      "step": 4235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008305599999999999,
+      "loss": 1.2021,
+      "step": 4236
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008305200000000001,
+      "loss": 1.2235,
+      "step": 4237
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083048,
+      "loss": 1.2596,
+      "step": 4238
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083044,
+      "loss": 1.3161,
+      "step": 4239
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008304,
+      "loss": 1.2957,
+      "step": 4240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083036,
+      "loss": 1.2718,
+      "step": 4241
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083032,
+      "loss": 1.1601,
+      "step": 4242
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008302800000000001,
+      "loss": 1.4124,
+      "step": 4243
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083024,
+      "loss": 1.3199,
+      "step": 4244
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008302,
+      "loss": 1.2797,
+      "step": 4245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083016,
+      "loss": 1.3363,
+      "step": 4246
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083012,
+      "loss": 1.2367,
+      "step": 4247
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008300800000000001,
+      "loss": 1.2386,
+      "step": 4248
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083004,
+      "loss": 1.3203,
+      "step": 4249
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00083,
+      "loss": 1.1305,
+      "step": 4250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082996,
+      "loss": 1.3308,
+      "step": 4251
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082992,
+      "loss": 1.1902,
+      "step": 4252
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082988,
+      "loss": 1.2805,
+      "step": 4253
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008298400000000001,
+      "loss": 1.2432,
+      "step": 4254
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008298,
+      "loss": 1.2843,
+      "step": 4255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082976,
+      "loss": 1.3225,
+      "step": 4256
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082972,
+      "loss": 1.3424,
+      "step": 4257
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082968,
+      "loss": 1.2517,
+      "step": 4258
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008296400000000001,
+      "loss": 1.3234,
+      "step": 4259
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008296,
+      "loss": 1.2184,
+      "step": 4260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082956,
+      "loss": 1.2652,
+      "step": 4261
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082952,
+      "loss": 1.2008,
+      "step": 4262
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008294800000000001,
+      "loss": 1.3181,
+      "step": 4263
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082944,
+      "loss": 1.3058,
+      "step": 4264
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008294,
+      "loss": 1.2448,
+      "step": 4265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082936,
+      "loss": 1.3346,
+      "step": 4266
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008293199999999999,
+      "loss": 1.3086,
+      "step": 4267
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008292800000000001,
+      "loss": 1.2481,
+      "step": 4268
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082924,
+      "loss": 1.3121,
+      "step": 4269
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008292000000000001,
+      "loss": 1.2889,
+      "step": 4270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082916,
+      "loss": 1.2119,
+      "step": 4271
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082912,
+      "loss": 1.283,
+      "step": 4272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082908,
+      "loss": 1.2637,
+      "step": 4273
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082904,
+      "loss": 1.2601,
+      "step": 4274
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000829,
+      "loss": 1.2257,
+      "step": 4275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082896,
+      "loss": 1.2865,
+      "step": 4276
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082892,
+      "loss": 1.3149,
+      "step": 4277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008288799999999999,
+      "loss": 1.2839,
+      "step": 4278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008288400000000001,
+      "loss": 1.2537,
+      "step": 4279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008288,
+      "loss": 1.2951,
+      "step": 4280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008287600000000001,
+      "loss": 1.2496,
+      "step": 4281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082872,
+      "loss": 1.2778,
+      "step": 4282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082868,
+      "loss": 1.3561,
+      "step": 4283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008286400000000001,
+      "loss": 1.2721,
+      "step": 4284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008286,
+      "loss": 1.2664,
+      "step": 4285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082856,
+      "loss": 1.2903,
+      "step": 4286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082852,
+      "loss": 1.2924,
+      "step": 4287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082848,
+      "loss": 1.2421,
+      "step": 4288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082844,
+      "loss": 1.2115,
+      "step": 4289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008284000000000001,
+      "loss": 1.3141,
+      "step": 4290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082836,
+      "loss": 1.2786,
+      "step": 4291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082832,
+      "loss": 1.27,
+      "step": 4292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082828,
+      "loss": 1.325,
+      "step": 4293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008282399999999999,
+      "loss": 1.1639,
+      "step": 4294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008282000000000001,
+      "loss": 1.3204,
+      "step": 4295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082816,
+      "loss": 1.234,
+      "step": 4296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082812,
+      "loss": 1.2398,
+      "step": 4297
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082808,
+      "loss": 1.3057,
+      "step": 4298
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082804,
+      "loss": 1.2708,
+      "step": 4299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000828,
+      "loss": 1.2193,
+      "step": 4300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008279600000000001,
+      "loss": 1.2715,
+      "step": 4301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082792,
+      "loss": 1.2418,
+      "step": 4302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008278799999999999,
+      "loss": 1.2806,
+      "step": 4303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082784,
+      "loss": 1.2596,
+      "step": 4304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008278,
+      "loss": 1.1999,
+      "step": 4305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008277600000000001,
+      "loss": 1.2506,
+      "step": 4306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082772,
+      "loss": 1.3114,
+      "step": 4307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082768,
+      "loss": 1.2547,
+      "step": 4308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082764,
+      "loss": 1.271,
+      "step": 4309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008276000000000001,
+      "loss": 1.3427,
+      "step": 4310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082756,
+      "loss": 1.4067,
+      "step": 4311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008275200000000001,
+      "loss": 1.2814,
+      "step": 4312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082748,
+      "loss": 1.2841,
+      "step": 4313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008274399999999999,
+      "loss": 1.3518,
+      "step": 4314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008274,
+      "loss": 1.2409,
+      "step": 4315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082736,
+      "loss": 1.2622,
+      "step": 4316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008273200000000001,
+      "loss": 1.267,
+      "step": 4317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082728,
+      "loss": 1.2418,
+      "step": 4318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082724,
+      "loss": 1.3187,
+      "step": 4319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008272,
+      "loss": 1.269,
+      "step": 4320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008271600000000001,
+      "loss": 1.2088,
+      "step": 4321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082712,
+      "loss": 1.3751,
+      "step": 4322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082708,
+      "loss": 1.3063,
+      "step": 4323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082704,
+      "loss": 1.3445,
+      "step": 4324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008269999999999999,
+      "loss": 1.2586,
+      "step": 4325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008269600000000001,
+      "loss": 1.2389,
+      "step": 4326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082692,
+      "loss": 1.3408,
+      "step": 4327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082688,
+      "loss": 1.3764,
+      "step": 4328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082684,
+      "loss": 1.1712,
+      "step": 4329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008268,
+      "loss": 1.2537,
+      "step": 4330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008267600000000001,
+      "loss": 1.2391,
+      "step": 4331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008267200000000001,
+      "loss": 1.4202,
+      "step": 4332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082668,
+      "loss": 1.3355,
+      "step": 4333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082664,
+      "loss": 1.2921,
+      "step": 4334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008266,
+      "loss": 1.182,
+      "step": 4335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082656,
+      "loss": 1.3005,
+      "step": 4336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008265200000000001,
+      "loss": 1.2622,
+      "step": 4337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082648,
+      "loss": 1.1517,
+      "step": 4338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082644,
+      "loss": 1.2433,
+      "step": 4339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008264,
+      "loss": 1.3355,
+      "step": 4340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082636,
+      "loss": 1.2434,
+      "step": 4341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008263200000000001,
+      "loss": 1.2659,
+      "step": 4342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082628,
+      "loss": 1.2669,
+      "step": 4343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082624,
+      "loss": 1.2387,
+      "step": 4344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008262,
+      "loss": 1.3007,
+      "step": 4345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082616,
+      "loss": 1.2827,
+      "step": 4346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082612,
+      "loss": 1.3159,
+      "step": 4347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008260800000000001,
+      "loss": 1.2421,
+      "step": 4348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082604,
+      "loss": 1.3323,
+      "step": 4349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000826,
+      "loss": 1.2818,
+      "step": 4350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082596,
+      "loss": 1.2903,
+      "step": 4351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082592,
+      "loss": 1.2742,
+      "step": 4352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082588,
+      "loss": 1.3316,
+      "step": 4353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082584,
+      "loss": 1.2212,
+      "step": 4354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008258,
+      "loss": 1.3155,
+      "step": 4355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082576,
+      "loss": 1.2623,
+      "step": 4356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008257200000000001,
+      "loss": 1.2211,
+      "step": 4357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082568,
+      "loss": 1.3884,
+      "step": 4358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008256400000000001,
+      "loss": 1.1716,
+      "step": 4359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008256,
+      "loss": 1.2722,
+      "step": 4360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082556,
+      "loss": 1.2571,
+      "step": 4361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082552,
+      "loss": 1.2465,
+      "step": 4362
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082548,
+      "loss": 1.2855,
+      "step": 4363
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082544,
+      "loss": 1.2887,
+      "step": 4364
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008254,
+      "loss": 1.3421,
+      "step": 4365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082536,
+      "loss": 1.3684,
+      "step": 4366
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082532,
+      "loss": 1.2686,
+      "step": 4367
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008252800000000001,
+      "loss": 1.2523,
+      "step": 4368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082524,
+      "loss": 1.2326,
+      "step": 4369
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008252000000000001,
+      "loss": 1.245,
+      "step": 4370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082516,
+      "loss": 1.1972,
+      "step": 4371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008251199999999999,
+      "loss": 1.2128,
+      "step": 4372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008250800000000001,
+      "loss": 1.345,
+      "step": 4373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082504,
+      "loss": 1.2649,
+      "step": 4374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000825,
+      "loss": 1.2611,
+      "step": 4375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082496,
+      "loss": 1.2799,
+      "step": 4376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082492,
+      "loss": 1.3386,
+      "step": 4377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082488,
+      "loss": 1.2854,
+      "step": 4378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008248400000000001,
+      "loss": 1.4012,
+      "step": 4379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008248,
+      "loss": 1.273,
+      "step": 4380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008247600000000001,
+      "loss": 1.2091,
+      "step": 4381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082472,
+      "loss": 1.2975,
+      "step": 4382
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008246799999999999,
+      "loss": 1.337,
+      "step": 4383
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008246400000000001,
+      "loss": 1.2724,
+      "step": 4384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008246,
+      "loss": 1.3352,
+      "step": 4385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082456,
+      "loss": 1.2299,
+      "step": 4386
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082452,
+      "loss": 1.2201,
+      "step": 4387
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082448,
+      "loss": 1.2186,
+      "step": 4388
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082444,
+      "loss": 1.2232,
+      "step": 4389
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008244000000000001,
+      "loss": 1.3016,
+      "step": 4390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082436,
+      "loss": 1.2458,
+      "step": 4391
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082432,
+      "loss": 1.3448,
+      "step": 4392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082428,
+      "loss": 1.3373,
+      "step": 4393
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082424,
+      "loss": 1.2765,
+      "step": 4394
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008242000000000001,
+      "loss": 1.3367,
+      "step": 4395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082416,
+      "loss": 1.2113,
+      "step": 4396
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082412,
+      "loss": 1.3111,
+      "step": 4397
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082408,
+      "loss": 1.2187,
+      "step": 4398
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008240400000000001,
+      "loss": 1.2319,
+      "step": 4399
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000824,
+      "loss": 1.3228,
+      "step": 4400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082396,
+      "loss": 1.3096,
+      "step": 4401
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082392,
+      "loss": 1.2313,
+      "step": 4402
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008238799999999999,
+      "loss": 1.2263,
+      "step": 4403
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082384,
+      "loss": 1.3558,
+      "step": 4404
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008238,
+      "loss": 1.2755,
+      "step": 4405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008237600000000001,
+      "loss": 1.2954,
+      "step": 4406
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082372,
+      "loss": 1.1878,
+      "step": 4407
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082368,
+      "loss": 1.2549,
+      "step": 4408
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082364,
+      "loss": 1.3318,
+      "step": 4409
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008236000000000001,
+      "loss": 1.3491,
+      "step": 4410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082356,
+      "loss": 1.2775,
+      "step": 4411
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082352,
+      "loss": 1.2534,
+      "step": 4412
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082348,
+      "loss": 1.3615,
+      "step": 4413
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008234399999999999,
+      "loss": 1.2478,
+      "step": 4414
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008234000000000001,
+      "loss": 1.2074,
+      "step": 4415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082336,
+      "loss": 1.2287,
+      "step": 4416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008233200000000001,
+      "loss": 1.3201,
+      "step": 4417
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082328,
+      "loss": 1.2671,
+      "step": 4418
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082324,
+      "loss": 1.2999,
+      "step": 4419
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008232000000000001,
+      "loss": 1.2458,
+      "step": 4420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082316,
+      "loss": 1.2379,
+      "step": 4421
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082312,
+      "loss": 1.299,
+      "step": 4422
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082308,
+      "loss": 1.2493,
+      "step": 4423
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082304,
+      "loss": 1.2878,
+      "step": 4424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000823,
+      "loss": 1.2292,
+      "step": 4425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008229600000000001,
+      "loss": 1.2831,
+      "step": 4426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082292,
+      "loss": 1.3348,
+      "step": 4427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082288,
+      "loss": 1.2363,
+      "step": 4428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082284,
+      "loss": 1.2334,
+      "step": 4429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008227999999999999,
+      "loss": 1.2741,
+      "step": 4430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0008227600000000001,
+      "loss": 1.2512,
+      "step": 4431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082272,
+      "loss": 1.2417,
+      "step": 4432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082268,
+      "loss": 1.3448,
+      "step": 4433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00082264,
+      "loss": 1.297,
+      "step": 4434
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008226,
+      "loss": 1.3218,
+      "step": 4435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082256,
+      "loss": 1.3017,
+      "step": 4436
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008225200000000001,
+      "loss": 1.249,
+      "step": 4437
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082248,
+      "loss": 1.2876,
+      "step": 4438
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082244,
+      "loss": 1.272,
+      "step": 4439
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008224,
+      "loss": 1.2829,
+      "step": 4440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082236,
+      "loss": 1.3008,
+      "step": 4441
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008223200000000001,
+      "loss": 1.2152,
+      "step": 4442
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082228,
+      "loss": 1.3167,
+      "step": 4443
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082224,
+      "loss": 1.3316,
+      "step": 4444
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008222,
+      "loss": 1.237,
+      "step": 4445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008221600000000001,
+      "loss": 1.2882,
+      "step": 4446
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082212,
+      "loss": 1.1197,
+      "step": 4447
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008220800000000001,
+      "loss": 1.3699,
+      "step": 4448
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082204,
+      "loss": 1.2502,
+      "step": 4449
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008219999999999999,
+      "loss": 1.2414,
+      "step": 4450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082196,
+      "loss": 1.2611,
+      "step": 4451
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082192,
+      "loss": 1.2164,
+      "step": 4452
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008218800000000001,
+      "loss": 1.2498,
+      "step": 4453
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082184,
+      "loss": 1.2377,
+      "step": 4454
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008218,
+      "loss": 1.3312,
+      "step": 4455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082176,
+      "loss": 1.2319,
+      "step": 4456
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008217200000000001,
+      "loss": 1.2507,
+      "step": 4457
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082168,
+      "loss": 1.2356,
+      "step": 4458
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082164,
+      "loss": 1.292,
+      "step": 4459
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008216,
+      "loss": 1.257,
+      "step": 4460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008215599999999999,
+      "loss": 1.207,
+      "step": 4461
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008215200000000001,
+      "loss": 1.2846,
+      "step": 4462
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082148,
+      "loss": 1.2134,
+      "step": 4463
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082144,
+      "loss": 1.3224,
+      "step": 4464
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008214,
+      "loss": 1.2812,
+      "step": 4465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082136,
+      "loss": 1.307,
+      "step": 4466
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008213200000000001,
+      "loss": 1.3372,
+      "step": 4467
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008212800000000001,
+      "loss": 1.2144,
+      "step": 4468
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082124,
+      "loss": 1.2619,
+      "step": 4469
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008212,
+      "loss": 1.2795,
+      "step": 4470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082116,
+      "loss": 1.253,
+      "step": 4471
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008211199999999999,
+      "loss": 1.2788,
+      "step": 4472
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008210800000000001,
+      "loss": 1.1976,
+      "step": 4473
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082104,
+      "loss": 1.3318,
+      "step": 4474
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000821,
+      "loss": 1.2816,
+      "step": 4475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082096,
+      "loss": 1.2694,
+      "step": 4476
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082092,
+      "loss": 1.2407,
+      "step": 4477
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008208800000000001,
+      "loss": 1.3368,
+      "step": 4478
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082084,
+      "loss": 1.3117,
+      "step": 4479
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008208,
+      "loss": 1.3221,
+      "step": 4480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082076,
+      "loss": 1.2529,
+      "step": 4481
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082072,
+      "loss": 1.2766,
+      "step": 4482
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082068,
+      "loss": 1.339,
+      "step": 4483
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008206400000000001,
+      "loss": 1.2312,
+      "step": 4484
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008206,
+      "loss": 1.1753,
+      "step": 4485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082056,
+      "loss": 1.2977,
+      "step": 4486
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082052,
+      "loss": 1.2562,
+      "step": 4487
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082048,
+      "loss": 1.3219,
+      "step": 4488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082044,
+      "loss": 1.2923,
+      "step": 4489
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008204,
+      "loss": 1.3186,
+      "step": 4490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082036,
+      "loss": 1.3001,
+      "step": 4491
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082032,
+      "loss": 1.3159,
+      "step": 4492
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008202800000000001,
+      "loss": 1.3404,
+      "step": 4493
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082024,
+      "loss": 1.288,
+      "step": 4494
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008202000000000001,
+      "loss": 1.1689,
+      "step": 4495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082016,
+      "loss": 1.1709,
+      "step": 4496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082012,
+      "loss": 1.3222,
+      "step": 4497
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082008,
+      "loss": 1.2064,
+      "step": 4498
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082004,
+      "loss": 1.3112,
+      "step": 4499
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00082,
+      "loss": 1.251,
+      "step": 4500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081996,
+      "loss": 1.2539,
+      "step": 4501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081992,
+      "loss": 1.2701,
+      "step": 4502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081988,
+      "loss": 1.2279,
+      "step": 4503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008198400000000001,
+      "loss": 1.3751,
+      "step": 4504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008198,
+      "loss": 1.1885,
+      "step": 4505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008197600000000001,
+      "loss": 1.388,
+      "step": 4506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081972,
+      "loss": 1.3783,
+      "step": 4507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008196799999999999,
+      "loss": 1.2742,
+      "step": 4508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008196400000000001,
+      "loss": 1.2675,
+      "step": 4509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008196,
+      "loss": 1.2235,
+      "step": 4510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081956,
+      "loss": 1.2046,
+      "step": 4511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081952,
+      "loss": 1.2524,
+      "step": 4512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081948,
+      "loss": 1.2871,
+      "step": 4513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081944,
+      "loss": 1.2816,
+      "step": 4514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008194000000000001,
+      "loss": 1.2417,
+      "step": 4515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081936,
+      "loss": 1.2553,
+      "step": 4516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008193200000000001,
+      "loss": 1.2073,
+      "step": 4517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081928,
+      "loss": 1.2935,
+      "step": 4518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008192399999999999,
+      "loss": 1.2542,
+      "step": 4519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008192000000000001,
+      "loss": 1.2838,
+      "step": 4520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081916,
+      "loss": 1.1605,
+      "step": 4521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081912,
+      "loss": 1.275,
+      "step": 4522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081908,
+      "loss": 1.3002,
+      "step": 4523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081904,
+      "loss": 1.2214,
+      "step": 4524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000819,
+      "loss": 1.2996,
+      "step": 4525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008189600000000001,
+      "loss": 1.2842,
+      "step": 4526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081892,
+      "loss": 1.2664,
+      "step": 4527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081888,
+      "loss": 1.3081,
+      "step": 4528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081884,
+      "loss": 1.2501,
+      "step": 4529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008188,
+      "loss": 1.2454,
+      "step": 4530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008187600000000001,
+      "loss": 1.3073,
+      "step": 4531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081872,
+      "loss": 1.2652,
+      "step": 4532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081868,
+      "loss": 1.2333,
+      "step": 4533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081864,
+      "loss": 1.2506,
+      "step": 4534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008186000000000001,
+      "loss": 1.2293,
+      "step": 4535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081856,
+      "loss": 1.2687,
+      "step": 4536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081852,
+      "loss": 1.2607,
+      "step": 4537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081848,
+      "loss": 1.2045,
+      "step": 4538
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008184399999999999,
+      "loss": 1.3233,
+      "step": 4539
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008184,
+      "loss": 1.3727,
+      "step": 4540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081836,
+      "loss": 1.2348,
+      "step": 4541
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008183200000000001,
+      "loss": 1.2278,
+      "step": 4542
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081828,
+      "loss": 1.2377,
+      "step": 4543
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081824,
+      "loss": 1.3047,
+      "step": 4544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008182,
+      "loss": 1.2328,
+      "step": 4545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008181600000000001,
+      "loss": 1.1772,
+      "step": 4546
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081812,
+      "loss": 1.2529,
+      "step": 4547
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081808,
+      "loss": 1.2488,
+      "step": 4548
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081804,
+      "loss": 1.3519,
+      "step": 4549
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008179999999999999,
+      "loss": 1.2252,
+      "step": 4550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008179600000000001,
+      "loss": 1.2935,
+      "step": 4551
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081792,
+      "loss": 1.1971,
+      "step": 4552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008178800000000001,
+      "loss": 1.2779,
+      "step": 4553
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081784,
+      "loss": 1.4053,
+      "step": 4554
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008178,
+      "loss": 1.2069,
+      "step": 4555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008177600000000001,
+      "loss": 1.1668,
+      "step": 4556
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081772,
+      "loss": 1.2148,
+      "step": 4557
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081768,
+      "loss": 1.2063,
+      "step": 4558
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081764,
+      "loss": 1.378,
+      "step": 4559
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008176,
+      "loss": 1.2286,
+      "step": 4560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008175599999999999,
+      "loss": 1.2535,
+      "step": 4561
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008175200000000001,
+      "loss": 1.3079,
+      "step": 4562
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081748,
+      "loss": 1.2499,
+      "step": 4563
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008174400000000001,
+      "loss": 1.2894,
+      "step": 4564
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008174,
+      "loss": 1.2348,
+      "step": 4565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008173599999999999,
+      "loss": 1.2589,
+      "step": 4566
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008173200000000001,
+      "loss": 1.2557,
+      "step": 4567
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081728,
+      "loss": 1.1768,
+      "step": 4568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081724,
+      "loss": 1.1691,
+      "step": 4569
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008172,
+      "loss": 1.2745,
+      "step": 4570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081716,
+      "loss": 1.331,
+      "step": 4571
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081712,
+      "loss": 1.2304,
+      "step": 4572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008170800000000001,
+      "loss": 1.3989,
+      "step": 4573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081704,
+      "loss": 1.1973,
+      "step": 4574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000817,
+      "loss": 1.2711,
+      "step": 4575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081696,
+      "loss": 1.2869,
+      "step": 4576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081692,
+      "loss": 1.2844,
+      "step": 4577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008168800000000001,
+      "loss": 1.2388,
+      "step": 4578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081684,
+      "loss": 1.3143,
+      "step": 4579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008168,
+      "loss": 1.2569,
+      "step": 4580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081676,
+      "loss": 1.1625,
+      "step": 4581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008167200000000001,
+      "loss": 1.213,
+      "step": 4582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081668,
+      "loss": 1.1869,
+      "step": 4583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008166400000000001,
+      "loss": 1.2822,
+      "step": 4584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008166,
+      "loss": 1.2184,
+      "step": 4585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008165599999999999,
+      "loss": 1.275,
+      "step": 4586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081652,
+      "loss": 1.1613,
+      "step": 4587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081648,
+      "loss": 1.3555,
+      "step": 4588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008164400000000001,
+      "loss": 1.3163,
+      "step": 4589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008164,
+      "loss": 1.298,
+      "step": 4590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081636,
+      "loss": 1.2465,
+      "step": 4591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081632,
+      "loss": 1.3093,
+      "step": 4592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008162800000000001,
+      "loss": 1.2753,
+      "step": 4593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081624,
+      "loss": 1.21,
+      "step": 4594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008162000000000001,
+      "loss": 1.2872,
+      "step": 4595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081616,
+      "loss": 1.2329,
+      "step": 4596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008161199999999999,
+      "loss": 1.2401,
+      "step": 4597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008160800000000001,
+      "loss": 1.2578,
+      "step": 4598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081604,
+      "loss": 1.3106,
+      "step": 4599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000816,
+      "loss": 1.1948,
+      "step": 4600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081596,
+      "loss": 1.3176,
+      "step": 4601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081592,
+      "loss": 1.3329,
+      "step": 4602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008158800000000001,
+      "loss": 1.2785,
+      "step": 4603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008158400000000001,
+      "loss": 1.2307,
+      "step": 4604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008158,
+      "loss": 1.342,
+      "step": 4605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081576,
+      "loss": 1.2844,
+      "step": 4606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081572,
+      "loss": 1.3049,
+      "step": 4607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008156799999999999,
+      "loss": 1.2968,
+      "step": 4608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008156400000000001,
+      "loss": 1.2829,
+      "step": 4609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008156,
+      "loss": 1.2188,
+      "step": 4610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081556,
+      "loss": 1.297,
+      "step": 4611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081552,
+      "loss": 1.2345,
+      "step": 4612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081548,
+      "loss": 1.2377,
+      "step": 4613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008154400000000001,
+      "loss": 1.1917,
+      "step": 4614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008154,
+      "loss": 1.2836,
+      "step": 4615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081536,
+      "loss": 1.2363,
+      "step": 4616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081532,
+      "loss": 1.3201,
+      "step": 4617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081528,
+      "loss": 1.2882,
+      "step": 4618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081524,
+      "loss": 1.3091,
+      "step": 4619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008152000000000001,
+      "loss": 1.3261,
+      "step": 4620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081516,
+      "loss": 1.2758,
+      "step": 4621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081512,
+      "loss": 1.2678,
+      "step": 4622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081508,
+      "loss": 1.2806,
+      "step": 4623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008150400000000001,
+      "loss": 1.2618,
+      "step": 4624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000815,
+      "loss": 1.2703,
+      "step": 4625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081496,
+      "loss": 1.2835,
+      "step": 4626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081492,
+      "loss": 1.3313,
+      "step": 4627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081488,
+      "loss": 1.2146,
+      "step": 4628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081484,
+      "loss": 1.2604,
+      "step": 4629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008148,
+      "loss": 1.255,
+      "step": 4630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008147600000000001,
+      "loss": 1.3033,
+      "step": 4631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081472,
+      "loss": 1.225,
+      "step": 4632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081468,
+      "loss": 1.2154,
+      "step": 4633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081464,
+      "loss": 1.3099,
+      "step": 4634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008146,
+      "loss": 1.2523,
+      "step": 4635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081456,
+      "loss": 1.2103,
+      "step": 4636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081452,
+      "loss": 1.3088,
+      "step": 4637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081448,
+      "loss": 1.3007,
+      "step": 4638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081444,
+      "loss": 1.2518,
+      "step": 4639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008144000000000001,
+      "loss": 1.3425,
+      "step": 4640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081436,
+      "loss": 1.3474,
+      "step": 4641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008143200000000001,
+      "loss": 1.2695,
+      "step": 4642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081428,
+      "loss": 1.198,
+      "step": 4643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008142399999999999,
+      "loss": 1.3246,
+      "step": 4644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008142000000000001,
+      "loss": 1.2434,
+      "step": 4645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081416,
+      "loss": 1.2459,
+      "step": 4646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081412,
+      "loss": 1.1823,
+      "step": 4647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081408,
+      "loss": 1.1869,
+      "step": 4648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081404,
+      "loss": 1.3078,
+      "step": 4649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008139999999999999,
+      "loss": 1.224,
+      "step": 4650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008139600000000001,
+      "loss": 1.2272,
+      "step": 4651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081392,
+      "loss": 1.2294,
+      "step": 4652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008138800000000001,
+      "loss": 1.1577,
+      "step": 4653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081384,
+      "loss": 1.24,
+      "step": 4654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008137999999999999,
+      "loss": 1.3023,
+      "step": 4655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008137600000000001,
+      "loss": 1.2037,
+      "step": 4656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081372,
+      "loss": 1.2444,
+      "step": 4657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081368,
+      "loss": 1.1671,
+      "step": 4658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081364,
+      "loss": 1.1992,
+      "step": 4659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008136,
+      "loss": 1.3215,
+      "step": 4660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081356,
+      "loss": 1.2664,
+      "step": 4661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008135200000000001,
+      "loss": 1.2578,
+      "step": 4662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081348,
+      "loss": 1.2375,
+      "step": 4663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081344,
+      "loss": 1.2456,
+      "step": 4664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008134,
+      "loss": 1.2769,
+      "step": 4665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081336,
+      "loss": 1.2435,
+      "step": 4666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008133200000000001,
+      "loss": 1.3494,
+      "step": 4667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081328,
+      "loss": 1.2031,
+      "step": 4668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081324,
+      "loss": 1.32,
+      "step": 4669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008132,
+      "loss": 1.2914,
+      "step": 4670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008131600000000001,
+      "loss": 1.2599,
+      "step": 4671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081312,
+      "loss": 1.2609,
+      "step": 4672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008130800000000001,
+      "loss": 1.2276,
+      "step": 4673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081304,
+      "loss": 1.2382,
+      "step": 4674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008129999999999999,
+      "loss": 1.2786,
+      "step": 4675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081296,
+      "loss": 1.2729,
+      "step": 4676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081292,
+      "loss": 1.2121,
+      "step": 4677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008128800000000001,
+      "loss": 1.1675,
+      "step": 4678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081284,
+      "loss": 1.2211,
+      "step": 4679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008128,
+      "loss": 1.3166,
+      "step": 4680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081276,
+      "loss": 1.3952,
+      "step": 4681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008127200000000001,
+      "loss": 1.2839,
+      "step": 4682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081268,
+      "loss": 1.2259,
+      "step": 4683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00081264,
+      "loss": 1.207,
+      "step": 4684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008126,
+      "loss": 1.2769,
+      "step": 4685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008125599999999999,
+      "loss": 1.2195,
+      "step": 4686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0008125200000000001,
+      "loss": 1.2177,
+      "step": 4687
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081248,
+      "loss": 1.3244,
+      "step": 4688
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008124400000000001,
+      "loss": 1.2794,
+      "step": 4689
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008124,
+      "loss": 1.3324,
+      "step": 4690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081236,
+      "loss": 1.2929,
+      "step": 4691
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008123200000000001,
+      "loss": 1.2817,
+      "step": 4692
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081228,
+      "loss": 1.2496,
+      "step": 4693
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081224,
+      "loss": 1.3107,
+      "step": 4694
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008122,
+      "loss": 1.2686,
+      "step": 4695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081216,
+      "loss": 1.1703,
+      "step": 4696
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008121199999999999,
+      "loss": 1.3444,
+      "step": 4697
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008120800000000001,
+      "loss": 1.2677,
+      "step": 4698
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081204,
+      "loss": 1.3778,
+      "step": 4699
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008120000000000001,
+      "loss": 1.3067,
+      "step": 4700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081196,
+      "loss": 1.1884,
+      "step": 4701
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008119199999999999,
+      "loss": 1.2821,
+      "step": 4702
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008118800000000001,
+      "loss": 1.2478,
+      "step": 4703
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081184,
+      "loss": 1.274,
+      "step": 4704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008118,
+      "loss": 1.2625,
+      "step": 4705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081176,
+      "loss": 1.2218,
+      "step": 4706
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081172,
+      "loss": 1.2767,
+      "step": 4707
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081168,
+      "loss": 1.2492,
+      "step": 4708
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008116400000000001,
+      "loss": 1.3486,
+      "step": 4709
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008116,
+      "loss": 1.2698,
+      "step": 4710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081156,
+      "loss": 1.2304,
+      "step": 4711
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081152,
+      "loss": 1.274,
+      "step": 4712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081148,
+      "loss": 1.2363,
+      "step": 4713
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008114400000000001,
+      "loss": 1.2074,
+      "step": 4714
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008114,
+      "loss": 1.2327,
+      "step": 4715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081136,
+      "loss": 1.2719,
+      "step": 4716
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081132,
+      "loss": 1.3274,
+      "step": 4717
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081128,
+      "loss": 1.2645,
+      "step": 4718
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081124,
+      "loss": 1.3009,
+      "step": 4719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008112000000000001,
+      "loss": 1.2814,
+      "step": 4720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081116,
+      "loss": 1.1159,
+      "step": 4721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008111199999999999,
+      "loss": 1.2132,
+      "step": 4722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081108,
+      "loss": 1.266,
+      "step": 4723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081104,
+      "loss": 1.2782,
+      "step": 4724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008110000000000001,
+      "loss": 1.2423,
+      "step": 4725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081096,
+      "loss": 1.3146,
+      "step": 4726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081092,
+      "loss": 1.2515,
+      "step": 4727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081088,
+      "loss": 1.2415,
+      "step": 4728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008108400000000001,
+      "loss": 1.274,
+      "step": 4729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008108,
+      "loss": 1.2713,
+      "step": 4730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008107600000000001,
+      "loss": 1.3919,
+      "step": 4731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081072,
+      "loss": 1.1503,
+      "step": 4732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008106799999999999,
+      "loss": 1.2896,
+      "step": 4733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008106400000000001,
+      "loss": 1.3031,
+      "step": 4734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008106,
+      "loss": 1.3332,
+      "step": 4735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081056,
+      "loss": 1.2532,
+      "step": 4736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081052,
+      "loss": 1.2483,
+      "step": 4737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081048,
+      "loss": 1.288,
+      "step": 4738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081044,
+      "loss": 1.2912,
+      "step": 4739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008104000000000001,
+      "loss": 1.3463,
+      "step": 4740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081036,
+      "loss": 1.3393,
+      "step": 4741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081032,
+      "loss": 1.1839,
+      "step": 4742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081028,
+      "loss": 1.3517,
+      "step": 4743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008102399999999999,
+      "loss": 1.2527,
+      "step": 4744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008102000000000001,
+      "loss": 1.2729,
+      "step": 4745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081016,
+      "loss": 1.3701,
+      "step": 4746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081012,
+      "loss": 1.2911,
+      "step": 4747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081008,
+      "loss": 1.259,
+      "step": 4748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00081004,
+      "loss": 1.2291,
+      "step": 4749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008100000000000001,
+      "loss": 1.238,
+      "step": 4750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080996,
+      "loss": 1.31,
+      "step": 4751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080992,
+      "loss": 1.2514,
+      "step": 4752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080988,
+      "loss": 1.2065,
+      "step": 4753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080984,
+      "loss": 1.1762,
+      "step": 4754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008098,
+      "loss": 1.23,
+      "step": 4755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008097600000000001,
+      "loss": 1.4217,
+      "step": 4756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080972,
+      "loss": 1.2969,
+      "step": 4757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080968,
+      "loss": 1.352,
+      "step": 4758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080964,
+      "loss": 1.2769,
+      "step": 4759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008096000000000001,
+      "loss": 1.2319,
+      "step": 4760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080956,
+      "loss": 1.3283,
+      "step": 4761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080952,
+      "loss": 1.2693,
+      "step": 4762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080948,
+      "loss": 1.2478,
+      "step": 4763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080944,
+      "loss": 1.2701,
+      "step": 4764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008094,
+      "loss": 1.2869,
+      "step": 4765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080936,
+      "loss": 1.2503,
+      "step": 4766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008093200000000001,
+      "loss": 1.1826,
+      "step": 4767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080928,
+      "loss": 1.354,
+      "step": 4768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080924,
+      "loss": 1.3066,
+      "step": 4769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008092,
+      "loss": 1.2723,
+      "step": 4770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080916,
+      "loss": 1.2696,
+      "step": 4771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080912,
+      "loss": 1.329,
+      "step": 4772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080908,
+      "loss": 1.2999,
+      "step": 4773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080904,
+      "loss": 1.2894,
+      "step": 4774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000809,
+      "loss": 1.268,
+      "step": 4775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008089600000000001,
+      "loss": 1.225,
+      "step": 4776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080892,
+      "loss": 1.2181,
+      "step": 4777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008088800000000001,
+      "loss": 1.2143,
+      "step": 4778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080884,
+      "loss": 1.2727,
+      "step": 4779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008087999999999999,
+      "loss": 1.3184,
+      "step": 4780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008087600000000001,
+      "loss": 1.2229,
+      "step": 4781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080872,
+      "loss": 1.2445,
+      "step": 4782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080868,
+      "loss": 1.163,
+      "step": 4783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080864,
+      "loss": 1.2687,
+      "step": 4784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008086,
+      "loss": 1.2081,
+      "step": 4785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008085599999999999,
+      "loss": 1.1435,
+      "step": 4786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008085200000000001,
+      "loss": 1.2615,
+      "step": 4787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080848,
+      "loss": 1.2706,
+      "step": 4788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008084400000000001,
+      "loss": 1.2634,
+      "step": 4789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008084,
+      "loss": 1.2843,
+      "step": 4790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008083599999999999,
+      "loss": 1.2108,
+      "step": 4791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008083200000000001,
+      "loss": 1.3721,
+      "step": 4792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080828,
+      "loss": 1.3175,
+      "step": 4793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080824,
+      "loss": 1.3463,
+      "step": 4794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008082,
+      "loss": 1.2625,
+      "step": 4795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080816,
+      "loss": 1.2292,
+      "step": 4796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080812,
+      "loss": 1.2545,
+      "step": 4797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008080800000000001,
+      "loss": 1.2051,
+      "step": 4798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080804,
+      "loss": 1.3125,
+      "step": 4799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000808,
+      "loss": 1.1941,
+      "step": 4800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080796,
+      "loss": 1.1473,
+      "step": 4801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080792,
+      "loss": 1.1224,
+      "step": 4802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008078800000000001,
+      "loss": 1.1654,
+      "step": 4803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080784,
+      "loss": 1.2321,
+      "step": 4804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008078,
+      "loss": 1.2663,
+      "step": 4805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080776,
+      "loss": 1.2541,
+      "step": 4806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080772,
+      "loss": 1.2417,
+      "step": 4807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080768,
+      "loss": 1.1693,
+      "step": 4808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008076400000000001,
+      "loss": 1.2725,
+      "step": 4809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008076,
+      "loss": 1.1879,
+      "step": 4810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008075599999999999,
+      "loss": 1.2846,
+      "step": 4811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080752,
+      "loss": 1.2561,
+      "step": 4812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080748,
+      "loss": 1.2338,
+      "step": 4813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008074400000000001,
+      "loss": 1.231,
+      "step": 4814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008074,
+      "loss": 1.2918,
+      "step": 4815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080736,
+      "loss": 1.2771,
+      "step": 4816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080732,
+      "loss": 1.1892,
+      "step": 4817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008072800000000001,
+      "loss": 1.2882,
+      "step": 4818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080724,
+      "loss": 1.1767,
+      "step": 4819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008072,
+      "loss": 1.1775,
+      "step": 4820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080716,
+      "loss": 1.2025,
+      "step": 4821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008071199999999999,
+      "loss": 1.0598,
+      "step": 4822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008070800000000001,
+      "loss": 1.2817,
+      "step": 4823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080704,
+      "loss": 1.2823,
+      "step": 4824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008070000000000001,
+      "loss": 1.3311,
+      "step": 4825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080696,
+      "loss": 1.3725,
+      "step": 4826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080692,
+      "loss": 1.2351,
+      "step": 4827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080688,
+      "loss": 1.2966,
+      "step": 4828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080684,
+      "loss": 1.324,
+      "step": 4829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008068,
+      "loss": 1.2571,
+      "step": 4830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080676,
+      "loss": 1.2954,
+      "step": 4831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080672,
+      "loss": 1.2329,
+      "step": 4832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008066799999999999,
+      "loss": 1.2459,
+      "step": 4833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008066400000000001,
+      "loss": 1.2732,
+      "step": 4834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008066,
+      "loss": 1.2325,
+      "step": 4835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008065600000000001,
+      "loss": 1.2212,
+      "step": 4836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080652,
+      "loss": 1.3141,
+      "step": 4837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080648,
+      "loss": 1.3532,
+      "step": 4838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008064400000000001,
+      "loss": 1.2799,
+      "step": 4839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008064,
+      "loss": 1.3736,
+      "step": 4840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080636,
+      "loss": 1.2225,
+      "step": 4841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080632,
+      "loss": 1.262,
+      "step": 4842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080628,
+      "loss": 1.2646,
+      "step": 4843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080624,
+      "loss": 1.2398,
+      "step": 4844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008062000000000001,
+      "loss": 1.2769,
+      "step": 4845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080616,
+      "loss": 1.2787,
+      "step": 4846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080612,
+      "loss": 1.2638,
+      "step": 4847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080608,
+      "loss": 1.2556,
+      "step": 4848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080604,
+      "loss": 1.2031,
+      "step": 4849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008060000000000001,
+      "loss": 1.258,
+      "step": 4850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080596,
+      "loss": 1.308,
+      "step": 4851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080592,
+      "loss": 1.1979,
+      "step": 4852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080588,
+      "loss": 1.203,
+      "step": 4853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080584,
+      "loss": 1.1539,
+      "step": 4854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008058,
+      "loss": 1.2204,
+      "step": 4855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008057600000000001,
+      "loss": 1.2452,
+      "step": 4856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080572,
+      "loss": 1.247,
+      "step": 4857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008056799999999999,
+      "loss": 1.2539,
+      "step": 4858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080564,
+      "loss": 1.2732,
+      "step": 4859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008056,
+      "loss": 1.2615,
+      "step": 4860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008055600000000001,
+      "loss": 1.3379,
+      "step": 4861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080552,
+      "loss": 1.3255,
+      "step": 4862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080548,
+      "loss": 1.2054,
+      "step": 4863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080544,
+      "loss": 1.2665,
+      "step": 4864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008054000000000001,
+      "loss": 1.2819,
+      "step": 4865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080536,
+      "loss": 1.3482,
+      "step": 4866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008053200000000001,
+      "loss": 1.264,
+      "step": 4867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080528,
+      "loss": 1.2238,
+      "step": 4868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008052399999999999,
+      "loss": 1.2949,
+      "step": 4869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008052000000000001,
+      "loss": 1.2276,
+      "step": 4870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080516,
+      "loss": 1.2461,
+      "step": 4871
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080512,
+      "loss": 1.1896,
+      "step": 4872
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080508,
+      "loss": 1.2327,
+      "step": 4873
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080504,
+      "loss": 1.2732,
+      "step": 4874
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000805,
+      "loss": 1.3243,
+      "step": 4875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008049600000000001,
+      "loss": 1.2843,
+      "step": 4876
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080492,
+      "loss": 1.2084,
+      "step": 4877
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080488,
+      "loss": 1.2133,
+      "step": 4878
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080484,
+      "loss": 1.2808,
+      "step": 4879
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008047999999999999,
+      "loss": 1.2808,
+      "step": 4880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008047600000000001,
+      "loss": 1.2616,
+      "step": 4881
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080472,
+      "loss": 1.2946,
+      "step": 4882
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080468,
+      "loss": 1.2561,
+      "step": 4883
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080464,
+      "loss": 1.1942,
+      "step": 4884
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008046,
+      "loss": 1.2402,
+      "step": 4885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008045600000000001,
+      "loss": 1.2026,
+      "step": 4886
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008045200000000001,
+      "loss": 1.2853,
+      "step": 4887
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080448,
+      "loss": 1.2505,
+      "step": 4888
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080444,
+      "loss": 1.2062,
+      "step": 4889
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008044,
+      "loss": 1.1773,
+      "step": 4890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080436,
+      "loss": 1.1695,
+      "step": 4891
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008043200000000001,
+      "loss": 1.2024,
+      "step": 4892
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080428,
+      "loss": 1.1147,
+      "step": 4893
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080424,
+      "loss": 1.334,
+      "step": 4894
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008042,
+      "loss": 1.2794,
+      "step": 4895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080416,
+      "loss": 1.2605,
+      "step": 4896
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080412,
+      "loss": 1.2765,
+      "step": 4897
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080408,
+      "loss": 1.1808,
+      "step": 4898
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080404,
+      "loss": 1.221,
+      "step": 4899
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000804,
+      "loss": 1.3009,
+      "step": 4900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080396,
+      "loss": 1.2552,
+      "step": 4901
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080392,
+      "loss": 1.1613,
+      "step": 4902
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008038800000000001,
+      "loss": 1.2321,
+      "step": 4903
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080384,
+      "loss": 1.2301,
+      "step": 4904
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008038,
+      "loss": 1.2265,
+      "step": 4905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080376,
+      "loss": 1.2742,
+      "step": 4906
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080372,
+      "loss": 1.2372,
+      "step": 4907
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080368,
+      "loss": 1.2179,
+      "step": 4908
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080364,
+      "loss": 1.1899,
+      "step": 4909
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008036,
+      "loss": 1.267,
+      "step": 4910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080356,
+      "loss": 1.2513,
+      "step": 4911
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008035200000000001,
+      "loss": 1.2471,
+      "step": 4912
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080348,
+      "loss": 1.2501,
+      "step": 4913
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008034400000000001,
+      "loss": 1.3068,
+      "step": 4914
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008034,
+      "loss": 1.3206,
+      "step": 4915
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080336,
+      "loss": 1.1695,
+      "step": 4916
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080332,
+      "loss": 1.1822,
+      "step": 4917
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080328,
+      "loss": 1.179,
+      "step": 4918
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080324,
+      "loss": 1.2569,
+      "step": 4919
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008032,
+      "loss": 1.3188,
+      "step": 4920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080316,
+      "loss": 1.2796,
+      "step": 4921
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008031199999999999,
+      "loss": 1.2778,
+      "step": 4922
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008030800000000001,
+      "loss": 1.1907,
+      "step": 4923
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080304,
+      "loss": 1.2463,
+      "step": 4924
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008030000000000001,
+      "loss": 1.2034,
+      "step": 4925
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080296,
+      "loss": 1.181,
+      "step": 4926
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008029199999999999,
+      "loss": 1.2212,
+      "step": 4927
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008028800000000001,
+      "loss": 1.1631,
+      "step": 4928
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080284,
+      "loss": 1.2576,
+      "step": 4929
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008028,
+      "loss": 1.2301,
+      "step": 4930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080276,
+      "loss": 1.2548,
+      "step": 4931
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080272,
+      "loss": 1.3372,
+      "step": 4932
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080268,
+      "loss": 1.1466,
+      "step": 4933
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008026400000000001,
+      "loss": 1.2062,
+      "step": 4934
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008026,
+      "loss": 1.1853,
+      "step": 4935
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080256,
+      "loss": 1.2396,
+      "step": 4936
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080252,
+      "loss": 1.375,
+      "step": 4937
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080248,
+      "loss": 1.3365,
+      "step": 4938
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008024400000000001,
+      "loss": 1.2035,
+      "step": 4939
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0008024,
+      "loss": 1.2041,
+      "step": 4940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00080236,
+      "loss": 1.23,
+      "step": 4941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080232,
+      "loss": 1.1968,
+      "step": 4942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080228,
+      "loss": 1.1993,
+      "step": 4943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080224,
+      "loss": 1.3118,
+      "step": 4944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008022000000000001,
+      "loss": 1.3305,
+      "step": 4945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080216,
+      "loss": 1.2323,
+      "step": 4946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080212,
+      "loss": 1.204,
+      "step": 4947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080208,
+      "loss": 1.2549,
+      "step": 4948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080204,
+      "loss": 1.1573,
+      "step": 4949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008020000000000001,
+      "loss": 1.1906,
+      "step": 4950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080196,
+      "loss": 1.303,
+      "step": 4951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080192,
+      "loss": 1.1835,
+      "step": 4952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080188,
+      "loss": 1.2607,
+      "step": 4953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008018400000000001,
+      "loss": 1.1914,
+      "step": 4954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008018,
+      "loss": 1.3134,
+      "step": 4955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080176,
+      "loss": 1.2656,
+      "step": 4956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080172,
+      "loss": 1.3231,
+      "step": 4957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008016799999999999,
+      "loss": 1.2374,
+      "step": 4958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008016400000000001,
+      "loss": 1.307,
+      "step": 4959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008016,
+      "loss": 1.1748,
+      "step": 4960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008015600000000001,
+      "loss": 1.3116,
+      "step": 4961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080152,
+      "loss": 1.2224,
+      "step": 4962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080148,
+      "loss": 1.3143,
+      "step": 4963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080144,
+      "loss": 1.2397,
+      "step": 4964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008014,
+      "loss": 1.3031,
+      "step": 4965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080136,
+      "loss": 1.1563,
+      "step": 4966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080132,
+      "loss": 1.1747,
+      "step": 4967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080128,
+      "loss": 1.2286,
+      "step": 4968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008012399999999999,
+      "loss": 1.1564,
+      "step": 4969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008012000000000001,
+      "loss": 1.2529,
+      "step": 4970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080116,
+      "loss": 1.1845,
+      "step": 4971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008011200000000001,
+      "loss": 1.3174,
+      "step": 4972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080108,
+      "loss": 1.2921,
+      "step": 4973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080104,
+      "loss": 1.2102,
+      "step": 4974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008010000000000001,
+      "loss": 1.2104,
+      "step": 4975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080096,
+      "loss": 1.2524,
+      "step": 4976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080092,
+      "loss": 1.3048,
+      "step": 4977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080088,
+      "loss": 1.2702,
+      "step": 4978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080084,
+      "loss": 1.3114,
+      "step": 4979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008008,
+      "loss": 1.3331,
+      "step": 4980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008007600000000001,
+      "loss": 1.2098,
+      "step": 4981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080072,
+      "loss": 1.1731,
+      "step": 4982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080068,
+      "loss": 1.1966,
+      "step": 4983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080064,
+      "loss": 1.2141,
+      "step": 4984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008005999999999999,
+      "loss": 1.2274,
+      "step": 4985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008005600000000001,
+      "loss": 1.2179,
+      "step": 4986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080052,
+      "loss": 1.2988,
+      "step": 4987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080048,
+      "loss": 1.2727,
+      "step": 4988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080044,
+      "loss": 1.2535,
+      "step": 4989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008004,
+      "loss": 1.2883,
+      "step": 4990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080036,
+      "loss": 1.2273,
+      "step": 4991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008003200000000001,
+      "loss": 1.1844,
+      "step": 4992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080028,
+      "loss": 1.2058,
+      "step": 4993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008002399999999999,
+      "loss": 1.2638,
+      "step": 4994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008002,
+      "loss": 1.23,
+      "step": 4995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080016,
+      "loss": 1.2565,
+      "step": 4996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008001200000000001,
+      "loss": 1.2386,
+      "step": 4997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080008,
+      "loss": 1.2604,
+      "step": 4998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00080004,
+      "loss": 1.2698,
+      "step": 4999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0008,
+      "loss": 1.1832,
+      "step": 5000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007999600000000001,
+      "loss": 1.1908,
+      "step": 5001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079992,
+      "loss": 1.1998,
+      "step": 5002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007998800000000001,
+      "loss": 1.3199,
+      "step": 5003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079984,
+      "loss": 1.1534,
+      "step": 5004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007997999999999999,
+      "loss": 1.2475,
+      "step": 5005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079976,
+      "loss": 1.2274,
+      "step": 5006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079972,
+      "loss": 1.2683,
+      "step": 5007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079968,
+      "loss": 1.2596,
+      "step": 5008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079964,
+      "loss": 1.2912,
+      "step": 5009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007996,
+      "loss": 1.1934,
+      "step": 5010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079956,
+      "loss": 1.1184,
+      "step": 5011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007995200000000001,
+      "loss": 1.1981,
+      "step": 5012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079948,
+      "loss": 1.2573,
+      "step": 5013
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079944,
+      "loss": 1.2053,
+      "step": 5014
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007994,
+      "loss": 1.2135,
+      "step": 5015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007993599999999999,
+      "loss": 1.2102,
+      "step": 5016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007993200000000001,
+      "loss": 1.2895,
+      "step": 5017
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079928,
+      "loss": 1.2803,
+      "step": 5018
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079924,
+      "loss": 1.3063,
+      "step": 5019
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007992,
+      "loss": 1.2321,
+      "step": 5020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079916,
+      "loss": 1.1724,
+      "step": 5021
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007991200000000001,
+      "loss": 1.1918,
+      "step": 5022
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007990800000000001,
+      "loss": 1.2219,
+      "step": 5023
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079904,
+      "loss": 1.2674,
+      "step": 5024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000799,
+      "loss": 1.298,
+      "step": 5025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079896,
+      "loss": 1.1875,
+      "step": 5026
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079892,
+      "loss": 1.3181,
+      "step": 5027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007988800000000001,
+      "loss": 1.3023,
+      "step": 5028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079884,
+      "loss": 1.251,
+      "step": 5029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007988,
+      "loss": 1.2862,
+      "step": 5030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079876,
+      "loss": 1.2014,
+      "step": 5031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079872,
+      "loss": 1.2921,
+      "step": 5032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079868,
+      "loss": 1.2902,
+      "step": 5033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079864,
+      "loss": 1.249,
+      "step": 5034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007986,
+      "loss": 1.2042,
+      "step": 5035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079856,
+      "loss": 1.2454,
+      "step": 5036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079852,
+      "loss": 1.2784,
+      "step": 5037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079848,
+      "loss": 1.2617,
+      "step": 5038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007984400000000001,
+      "loss": 1.1721,
+      "step": 5039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007984,
+      "loss": 1.357,
+      "step": 5040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079836,
+      "loss": 1.2121,
+      "step": 5041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079832,
+      "loss": 1.2242,
+      "step": 5042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079828,
+      "loss": 1.2733,
+      "step": 5043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079824,
+      "loss": 1.2186,
+      "step": 5044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007982,
+      "loss": 1.3112,
+      "step": 5045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079816,
+      "loss": 1.2576,
+      "step": 5046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079812,
+      "loss": 1.2205,
+      "step": 5047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007980800000000001,
+      "loss": 1.1548,
+      "step": 5048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079804,
+      "loss": 1.2549,
+      "step": 5049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007980000000000001,
+      "loss": 1.1662,
+      "step": 5050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079796,
+      "loss": 1.1465,
+      "step": 5051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079792,
+      "loss": 1.1745,
+      "step": 5052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079788,
+      "loss": 1.2947,
+      "step": 5053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079784,
+      "loss": 1.2774,
+      "step": 5054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007978,
+      "loss": 1.241,
+      "step": 5055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079776,
+      "loss": 1.2248,
+      "step": 5056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079772,
+      "loss": 1.2752,
+      "step": 5057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007976799999999999,
+      "loss": 1.3073,
+      "step": 5058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007976400000000001,
+      "loss": 1.2376,
+      "step": 5059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007976,
+      "loss": 1.2451,
+      "step": 5060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007975600000000001,
+      "loss": 1.2064,
+      "step": 5061
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079752,
+      "loss": 1.1987,
+      "step": 5062
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007974799999999999,
+      "loss": 1.2982,
+      "step": 5063
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007974400000000001,
+      "loss": 1.2699,
+      "step": 5064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007974,
+      "loss": 1.1341,
+      "step": 5065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079736,
+      "loss": 1.2505,
+      "step": 5066
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079732,
+      "loss": 1.2357,
+      "step": 5067
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079728,
+      "loss": 1.1912,
+      "step": 5068
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079724,
+      "loss": 1.3581,
+      "step": 5069
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007972000000000001,
+      "loss": 1.2727,
+      "step": 5070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079716,
+      "loss": 1.2675,
+      "step": 5071
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007971200000000001,
+      "loss": 1.2538,
+      "step": 5072
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079708,
+      "loss": 1.3005,
+      "step": 5073
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007970399999999999,
+      "loss": 1.2681,
+      "step": 5074
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007970000000000001,
+      "loss": 1.2393,
+      "step": 5075
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079696,
+      "loss": 1.198,
+      "step": 5076
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079692,
+      "loss": 1.2594,
+      "step": 5077
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079688,
+      "loss": 1.3012,
+      "step": 5078
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079684,
+      "loss": 1.294,
+      "step": 5079
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007968,
+      "loss": 1.1888,
+      "step": 5080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007967600000000001,
+      "loss": 1.2652,
+      "step": 5081
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079672,
+      "loss": 1.2077,
+      "step": 5082
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079668,
+      "loss": 1.2642,
+      "step": 5083
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079664,
+      "loss": 1.2152,
+      "step": 5084
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007966,
+      "loss": 1.2733,
+      "step": 5085
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007965600000000001,
+      "loss": 1.2888,
+      "step": 5086
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079652,
+      "loss": 1.1997,
+      "step": 5087
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079648,
+      "loss": 1.1767,
+      "step": 5088
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079644,
+      "loss": 1.2083,
+      "step": 5089
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007964000000000001,
+      "loss": 1.2651,
+      "step": 5090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079636,
+      "loss": 1.1731,
+      "step": 5091
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079632,
+      "loss": 1.2468,
+      "step": 5092
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079628,
+      "loss": 1.3077,
+      "step": 5093
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007962399999999999,
+      "loss": 1.2884,
+      "step": 5094
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007962000000000001,
+      "loss": 1.2488,
+      "step": 5095
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079616,
+      "loss": 1.1299,
+      "step": 5096
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007961200000000001,
+      "loss": 1.3007,
+      "step": 5097
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079608,
+      "loss": 1.2843,
+      "step": 5098
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079604,
+      "loss": 1.3054,
+      "step": 5099
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000796,
+      "loss": 1.2513,
+      "step": 5100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007959600000000001,
+      "loss": 1.1797,
+      "step": 5101
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079592,
+      "loss": 1.2638,
+      "step": 5102
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079588,
+      "loss": 1.1537,
+      "step": 5103
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079584,
+      "loss": 1.2052,
+      "step": 5104
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007957999999999999,
+      "loss": 1.1893,
+      "step": 5105
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007957600000000001,
+      "loss": 1.2238,
+      "step": 5106
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079572,
+      "loss": 1.2339,
+      "step": 5107
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007956800000000001,
+      "loss": 1.2796,
+      "step": 5108
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079564,
+      "loss": 1.2417,
+      "step": 5109
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007956,
+      "loss": 1.2841,
+      "step": 5110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007955600000000001,
+      "loss": 1.2239,
+      "step": 5111
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079552,
+      "loss": 1.234,
+      "step": 5112
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079548,
+      "loss": 1.1946,
+      "step": 5113
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079544,
+      "loss": 1.2843,
+      "step": 5114
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007954,
+      "loss": 1.3221,
+      "step": 5115
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079536,
+      "loss": 1.2925,
+      "step": 5116
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007953200000000001,
+      "loss": 1.2523,
+      "step": 5117
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079528,
+      "loss": 1.2477,
+      "step": 5118
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079524,
+      "loss": 1.2521,
+      "step": 5119
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007952,
+      "loss": 1.187,
+      "step": 5120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007951599999999999,
+      "loss": 1.3419,
+      "step": 5121
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007951200000000001,
+      "loss": 1.2628,
+      "step": 5122
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079508,
+      "loss": 1.196,
+      "step": 5123
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079504,
+      "loss": 1.2727,
+      "step": 5124
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000795,
+      "loss": 1.2728,
+      "step": 5125
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079496,
+      "loss": 1.2913,
+      "step": 5126
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079492,
+      "loss": 1.2085,
+      "step": 5127
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007948800000000001,
+      "loss": 1.2762,
+      "step": 5128
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079484,
+      "loss": 1.1514,
+      "step": 5129
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007948,
+      "loss": 1.1899,
+      "step": 5130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079476,
+      "loss": 1.2026,
+      "step": 5131
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079472,
+      "loss": 1.2438,
+      "step": 5132
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007946800000000001,
+      "loss": 1.2303,
+      "step": 5133
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079464,
+      "loss": 1.2694,
+      "step": 5134
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007946,
+      "loss": 1.1821,
+      "step": 5135
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079456,
+      "loss": 1.1734,
+      "step": 5136
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007945200000000001,
+      "loss": 1.1586,
+      "step": 5137
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079448,
+      "loss": 1.2404,
+      "step": 5138
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007944400000000001,
+      "loss": 1.2702,
+      "step": 5139
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007944,
+      "loss": 1.2286,
+      "step": 5140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007943599999999999,
+      "loss": 1.1481,
+      "step": 5141
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079432,
+      "loss": 1.2706,
+      "step": 5142
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079428,
+      "loss": 1.1924,
+      "step": 5143
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079424,
+      "loss": 1.2246,
+      "step": 5144
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007942,
+      "loss": 1.3229,
+      "step": 5145
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079416,
+      "loss": 1.2808,
+      "step": 5146
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079412,
+      "loss": 1.1798,
+      "step": 5147
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007940800000000001,
+      "loss": 1.2279,
+      "step": 5148
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079404,
+      "loss": 1.2179,
+      "step": 5149
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007940000000000001,
+      "loss": 1.2527,
+      "step": 5150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079396,
+      "loss": 1.3117,
+      "step": 5151
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007939199999999999,
+      "loss": 1.2063,
+      "step": 5152
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007938800000000001,
+      "loss": 1.1741,
+      "step": 5153
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079384,
+      "loss": 1.2573,
+      "step": 5154
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007938,
+      "loss": 1.1788,
+      "step": 5155
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079376,
+      "loss": 1.1649,
+      "step": 5156
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079372,
+      "loss": 1.2239,
+      "step": 5157
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007936800000000001,
+      "loss": 1.2889,
+      "step": 5158
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007936400000000001,
+      "loss": 1.2084,
+      "step": 5159
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007936,
+      "loss": 1.3038,
+      "step": 5160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079356,
+      "loss": 1.2567,
+      "step": 5161
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079352,
+      "loss": 1.2667,
+      "step": 5162
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007934799999999999,
+      "loss": 1.2164,
+      "step": 5163
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007934400000000001,
+      "loss": 1.207,
+      "step": 5164
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007934,
+      "loss": 1.2304,
+      "step": 5165
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079336,
+      "loss": 1.2052,
+      "step": 5166
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079332,
+      "loss": 1.2501,
+      "step": 5167
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079328,
+      "loss": 1.2656,
+      "step": 5168
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079324,
+      "loss": 1.2564,
+      "step": 5169
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007932,
+      "loss": 1.2336,
+      "step": 5170
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079316,
+      "loss": 1.1688,
+      "step": 5171
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079312,
+      "loss": 1.2737,
+      "step": 5172
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079308,
+      "loss": 1.2161,
+      "step": 5173
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079304,
+      "loss": 1.2456,
+      "step": 5174
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007930000000000001,
+      "loss": 1.1912,
+      "step": 5175
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079296,
+      "loss": 1.2074,
+      "step": 5176
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079292,
+      "loss": 1.2326,
+      "step": 5177
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079288,
+      "loss": 1.2756,
+      "step": 5178
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007928400000000001,
+      "loss": 1.2349,
+      "step": 5179
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007928,
+      "loss": 1.1745,
+      "step": 5180
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079276,
+      "loss": 1.2281,
+      "step": 5181
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079272,
+      "loss": 1.1941,
+      "step": 5182
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079268,
+      "loss": 1.2499,
+      "step": 5183
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007926400000000001,
+      "loss": 1.2612,
+      "step": 5184
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007926,
+      "loss": 1.2564,
+      "step": 5185
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007925600000000001,
+      "loss": 1.1978,
+      "step": 5186
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079252,
+      "loss": 1.2107,
+      "step": 5187
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079248,
+      "loss": 1.2565,
+      "step": 5188
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079244,
+      "loss": 1.1209,
+      "step": 5189
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0007924,
+      "loss": 1.231,
+      "step": 5190
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079236,
+      "loss": 1.2489,
+      "step": 5191
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079232,
+      "loss": 1.311,
+      "step": 5192
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079228,
+      "loss": 1.3063,
+      "step": 5193
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00079224,
+      "loss": 1.2599,
+      "step": 5194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007922000000000001,
+      "loss": 1.2612,
+      "step": 5195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079216,
+      "loss": 1.2344,
+      "step": 5196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007921200000000001,
+      "loss": 1.2147,
+      "step": 5197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079208,
+      "loss": 1.1663,
+      "step": 5198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007920399999999999,
+      "loss": 1.334,
+      "step": 5199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007920000000000001,
+      "loss": 1.1489,
+      "step": 5200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079196,
+      "loss": 1.1759,
+      "step": 5201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079192,
+      "loss": 1.2343,
+      "step": 5202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079188,
+      "loss": 1.1955,
+      "step": 5203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079184,
+      "loss": 1.335,
+      "step": 5204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007918,
+      "loss": 1.1491,
+      "step": 5205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007917600000000001,
+      "loss": 1.2604,
+      "step": 5206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079172,
+      "loss": 1.2106,
+      "step": 5207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007916800000000001,
+      "loss": 1.211,
+      "step": 5208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079164,
+      "loss": 1.2276,
+      "step": 5209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007915999999999999,
+      "loss": 1.2832,
+      "step": 5210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007915600000000001,
+      "loss": 1.1882,
+      "step": 5211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079152,
+      "loss": 1.272,
+      "step": 5212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079148,
+      "loss": 1.2221,
+      "step": 5213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079144,
+      "loss": 1.3376,
+      "step": 5214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007914,
+      "loss": 1.2677,
+      "step": 5215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079136,
+      "loss": 1.2049,
+      "step": 5216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007913200000000001,
+      "loss": 1.297,
+      "step": 5217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079128,
+      "loss": 1.1713,
+      "step": 5218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079124,
+      "loss": 1.2641,
+      "step": 5219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007912,
+      "loss": 1.2608,
+      "step": 5220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079116,
+      "loss": 1.2343,
+      "step": 5221
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007911200000000001,
+      "loss": 1.2114,
+      "step": 5222
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079108,
+      "loss": 1.2232,
+      "step": 5223
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079104,
+      "loss": 1.2753,
+      "step": 5224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000791,
+      "loss": 1.2737,
+      "step": 5225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007909600000000001,
+      "loss": 1.1355,
+      "step": 5226
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079092,
+      "loss": 1.2767,
+      "step": 5227
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079088,
+      "loss": 1.2399,
+      "step": 5228
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079084,
+      "loss": 1.2145,
+      "step": 5229
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007907999999999999,
+      "loss": 1.3575,
+      "step": 5230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079076,
+      "loss": 1.2813,
+      "step": 5231
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079072,
+      "loss": 1.2943,
+      "step": 5232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007906800000000001,
+      "loss": 1.311,
+      "step": 5233
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079064,
+      "loss": 1.2564,
+      "step": 5234
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007906,
+      "loss": 1.2425,
+      "step": 5235
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079056,
+      "loss": 1.2415,
+      "step": 5236
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007905200000000001,
+      "loss": 1.2506,
+      "step": 5237
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079048,
+      "loss": 1.2655,
+      "step": 5238
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079044,
+      "loss": 1.2174,
+      "step": 5239
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007904,
+      "loss": 1.2146,
+      "step": 5240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007903599999999999,
+      "loss": 1.2581,
+      "step": 5241
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007903200000000001,
+      "loss": 1.253,
+      "step": 5242
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079028,
+      "loss": 1.2804,
+      "step": 5243
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007902400000000001,
+      "loss": 1.1774,
+      "step": 5244
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007902,
+      "loss": 1.3164,
+      "step": 5245
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079016,
+      "loss": 1.2181,
+      "step": 5246
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007901200000000001,
+      "loss": 1.2451,
+      "step": 5247
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079008,
+      "loss": 1.2685,
+      "step": 5248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079004,
+      "loss": 1.2418,
+      "step": 5249
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00079,
+      "loss": 1.2438,
+      "step": 5250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078996,
+      "loss": 1.2342,
+      "step": 5251
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007899199999999999,
+      "loss": 1.1699,
+      "step": 5252
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007898800000000001,
+      "loss": 1.2374,
+      "step": 5253
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078984,
+      "loss": 1.1984,
+      "step": 5254
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007898,
+      "loss": 1.1325,
+      "step": 5255
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078976,
+      "loss": 1.1441,
+      "step": 5256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007897199999999999,
+      "loss": 1.3189,
+      "step": 5257
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007896800000000001,
+      "loss": 1.2003,
+      "step": 5258
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078964,
+      "loss": 1.2061,
+      "step": 5259
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007896,
+      "loss": 1.304,
+      "step": 5260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078956,
+      "loss": 1.247,
+      "step": 5261
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078952,
+      "loss": 1.1881,
+      "step": 5262
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078948,
+      "loss": 1.2439,
+      "step": 5263
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007894400000000001,
+      "loss": 1.3112,
+      "step": 5264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007894,
+      "loss": 1.2501,
+      "step": 5265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078936,
+      "loss": 1.225,
+      "step": 5266
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078932,
+      "loss": 1.2388,
+      "step": 5267
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078928,
+      "loss": 1.1966,
+      "step": 5268
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007892400000000001,
+      "loss": 1.2108,
+      "step": 5269
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007892,
+      "loss": 1.2201,
+      "step": 5270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078916,
+      "loss": 1.2427,
+      "step": 5271
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078912,
+      "loss": 1.2124,
+      "step": 5272
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007890800000000001,
+      "loss": 1.231,
+      "step": 5273
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078904,
+      "loss": 1.29,
+      "step": 5274
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007890000000000001,
+      "loss": 1.1993,
+      "step": 5275
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078896,
+      "loss": 1.3001,
+      "step": 5276
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007889199999999999,
+      "loss": 1.1971,
+      "step": 5277
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078888,
+      "loss": 1.3051,
+      "step": 5278
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078884,
+      "loss": 1.1832,
+      "step": 5279
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007888,
+      "loss": 1.23,
+      "step": 5280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078876,
+      "loss": 1.188,
+      "step": 5281
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078872,
+      "loss": 1.2531,
+      "step": 5282
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078868,
+      "loss": 1.2706,
+      "step": 5283
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007886400000000001,
+      "loss": 1.2547,
+      "step": 5284
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007886,
+      "loss": 1.1838,
+      "step": 5285
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007885600000000001,
+      "loss": 1.2069,
+      "step": 5286
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078852,
+      "loss": 1.2456,
+      "step": 5287
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007884799999999999,
+      "loss": 1.2348,
+      "step": 5288
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007884400000000001,
+      "loss": 1.2493,
+      "step": 5289
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007884,
+      "loss": 1.143,
+      "step": 5290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078836,
+      "loss": 1.1713,
+      "step": 5291
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078832,
+      "loss": 1.2228,
+      "step": 5292
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078828,
+      "loss": 1.2326,
+      "step": 5293
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007882400000000001,
+      "loss": 1.3062,
+      "step": 5294
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007882000000000001,
+      "loss": 1.3171,
+      "step": 5295
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078816,
+      "loss": 1.2375,
+      "step": 5296
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078812,
+      "loss": 1.2341,
+      "step": 5297
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078808,
+      "loss": 1.1732,
+      "step": 5298
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007880399999999999,
+      "loss": 1.1835,
+      "step": 5299
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007880000000000001,
+      "loss": 1.1856,
+      "step": 5300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078796,
+      "loss": 1.3235,
+      "step": 5301
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078792,
+      "loss": 1.2802,
+      "step": 5302
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078788,
+      "loss": 1.2614,
+      "step": 5303
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078784,
+      "loss": 1.2635,
+      "step": 5304
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007878,
+      "loss": 1.2732,
+      "step": 5305
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078776,
+      "loss": 1.2845,
+      "step": 5306
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078772,
+      "loss": 1.3659,
+      "step": 5307
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078768,
+      "loss": 1.2992,
+      "step": 5308
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078764,
+      "loss": 1.1608,
+      "step": 5309
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007876,
+      "loss": 1.1497,
+      "step": 5310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007875600000000001,
+      "loss": 1.2559,
+      "step": 5311
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078752,
+      "loss": 1.207,
+      "step": 5312
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078748,
+      "loss": 1.2281,
+      "step": 5313
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078744,
+      "loss": 1.1695,
+      "step": 5314
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007874000000000001,
+      "loss": 1.2236,
+      "step": 5315
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078736,
+      "loss": 1.2273,
+      "step": 5316
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078732,
+      "loss": 1.2287,
+      "step": 5317
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078728,
+      "loss": 1.2511,
+      "step": 5318
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078724,
+      "loss": 1.1947,
+      "step": 5319
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007872,
+      "loss": 1.1681,
+      "step": 5320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078716,
+      "loss": 1.3459,
+      "step": 5321
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007871200000000001,
+      "loss": 1.2153,
+      "step": 5322
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078708,
+      "loss": 1.2125,
+      "step": 5323
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078704,
+      "loss": 1.3071,
+      "step": 5324
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000787,
+      "loss": 1.2302,
+      "step": 5325
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078696,
+      "loss": 1.1992,
+      "step": 5326
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078692,
+      "loss": 1.2486,
+      "step": 5327
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078688,
+      "loss": 1.1566,
+      "step": 5328
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078684,
+      "loss": 1.2323,
+      "step": 5329
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007868,
+      "loss": 1.1984,
+      "step": 5330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007867600000000001,
+      "loss": 1.3037,
+      "step": 5331
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078672,
+      "loss": 1.2157,
+      "step": 5332
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007866800000000001,
+      "loss": 1.2668,
+      "step": 5333
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078664,
+      "loss": 1.2498,
+      "step": 5334
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007865999999999999,
+      "loss": 1.2906,
+      "step": 5335
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007865600000000001,
+      "loss": 1.2623,
+      "step": 5336
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078652,
+      "loss": 1.2417,
+      "step": 5337
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078648,
+      "loss": 1.2254,
+      "step": 5338
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078644,
+      "loss": 1.2001,
+      "step": 5339
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007864,
+      "loss": 1.2451,
+      "step": 5340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007863599999999999,
+      "loss": 1.2084,
+      "step": 5341
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007863200000000001,
+      "loss": 1.2159,
+      "step": 5342
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078628,
+      "loss": 1.2631,
+      "step": 5343
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007862400000000001,
+      "loss": 1.1947,
+      "step": 5344
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007862,
+      "loss": 1.3705,
+      "step": 5345
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007861599999999999,
+      "loss": 1.2699,
+      "step": 5346
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007861200000000001,
+      "loss": 1.3125,
+      "step": 5347
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078608,
+      "loss": 1.2493,
+      "step": 5348
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078604,
+      "loss": 1.2988,
+      "step": 5349
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000786,
+      "loss": 1.2564,
+      "step": 5350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078596,
+      "loss": 1.1855,
+      "step": 5351
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078592,
+      "loss": 1.2629,
+      "step": 5352
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007858800000000001,
+      "loss": 1.183,
+      "step": 5353
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078584,
+      "loss": 1.277,
+      "step": 5354
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007858,
+      "loss": 1.1824,
+      "step": 5355
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078576,
+      "loss": 1.2323,
+      "step": 5356
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078572,
+      "loss": 1.2602,
+      "step": 5357
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007856800000000001,
+      "loss": 1.2103,
+      "step": 5358
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078564,
+      "loss": 1.2514,
+      "step": 5359
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007856,
+      "loss": 1.2453,
+      "step": 5360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078556,
+      "loss": 1.1986,
+      "step": 5361
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007855200000000001,
+      "loss": 1.2431,
+      "step": 5362
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078548,
+      "loss": 1.2104,
+      "step": 5363
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007854400000000001,
+      "loss": 1.2659,
+      "step": 5364
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007854,
+      "loss": 1.2148,
+      "step": 5365
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007853599999999999,
+      "loss": 1.2683,
+      "step": 5366
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078532,
+      "loss": 1.2199,
+      "step": 5367
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078528,
+      "loss": 1.2612,
+      "step": 5368
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007852400000000001,
+      "loss": 1.2227,
+      "step": 5369
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007852,
+      "loss": 1.2157,
+      "step": 5370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078516,
+      "loss": 1.2591,
+      "step": 5371
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078512,
+      "loss": 1.3042,
+      "step": 5372
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007850800000000001,
+      "loss": 1.2455,
+      "step": 5373
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078504,
+      "loss": 1.2389,
+      "step": 5374
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000785,
+      "loss": 1.1724,
+      "step": 5375
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078496,
+      "loss": 1.1361,
+      "step": 5376
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007849199999999999,
+      "loss": 1.1461,
+      "step": 5377
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007848800000000001,
+      "loss": 1.2359,
+      "step": 5378
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078484,
+      "loss": 1.2266,
+      "step": 5379
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007848000000000001,
+      "loss": 1.1858,
+      "step": 5380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078476,
+      "loss": 1.1683,
+      "step": 5381
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078472,
+      "loss": 1.1097,
+      "step": 5382
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007846800000000001,
+      "loss": 1.1655,
+      "step": 5383
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078464,
+      "loss": 1.2245,
+      "step": 5384
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007846,
+      "loss": 1.2094,
+      "step": 5385
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078456,
+      "loss": 1.3164,
+      "step": 5386
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078452,
+      "loss": 1.2051,
+      "step": 5387
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007844799999999999,
+      "loss": 1.339,
+      "step": 5388
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007844400000000001,
+      "loss": 1.1275,
+      "step": 5389
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007844,
+      "loss": 1.2255,
+      "step": 5390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078436,
+      "loss": 1.173,
+      "step": 5391
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078432,
+      "loss": 1.2271,
+      "step": 5392
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078428,
+      "loss": 1.298,
+      "step": 5393
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007842400000000001,
+      "loss": 1.2673,
+      "step": 5394
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007842,
+      "loss": 1.1599,
+      "step": 5395
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078416,
+      "loss": 1.245,
+      "step": 5396
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078412,
+      "loss": 1.2045,
+      "step": 5397
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078408,
+      "loss": 1.2245,
+      "step": 5398
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078404,
+      "loss": 1.2523,
+      "step": 5399
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007840000000000001,
+      "loss": 1.2973,
+      "step": 5400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078396,
+      "loss": 1.1799,
+      "step": 5401
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078392,
+      "loss": 1.2719,
+      "step": 5402
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078388,
+      "loss": 1.2545,
+      "step": 5403
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078384,
+      "loss": 1.2585,
+      "step": 5404
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007838000000000001,
+      "loss": 1.2182,
+      "step": 5405
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078376,
+      "loss": 1.2143,
+      "step": 5406
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078372,
+      "loss": 1.1989,
+      "step": 5407
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078368,
+      "loss": 1.2251,
+      "step": 5408
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078364,
+      "loss": 1.2556,
+      "step": 5409
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007836,
+      "loss": 1.2615,
+      "step": 5410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007835600000000001,
+      "loss": 1.2792,
+      "step": 5411
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078352,
+      "loss": 1.2837,
+      "step": 5412
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007834799999999999,
+      "loss": 1.1001,
+      "step": 5413
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078344,
+      "loss": 1.2124,
+      "step": 5414
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007834,
+      "loss": 1.1482,
+      "step": 5415
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078336,
+      "loss": 1.2677,
+      "step": 5416
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078332,
+      "loss": 1.285,
+      "step": 5417
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078328,
+      "loss": 1.2946,
+      "step": 5418
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078324,
+      "loss": 1.1926,
+      "step": 5419
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007832000000000001,
+      "loss": 1.3439,
+      "step": 5420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078316,
+      "loss": 1.2537,
+      "step": 5421
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007831200000000001,
+      "loss": 1.2485,
+      "step": 5422
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078308,
+      "loss": 1.1964,
+      "step": 5423
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007830399999999999,
+      "loss": 1.2974,
+      "step": 5424
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007830000000000001,
+      "loss": 1.2486,
+      "step": 5425
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078296,
+      "loss": 1.2003,
+      "step": 5426
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078292,
+      "loss": 1.2829,
+      "step": 5427
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078288,
+      "loss": 1.2405,
+      "step": 5428
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078284,
+      "loss": 1.1392,
+      "step": 5429
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007828,
+      "loss": 1.1559,
+      "step": 5430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007827600000000001,
+      "loss": 1.306,
+      "step": 5431
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078272,
+      "loss": 1.2211,
+      "step": 5432
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078268,
+      "loss": 1.3451,
+      "step": 5433
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078264,
+      "loss": 1.168,
+      "step": 5434
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007825999999999999,
+      "loss": 1.1871,
+      "step": 5435
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007825600000000001,
+      "loss": 1.1662,
+      "step": 5436
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078252,
+      "loss": 1.2541,
+      "step": 5437
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078248,
+      "loss": 1.2007,
+      "step": 5438
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078244,
+      "loss": 1.2632,
+      "step": 5439
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007824,
+      "loss": 1.227,
+      "step": 5440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007823600000000001,
+      "loss": 1.2158,
+      "step": 5441
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007823200000000001,
+      "loss": 1.2026,
+      "step": 5442
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078228,
+      "loss": 1.2159,
+      "step": 5443
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078224,
+      "loss": 1.2263,
+      "step": 5444
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007822,
+      "loss": 1.2126,
+      "step": 5445
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078216,
+      "loss": 1.1674,
+      "step": 5446
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0007821200000000001,
+      "loss": 1.2366,
+      "step": 5447
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00078208,
+      "loss": 1.1823,
+      "step": 5448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078204,
+      "loss": 1.3001,
+      "step": 5449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000782,
+      "loss": 1.2478,
+      "step": 5450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007819600000000001,
+      "loss": 1.2687,
+      "step": 5451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078192,
+      "loss": 1.2378,
+      "step": 5452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078188,
+      "loss": 1.1664,
+      "step": 5453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078184,
+      "loss": 1.1047,
+      "step": 5454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007818,
+      "loss": 1.1977,
+      "step": 5455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078176,
+      "loss": 1.2367,
+      "step": 5456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078172,
+      "loss": 1.2233,
+      "step": 5457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007816800000000001,
+      "loss": 1.2773,
+      "step": 5458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078164,
+      "loss": 1.2356,
+      "step": 5459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007816,
+      "loss": 1.2105,
+      "step": 5460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078156,
+      "loss": 1.2188,
+      "step": 5461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078152,
+      "loss": 1.1959,
+      "step": 5462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078148,
+      "loss": 1.3066,
+      "step": 5463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078144,
+      "loss": 1.2377,
+      "step": 5464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007814,
+      "loss": 1.2199,
+      "step": 5465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078136,
+      "loss": 1.1318,
+      "step": 5466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007813200000000001,
+      "loss": 1.2549,
+      "step": 5467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078128,
+      "loss": 1.2115,
+      "step": 5468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007812400000000001,
+      "loss": 1.2222,
+      "step": 5469
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007812,
+      "loss": 1.2369,
+      "step": 5470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078116,
+      "loss": 1.261,
+      "step": 5471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007811200000000001,
+      "loss": 1.2047,
+      "step": 5472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078108,
+      "loss": 1.2104,
+      "step": 5473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078104,
+      "loss": 1.2043,
+      "step": 5474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000781,
+      "loss": 1.2299,
+      "step": 5475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078096,
+      "loss": 1.2633,
+      "step": 5476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007809199999999999,
+      "loss": 1.3132,
+      "step": 5477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007808800000000001,
+      "loss": 1.1576,
+      "step": 5478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078084,
+      "loss": 1.1641,
+      "step": 5479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007808000000000001,
+      "loss": 1.1593,
+      "step": 5480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078076,
+      "loss": 1.2068,
+      "step": 5481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007807199999999999,
+      "loss": 1.2046,
+      "step": 5482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007806800000000001,
+      "loss": 1.2474,
+      "step": 5483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078064,
+      "loss": 1.2219,
+      "step": 5484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007806,
+      "loss": 1.2052,
+      "step": 5485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078056,
+      "loss": 1.2123,
+      "step": 5486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078052,
+      "loss": 1.3034,
+      "step": 5487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078048,
+      "loss": 1.2467,
+      "step": 5488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007804400000000001,
+      "loss": 1.2848,
+      "step": 5489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007804,
+      "loss": 1.2621,
+      "step": 5490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078036,
+      "loss": 1.2562,
+      "step": 5491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078032,
+      "loss": 1.237,
+      "step": 5492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078028,
+      "loss": 1.2171,
+      "step": 5493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007802400000000001,
+      "loss": 1.2828,
+      "step": 5494
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007802,
+      "loss": 1.1785,
+      "step": 5495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078016,
+      "loss": 1.1953,
+      "step": 5496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078012,
+      "loss": 1.3212,
+      "step": 5497
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078008,
+      "loss": 1.1707,
+      "step": 5498
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00078004,
+      "loss": 1.1946,
+      "step": 5499
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007800000000000001,
+      "loss": 1.2199,
+      "step": 5500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077996,
+      "loss": 1.2599,
+      "step": 5501
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007799199999999999,
+      "loss": 1.2096,
+      "step": 5502
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077988,
+      "loss": 1.2157,
+      "step": 5503
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077984,
+      "loss": 1.24,
+      "step": 5504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007798000000000001,
+      "loss": 1.2254,
+      "step": 5505
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077976,
+      "loss": 1.1435,
+      "step": 5506
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077972,
+      "loss": 1.2031,
+      "step": 5507
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077968,
+      "loss": 1.238,
+      "step": 5508
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007796400000000001,
+      "loss": 1.2016,
+      "step": 5509
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007796,
+      "loss": 1.3239,
+      "step": 5510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077956,
+      "loss": 1.1828,
+      "step": 5511
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077952,
+      "loss": 1.2464,
+      "step": 5512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007794799999999999,
+      "loss": 1.1967,
+      "step": 5513
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007794400000000001,
+      "loss": 1.2385,
+      "step": 5514
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007794,
+      "loss": 1.2521,
+      "step": 5515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007793600000000001,
+      "loss": 1.1944,
+      "step": 5516
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077932,
+      "loss": 1.2331,
+      "step": 5517
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077928,
+      "loss": 1.1046,
+      "step": 5518
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077924,
+      "loss": 1.2384,
+      "step": 5519
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007792,
+      "loss": 1.2521,
+      "step": 5520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077916,
+      "loss": 1.2612,
+      "step": 5521
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077912,
+      "loss": 1.2237,
+      "step": 5522
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077908,
+      "loss": 1.2574,
+      "step": 5523
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007790399999999999,
+      "loss": 1.2186,
+      "step": 5524
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007790000000000001,
+      "loss": 1.3132,
+      "step": 5525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077896,
+      "loss": 1.3071,
+      "step": 5526
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077892,
+      "loss": 1.2295,
+      "step": 5527
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077888,
+      "loss": 1.2747,
+      "step": 5528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077884,
+      "loss": 1.1755,
+      "step": 5529
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007788000000000001,
+      "loss": 1.2092,
+      "step": 5530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077876,
+      "loss": 1.2509,
+      "step": 5531
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077872,
+      "loss": 1.2002,
+      "step": 5532
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077868,
+      "loss": 1.3172,
+      "step": 5533
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077864,
+      "loss": 1.1862,
+      "step": 5534
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007786,
+      "loss": 1.3387,
+      "step": 5535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007785600000000001,
+      "loss": 1.2211,
+      "step": 5536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077852,
+      "loss": 1.2285,
+      "step": 5537
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077848,
+      "loss": 1.2057,
+      "step": 5538
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077844,
+      "loss": 1.2461,
+      "step": 5539
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007784,
+      "loss": 1.2077,
+      "step": 5540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007783600000000001,
+      "loss": 1.1969,
+      "step": 5541
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077832,
+      "loss": 1.2399,
+      "step": 5542
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077828,
+      "loss": 1.1782,
+      "step": 5543
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077824,
+      "loss": 1.1905,
+      "step": 5544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007782,
+      "loss": 1.2144,
+      "step": 5545
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077816,
+      "loss": 1.1563,
+      "step": 5546
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007781200000000001,
+      "loss": 1.2273,
+      "step": 5547
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077808,
+      "loss": 1.2931,
+      "step": 5548
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007780399999999999,
+      "loss": 1.3254,
+      "step": 5549
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000778,
+      "loss": 1.1811,
+      "step": 5550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077796,
+      "loss": 1.2331,
+      "step": 5551
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077792,
+      "loss": 1.2275,
+      "step": 5552
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077788,
+      "loss": 1.2188,
+      "step": 5553
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077784,
+      "loss": 1.2794,
+      "step": 5554
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007778,
+      "loss": 1.172,
+      "step": 5555
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007777600000000001,
+      "loss": 1.2162,
+      "step": 5556
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077772,
+      "loss": 1.197,
+      "step": 5557
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007776800000000001,
+      "loss": 1.2075,
+      "step": 5558
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077764,
+      "loss": 1.2628,
+      "step": 5559
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007775999999999999,
+      "loss": 1.2059,
+      "step": 5560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007775600000000001,
+      "loss": 1.2041,
+      "step": 5561
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077752,
+      "loss": 1.2218,
+      "step": 5562
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077748,
+      "loss": 1.2454,
+      "step": 5563
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077744,
+      "loss": 1.2926,
+      "step": 5564
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007774,
+      "loss": 1.2175,
+      "step": 5565
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077736,
+      "loss": 1.2084,
+      "step": 5566
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007773200000000001,
+      "loss": 1.2134,
+      "step": 5567
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077728,
+      "loss": 1.1974,
+      "step": 5568
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077724,
+      "loss": 1.2022,
+      "step": 5569
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007772,
+      "loss": 1.18,
+      "step": 5570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007771599999999999,
+      "loss": 1.1535,
+      "step": 5571
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007771200000000001,
+      "loss": 1.3069,
+      "step": 5572
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077708,
+      "loss": 1.3165,
+      "step": 5573
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077704,
+      "loss": 1.1952,
+      "step": 5574
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000777,
+      "loss": 1.2835,
+      "step": 5575
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077696,
+      "loss": 1.2147,
+      "step": 5576
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007769200000000001,
+      "loss": 1.2221,
+      "step": 5577
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007768800000000001,
+      "loss": 1.1586,
+      "step": 5578
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077684,
+      "loss": 1.2536,
+      "step": 5579
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007768,
+      "loss": 1.1394,
+      "step": 5580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077676,
+      "loss": 1.2012,
+      "step": 5581
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077672,
+      "loss": 1.3557,
+      "step": 5582
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007766800000000001,
+      "loss": 1.1631,
+      "step": 5583
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077664,
+      "loss": 1.2184,
+      "step": 5584
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007766,
+      "loss": 1.3144,
+      "step": 5585
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077656,
+      "loss": 1.1872,
+      "step": 5586
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077652,
+      "loss": 1.2559,
+      "step": 5587
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077648,
+      "loss": 1.1799,
+      "step": 5588
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077644,
+      "loss": 1.1739,
+      "step": 5589
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007764,
+      "loss": 1.066,
+      "step": 5590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077636,
+      "loss": 1.2431,
+      "step": 5591
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077632,
+      "loss": 1.264,
+      "step": 5592
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077628,
+      "loss": 1.2611,
+      "step": 5593
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007762400000000001,
+      "loss": 1.1482,
+      "step": 5594
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007762,
+      "loss": 1.1337,
+      "step": 5595
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077616,
+      "loss": 1.2468,
+      "step": 5596
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077612,
+      "loss": 1.3221,
+      "step": 5597
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077608,
+      "loss": 1.2142,
+      "step": 5598
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077604,
+      "loss": 1.1496,
+      "step": 5599
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000776,
+      "loss": 1.2329,
+      "step": 5600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077596,
+      "loss": 1.274,
+      "step": 5601
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077592,
+      "loss": 1.2114,
+      "step": 5602
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007758800000000001,
+      "loss": 1.2754,
+      "step": 5603
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077584,
+      "loss": 1.2194,
+      "step": 5604
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007758000000000001,
+      "loss": 1.2294,
+      "step": 5605
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077576,
+      "loss": 1.2144,
+      "step": 5606
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077572,
+      "loss": 1.2336,
+      "step": 5607
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007756800000000001,
+      "loss": 1.2237,
+      "step": 5608
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077564,
+      "loss": 1.2683,
+      "step": 5609
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007756,
+      "loss": 1.1602,
+      "step": 5610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077556,
+      "loss": 1.2002,
+      "step": 5611
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077552,
+      "loss": 1.2723,
+      "step": 5612
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007754799999999999,
+      "loss": 1.2525,
+      "step": 5613
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007754400000000001,
+      "loss": 1.1883,
+      "step": 5614
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007754,
+      "loss": 1.2127,
+      "step": 5615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007753600000000001,
+      "loss": 1.2155,
+      "step": 5616
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077532,
+      "loss": 1.1894,
+      "step": 5617
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007752799999999999,
+      "loss": 1.1068,
+      "step": 5618
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007752400000000001,
+      "loss": 1.2266,
+      "step": 5619
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007752,
+      "loss": 1.2312,
+      "step": 5620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077516,
+      "loss": 1.2025,
+      "step": 5621
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077512,
+      "loss": 1.1927,
+      "step": 5622
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077508,
+      "loss": 1.1702,
+      "step": 5623
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077504,
+      "loss": 1.1937,
+      "step": 5624
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007750000000000001,
+      "loss": 1.2449,
+      "step": 5625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077496,
+      "loss": 1.1783,
+      "step": 5626
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007749200000000001,
+      "loss": 1.2177,
+      "step": 5627
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077488,
+      "loss": 1.1639,
+      "step": 5628
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077484,
+      "loss": 1.2676,
+      "step": 5629
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007748000000000001,
+      "loss": 1.1947,
+      "step": 5630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077476,
+      "loss": 1.3043,
+      "step": 5631
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077472,
+      "loss": 1.2008,
+      "step": 5632
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077468,
+      "loss": 1.1362,
+      "step": 5633
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077464,
+      "loss": 1.191,
+      "step": 5634
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007746,
+      "loss": 1.234,
+      "step": 5635
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007745600000000001,
+      "loss": 1.2055,
+      "step": 5636
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077452,
+      "loss": 1.2397,
+      "step": 5637
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007744799999999999,
+      "loss": 1.241,
+      "step": 5638
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077444,
+      "loss": 1.2571,
+      "step": 5639
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007744,
+      "loss": 1.1605,
+      "step": 5640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007743600000000001,
+      "loss": 1.2057,
+      "step": 5641
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077432,
+      "loss": 1.2499,
+      "step": 5642
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077428,
+      "loss": 1.2008,
+      "step": 5643
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077424,
+      "loss": 1.3037,
+      "step": 5644
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007742000000000001,
+      "loss": 1.2241,
+      "step": 5645
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077416,
+      "loss": 1.2876,
+      "step": 5646
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077412,
+      "loss": 1.1757,
+      "step": 5647
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077408,
+      "loss": 1.2041,
+      "step": 5648
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007740399999999999,
+      "loss": 1.2429,
+      "step": 5649
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007740000000000001,
+      "loss": 1.2225,
+      "step": 5650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077396,
+      "loss": 1.2105,
+      "step": 5651
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007739200000000001,
+      "loss": 1.2154,
+      "step": 5652
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077388,
+      "loss": 1.1572,
+      "step": 5653
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077384,
+      "loss": 1.2849,
+      "step": 5654
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007738,
+      "loss": 1.2497,
+      "step": 5655
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007737600000000001,
+      "loss": 1.1736,
+      "step": 5656
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077372,
+      "loss": 1.2471,
+      "step": 5657
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077368,
+      "loss": 1.1937,
+      "step": 5658
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077364,
+      "loss": 1.252,
+      "step": 5659
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007735999999999999,
+      "loss": 1.2246,
+      "step": 5660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007735600000000001,
+      "loss": 1.2156,
+      "step": 5661
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077352,
+      "loss": 1.1836,
+      "step": 5662
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077348,
+      "loss": 1.2259,
+      "step": 5663
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077344,
+      "loss": 1.2252,
+      "step": 5664
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007734,
+      "loss": 1.2519,
+      "step": 5665
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007733600000000001,
+      "loss": 1.2107,
+      "step": 5666
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077332,
+      "loss": 1.235,
+      "step": 5667
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077328,
+      "loss": 1.0791,
+      "step": 5668
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077324,
+      "loss": 1.1764,
+      "step": 5669
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007732,
+      "loss": 1.2525,
+      "step": 5670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077316,
+      "loss": 1.2567,
+      "step": 5671
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007731200000000001,
+      "loss": 1.1541,
+      "step": 5672
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077308,
+      "loss": 1.2803,
+      "step": 5673
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077304,
+      "loss": 1.2556,
+      "step": 5674
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000773,
+      "loss": 1.2297,
+      "step": 5675
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007729599999999999,
+      "loss": 1.1831,
+      "step": 5676
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007729200000000001,
+      "loss": 1.2374,
+      "step": 5677
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077288,
+      "loss": 1.199,
+      "step": 5678
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077284,
+      "loss": 1.2315,
+      "step": 5679
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007728,
+      "loss": 1.2324,
+      "step": 5680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077276,
+      "loss": 1.2324,
+      "step": 5681
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077272,
+      "loss": 1.2252,
+      "step": 5682
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007726800000000001,
+      "loss": 1.1438,
+      "step": 5683
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077264,
+      "loss": 1.189,
+      "step": 5684
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007726,
+      "loss": 1.1198,
+      "step": 5685
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077256,
+      "loss": 1.1549,
+      "step": 5686
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077252,
+      "loss": 1.2945,
+      "step": 5687
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007724800000000001,
+      "loss": 1.2599,
+      "step": 5688
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077244,
+      "loss": 1.2701,
+      "step": 5689
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007724,
+      "loss": 1.192,
+      "step": 5690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077236,
+      "loss": 1.1962,
+      "step": 5691
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007723200000000001,
+      "loss": 1.1844,
+      "step": 5692
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077228,
+      "loss": 1.2059,
+      "step": 5693
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007722400000000001,
+      "loss": 1.1728,
+      "step": 5694
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007722,
+      "loss": 1.3127,
+      "step": 5695
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007721599999999999,
+      "loss": 1.2069,
+      "step": 5696
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0007721200000000001,
+      "loss": 1.1633,
+      "step": 5697
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077208,
+      "loss": 1.2231,
+      "step": 5698
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077204,
+      "loss": 1.1615,
+      "step": 5699
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000772,
+      "loss": 1.2391,
+      "step": 5700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00077196,
+      "loss": 1.1006,
+      "step": 5701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077192,
+      "loss": 1.3325,
+      "step": 5702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007718800000000001,
+      "loss": 1.2189,
+      "step": 5703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077184,
+      "loss": 1.1507,
+      "step": 5704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007718000000000001,
+      "loss": 1.2262,
+      "step": 5705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077176,
+      "loss": 1.208,
+      "step": 5706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007717199999999999,
+      "loss": 1.1845,
+      "step": 5707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007716800000000001,
+      "loss": 1.1858,
+      "step": 5708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077164,
+      "loss": 1.1758,
+      "step": 5709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007716,
+      "loss": 1.2118,
+      "step": 5710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077156,
+      "loss": 1.2111,
+      "step": 5711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077152,
+      "loss": 1.2749,
+      "step": 5712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007714800000000001,
+      "loss": 1.1979,
+      "step": 5713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007714400000000001,
+      "loss": 1.2986,
+      "step": 5714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007714,
+      "loss": 1.131,
+      "step": 5715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077136,
+      "loss": 1.0831,
+      "step": 5716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077132,
+      "loss": 1.2778,
+      "step": 5717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077128,
+      "loss": 1.1457,
+      "step": 5718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007712400000000001,
+      "loss": 1.1879,
+      "step": 5719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007712,
+      "loss": 1.2494,
+      "step": 5720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077116,
+      "loss": 1.2344,
+      "step": 5721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077112,
+      "loss": 1.2297,
+      "step": 5722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077108,
+      "loss": 1.2078,
+      "step": 5723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077104,
+      "loss": 1.2517,
+      "step": 5724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000771,
+      "loss": 1.1516,
+      "step": 5725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077096,
+      "loss": 1.2559,
+      "step": 5726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077092,
+      "loss": 1.1575,
+      "step": 5727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077088,
+      "loss": 1.2651,
+      "step": 5728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077084,
+      "loss": 1.1901,
+      "step": 5729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007708000000000001,
+      "loss": 1.2072,
+      "step": 5730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077076,
+      "loss": 1.255,
+      "step": 5731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077072,
+      "loss": 1.209,
+      "step": 5732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077068,
+      "loss": 1.2728,
+      "step": 5733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077064,
+      "loss": 1.1871,
+      "step": 5734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007706,
+      "loss": 1.2049,
+      "step": 5735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077056,
+      "loss": 1.2487,
+      "step": 5736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077052,
+      "loss": 1.1213,
+      "step": 5737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077048,
+      "loss": 1.2026,
+      "step": 5738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007704400000000001,
+      "loss": 1.2452,
+      "step": 5739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007704,
+      "loss": 1.1964,
+      "step": 5740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007703600000000001,
+      "loss": 1.1722,
+      "step": 5741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077032,
+      "loss": 1.2046,
+      "step": 5742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077028,
+      "loss": 1.2078,
+      "step": 5743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077024,
+      "loss": 1.2028,
+      "step": 5744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007702,
+      "loss": 1.1985,
+      "step": 5745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077016,
+      "loss": 1.2201,
+      "step": 5746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077012,
+      "loss": 1.2332,
+      "step": 5747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00077008,
+      "loss": 1.2746,
+      "step": 5748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007700399999999999,
+      "loss": 1.1751,
+      "step": 5749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007700000000000001,
+      "loss": 1.2522,
+      "step": 5750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076996,
+      "loss": 1.1208,
+      "step": 5751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007699200000000001,
+      "loss": 1.2214,
+      "step": 5752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076988,
+      "loss": 1.1881,
+      "step": 5753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007698399999999999,
+      "loss": 1.1685,
+      "step": 5754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007698000000000001,
+      "loss": 1.2392,
+      "step": 5755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076976,
+      "loss": 1.2417,
+      "step": 5756
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076972,
+      "loss": 1.2086,
+      "step": 5757
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076968,
+      "loss": 1.1747,
+      "step": 5758
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076964,
+      "loss": 1.1868,
+      "step": 5759
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007696,
+      "loss": 1.2013,
+      "step": 5760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007695600000000001,
+      "loss": 1.1725,
+      "step": 5761
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076952,
+      "loss": 1.2101,
+      "step": 5762
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007694800000000001,
+      "loss": 1.1183,
+      "step": 5763
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076944,
+      "loss": 1.28,
+      "step": 5764
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007693999999999999,
+      "loss": 1.196,
+      "step": 5765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007693600000000001,
+      "loss": 1.271,
+      "step": 5766
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076932,
+      "loss": 1.264,
+      "step": 5767
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076928,
+      "loss": 1.2494,
+      "step": 5768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076924,
+      "loss": 1.214,
+      "step": 5769
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007692,
+      "loss": 1.1864,
+      "step": 5770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076916,
+      "loss": 1.2527,
+      "step": 5771
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007691200000000001,
+      "loss": 1.1702,
+      "step": 5772
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076908,
+      "loss": 1.2239,
+      "step": 5773
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007690399999999999,
+      "loss": 1.1798,
+      "step": 5774
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000769,
+      "loss": 1.1811,
+      "step": 5775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076896,
+      "loss": 1.1907,
+      "step": 5776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007689200000000001,
+      "loss": 1.2462,
+      "step": 5777
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076888,
+      "loss": 1.1635,
+      "step": 5778
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076884,
+      "loss": 1.2314,
+      "step": 5779
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007688,
+      "loss": 1.2141,
+      "step": 5780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007687600000000001,
+      "loss": 1.1805,
+      "step": 5781
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076872,
+      "loss": 1.1919,
+      "step": 5782
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076868,
+      "loss": 1.2456,
+      "step": 5783
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076864,
+      "loss": 1.2148,
+      "step": 5784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007685999999999999,
+      "loss": 1.1729,
+      "step": 5785
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007685600000000001,
+      "loss": 1.2315,
+      "step": 5786
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076852,
+      "loss": 1.2529,
+      "step": 5787
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007684800000000001,
+      "loss": 1.2755,
+      "step": 5788
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076844,
+      "loss": 1.153,
+      "step": 5789
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007684,
+      "loss": 1.2722,
+      "step": 5790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076836,
+      "loss": 1.1842,
+      "step": 5791
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007683200000000001,
+      "loss": 1.134,
+      "step": 5792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076828,
+      "loss": 1.2057,
+      "step": 5793
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076824,
+      "loss": 1.1463,
+      "step": 5794
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007682,
+      "loss": 1.2181,
+      "step": 5795
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007681599999999999,
+      "loss": 1.1694,
+      "step": 5796
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007681200000000001,
+      "loss": 1.2252,
+      "step": 5797
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076808,
+      "loss": 1.224,
+      "step": 5798
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076804,
+      "loss": 1.2298,
+      "step": 5799
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000768,
+      "loss": 1.2034,
+      "step": 5800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076796,
+      "loss": 1.1756,
+      "step": 5801
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007679200000000001,
+      "loss": 1.2057,
+      "step": 5802
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076788,
+      "loss": 1.1883,
+      "step": 5803
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076784,
+      "loss": 1.2476,
+      "step": 5804
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007678,
+      "loss": 1.1567,
+      "step": 5805
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076776,
+      "loss": 1.195,
+      "step": 5806
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076772,
+      "loss": 1.2419,
+      "step": 5807
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007676800000000001,
+      "loss": 1.1675,
+      "step": 5808
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076764,
+      "loss": 1.1945,
+      "step": 5809
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007676,
+      "loss": 1.1556,
+      "step": 5810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076756,
+      "loss": 1.2039,
+      "step": 5811
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007675199999999999,
+      "loss": 1.1835,
+      "step": 5812
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007674800000000001,
+      "loss": 1.1985,
+      "step": 5813
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076744,
+      "loss": 1.1596,
+      "step": 5814
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007674,
+      "loss": 1.1968,
+      "step": 5815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076736,
+      "loss": 1.1783,
+      "step": 5816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076732,
+      "loss": 1.1779,
+      "step": 5817
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076728,
+      "loss": 1.2076,
+      "step": 5818
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007672400000000001,
+      "loss": 1.2405,
+      "step": 5819
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007672,
+      "loss": 1.2335,
+      "step": 5820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076716,
+      "loss": 1.2214,
+      "step": 5821
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076712,
+      "loss": 1.237,
+      "step": 5822
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076708,
+      "loss": 1.1383,
+      "step": 5823
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007670400000000001,
+      "loss": 1.1831,
+      "step": 5824
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000767,
+      "loss": 1.1249,
+      "step": 5825
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076696,
+      "loss": 1.2479,
+      "step": 5826
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076692,
+      "loss": 1.325,
+      "step": 5827
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007668800000000001,
+      "loss": 1.2482,
+      "step": 5828
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076684,
+      "loss": 1.2191,
+      "step": 5829
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007668000000000001,
+      "loss": 1.2113,
+      "step": 5830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076676,
+      "loss": 1.1528,
+      "step": 5831
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007667199999999999,
+      "loss": 1.2139,
+      "step": 5832
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076668,
+      "loss": 1.2192,
+      "step": 5833
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076664,
+      "loss": 1.2786,
+      "step": 5834
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007666,
+      "loss": 1.3488,
+      "step": 5835
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076656,
+      "loss": 1.2434,
+      "step": 5836
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076652,
+      "loss": 1.1215,
+      "step": 5837
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076648,
+      "loss": 1.2957,
+      "step": 5838
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007664400000000001,
+      "loss": 1.1896,
+      "step": 5839
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007664,
+      "loss": 1.206,
+      "step": 5840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007663600000000001,
+      "loss": 1.1959,
+      "step": 5841
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076632,
+      "loss": 1.3243,
+      "step": 5842
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007662799999999999,
+      "loss": 1.2374,
+      "step": 5843
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007662400000000001,
+      "loss": 1.223,
+      "step": 5844
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007662,
+      "loss": 1.251,
+      "step": 5845
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076616,
+      "loss": 1.2015,
+      "step": 5846
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076612,
+      "loss": 1.1939,
+      "step": 5847
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076608,
+      "loss": 1.2329,
+      "step": 5848
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007660400000000001,
+      "loss": 1.2053,
+      "step": 5849
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007660000000000001,
+      "loss": 1.2199,
+      "step": 5850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076596,
+      "loss": 1.2697,
+      "step": 5851
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076592,
+      "loss": 1.3337,
+      "step": 5852
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076588,
+      "loss": 1.1374,
+      "step": 5853
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007658399999999999,
+      "loss": 1.1997,
+      "step": 5854
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007658000000000001,
+      "loss": 1.2414,
+      "step": 5855
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076576,
+      "loss": 1.2054,
+      "step": 5856
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076572,
+      "loss": 1.1189,
+      "step": 5857
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076568,
+      "loss": 1.1944,
+      "step": 5858
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076564,
+      "loss": 1.1464,
+      "step": 5859
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007656,
+      "loss": 1.1088,
+      "step": 5860
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076556,
+      "loss": 1.2157,
+      "step": 5861
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076552,
+      "loss": 1.2269,
+      "step": 5862
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076548,
+      "loss": 1.2321,
+      "step": 5863
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076544,
+      "loss": 1.1489,
+      "step": 5864
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007654,
+      "loss": 1.2074,
+      "step": 5865
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007653600000000001,
+      "loss": 1.2348,
+      "step": 5866
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076532,
+      "loss": 1.1811,
+      "step": 5867
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076528,
+      "loss": 1.2674,
+      "step": 5868
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076524,
+      "loss": 1.2193,
+      "step": 5869
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007652000000000001,
+      "loss": 1.1987,
+      "step": 5870
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076516,
+      "loss": 1.1929,
+      "step": 5871
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076512,
+      "loss": 1.1492,
+      "step": 5872
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076508,
+      "loss": 1.1667,
+      "step": 5873
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076504,
+      "loss": 1.1988,
+      "step": 5874
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007650000000000001,
+      "loss": 1.1561,
+      "step": 5875
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076496,
+      "loss": 1.1128,
+      "step": 5876
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007649200000000001,
+      "loss": 1.1727,
+      "step": 5877
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076488,
+      "loss": 1.1932,
+      "step": 5878
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076484,
+      "loss": 1.1868,
+      "step": 5879
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007648,
+      "loss": 1.1401,
+      "step": 5880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076476,
+      "loss": 1.2588,
+      "step": 5881
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076472,
+      "loss": 1.2311,
+      "step": 5882
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076468,
+      "loss": 1.2987,
+      "step": 5883
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076464,
+      "loss": 1.1214,
+      "step": 5884
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007645999999999999,
+      "loss": 1.1295,
+      "step": 5885
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007645600000000001,
+      "loss": 1.1735,
+      "step": 5886
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076452,
+      "loss": 1.2063,
+      "step": 5887
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007644800000000001,
+      "loss": 1.2395,
+      "step": 5888
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076444,
+      "loss": 1.1844,
+      "step": 5889
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007643999999999999,
+      "loss": 1.1976,
+      "step": 5890
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007643600000000001,
+      "loss": 1.209,
+      "step": 5891
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076432,
+      "loss": 1.2704,
+      "step": 5892
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076428,
+      "loss": 1.2431,
+      "step": 5893
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076424,
+      "loss": 1.2697,
+      "step": 5894
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007642,
+      "loss": 1.3142,
+      "step": 5895
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076416,
+      "loss": 1.2621,
+      "step": 5896
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007641200000000001,
+      "loss": 1.317,
+      "step": 5897
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076408,
+      "loss": 1.1432,
+      "step": 5898
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007640400000000001,
+      "loss": 1.3111,
+      "step": 5899
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000764,
+      "loss": 1.1516,
+      "step": 5900
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007639599999999999,
+      "loss": 1.1709,
+      "step": 5901
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007639200000000001,
+      "loss": 1.1385,
+      "step": 5902
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076388,
+      "loss": 1.1992,
+      "step": 5903
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076384,
+      "loss": 1.2219,
+      "step": 5904
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007638,
+      "loss": 1.1784,
+      "step": 5905
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076376,
+      "loss": 1.2,
+      "step": 5906
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076372,
+      "loss": 1.2561,
+      "step": 5907
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007636800000000001,
+      "loss": 1.1723,
+      "step": 5908
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076364,
+      "loss": 1.2186,
+      "step": 5909
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007635999999999999,
+      "loss": 1.2608,
+      "step": 5910
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076356,
+      "loss": 1.2152,
+      "step": 5911
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076352,
+      "loss": 1.1903,
+      "step": 5912
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007634800000000001,
+      "loss": 1.2353,
+      "step": 5913
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076344,
+      "loss": 1.1095,
+      "step": 5914
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007634,
+      "loss": 1.2246,
+      "step": 5915
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076336,
+      "loss": 1.1837,
+      "step": 5916
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007633200000000001,
+      "loss": 1.2134,
+      "step": 5917
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076328,
+      "loss": 1.2565,
+      "step": 5918
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007632400000000001,
+      "loss": 1.19,
+      "step": 5919
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007632,
+      "loss": 1.1887,
+      "step": 5920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007631599999999999,
+      "loss": 1.2852,
+      "step": 5921
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076312,
+      "loss": 1.1753,
+      "step": 5922
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076308,
+      "loss": 1.1338,
+      "step": 5923
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007630400000000001,
+      "loss": 1.163,
+      "step": 5924
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000763,
+      "loss": 1.2437,
+      "step": 5925
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076296,
+      "loss": 1.2101,
+      "step": 5926
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076292,
+      "loss": 1.2304,
+      "step": 5927
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007628800000000001,
+      "loss": 1.2263,
+      "step": 5928
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076284,
+      "loss": 1.2042,
+      "step": 5929
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007628,
+      "loss": 1.2431,
+      "step": 5930
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076276,
+      "loss": 1.2004,
+      "step": 5931
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007627199999999999,
+      "loss": 1.1765,
+      "step": 5932
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007626800000000001,
+      "loss": 1.2323,
+      "step": 5933
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076264,
+      "loss": 1.2718,
+      "step": 5934
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007626,
+      "loss": 1.2005,
+      "step": 5935
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076256,
+      "loss": 1.2126,
+      "step": 5936
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076252,
+      "loss": 1.1649,
+      "step": 5937
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007624800000000001,
+      "loss": 1.1688,
+      "step": 5938
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076244,
+      "loss": 1.2278,
+      "step": 5939
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007624,
+      "loss": 1.2093,
+      "step": 5940
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076236,
+      "loss": 1.1216,
+      "step": 5941
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076232,
+      "loss": 1.2081,
+      "step": 5942
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007622799999999999,
+      "loss": 1.2169,
+      "step": 5943
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007622400000000001,
+      "loss": 1.0748,
+      "step": 5944
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007622,
+      "loss": 1.1745,
+      "step": 5945
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076216,
+      "loss": 1.2213,
+      "step": 5946
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076212,
+      "loss": 1.2206,
+      "step": 5947
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076208,
+      "loss": 1.1135,
+      "step": 5948
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0007620400000000001,
+      "loss": 1.229,
+      "step": 5949
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000762,
+      "loss": 1.1897,
+      "step": 5950
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076196,
+      "loss": 1.1724,
+      "step": 5951
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076192,
+      "loss": 1.2224,
+      "step": 5952
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076188,
+      "loss": 1.2019,
+      "step": 5953
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00076184,
+      "loss": 1.3108,
+      "step": 5954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007618000000000001,
+      "loss": 1.2213,
+      "step": 5955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076176,
+      "loss": 1.1444,
+      "step": 5956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076172,
+      "loss": 1.248,
+      "step": 5957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076168,
+      "loss": 1.1447,
+      "step": 5958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076164,
+      "loss": 1.187,
+      "step": 5959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007616000000000001,
+      "loss": 1.2155,
+      "step": 5960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076156,
+      "loss": 1.2768,
+      "step": 5961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076152,
+      "loss": 1.2241,
+      "step": 5962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076148,
+      "loss": 1.1626,
+      "step": 5963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007614400000000001,
+      "loss": 1.2309,
+      "step": 5964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007614,
+      "loss": 1.1548,
+      "step": 5965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007613600000000001,
+      "loss": 1.1896,
+      "step": 5966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076132,
+      "loss": 1.1509,
+      "step": 5967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007612799999999999,
+      "loss": 1.2678,
+      "step": 5968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076124,
+      "loss": 1.1362,
+      "step": 5969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007612,
+      "loss": 1.2365,
+      "step": 5970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076116,
+      "loss": 1.1502,
+      "step": 5971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076112,
+      "loss": 1.2474,
+      "step": 5972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076108,
+      "loss": 1.1838,
+      "step": 5973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076104,
+      "loss": 1.2093,
+      "step": 5974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007610000000000001,
+      "loss": 1.2511,
+      "step": 5975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076096,
+      "loss": 1.2856,
+      "step": 5976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007609200000000001,
+      "loss": 1.2812,
+      "step": 5977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076088,
+      "loss": 1.1758,
+      "step": 5978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007608399999999999,
+      "loss": 1.1998,
+      "step": 5979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007608000000000001,
+      "loss": 1.2538,
+      "step": 5980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076076,
+      "loss": 1.2239,
+      "step": 5981
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076072,
+      "loss": 1.2026,
+      "step": 5982
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076068,
+      "loss": 1.1226,
+      "step": 5983
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076064,
+      "loss": 1.2986,
+      "step": 5984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007606000000000001,
+      "loss": 1.2562,
+      "step": 5985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007605600000000001,
+      "loss": 1.2275,
+      "step": 5986
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076052,
+      "loss": 1.3115,
+      "step": 5987
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076048,
+      "loss": 1.184,
+      "step": 5988
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076044,
+      "loss": 1.1679,
+      "step": 5989
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007603999999999999,
+      "loss": 1.1363,
+      "step": 5990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007603600000000001,
+      "loss": 1.2482,
+      "step": 5991
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076032,
+      "loss": 1.194,
+      "step": 5992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076028,
+      "loss": 1.2527,
+      "step": 5993
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076024,
+      "loss": 1.1401,
+      "step": 5994
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007602,
+      "loss": 1.1845,
+      "step": 5995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076016,
+      "loss": 1.1817,
+      "step": 5996
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076012,
+      "loss": 1.1698,
+      "step": 5997
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076008,
+      "loss": 1.153,
+      "step": 5998
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076004,
+      "loss": 1.2814,
+      "step": 5999
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00076,
+      "loss": 1.1685,
+      "step": 6000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075996,
+      "loss": 1.2076,
+      "step": 6001
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007599200000000001,
+      "loss": 1.2092,
+      "step": 6002
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075988,
+      "loss": 1.1408,
+      "step": 6003
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075984,
+      "loss": 1.1462,
+      "step": 6004
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007598,
+      "loss": 1.2214,
+      "step": 6005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007597600000000001,
+      "loss": 1.2494,
+      "step": 6006
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075972,
+      "loss": 1.1833,
+      "step": 6007
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075968,
+      "loss": 1.1541,
+      "step": 6008
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075964,
+      "loss": 1.1427,
+      "step": 6009
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007596,
+      "loss": 1.1908,
+      "step": 6010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075956,
+      "loss": 1.2025,
+      "step": 6011
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075952,
+      "loss": 1.1382,
+      "step": 6012
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007594800000000001,
+      "loss": 1.1771,
+      "step": 6013
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075944,
+      "loss": 1.2673,
+      "step": 6014
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007594,
+      "loss": 1.2236,
+      "step": 6015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075936,
+      "loss": 1.09,
+      "step": 6016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075932,
+      "loss": 1.2271,
+      "step": 6017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075928,
+      "loss": 1.2163,
+      "step": 6018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075924,
+      "loss": 1.2163,
+      "step": 6019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007592,
+      "loss": 1.22,
+      "step": 6020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007591599999999999,
+      "loss": 1.1737,
+      "step": 6021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007591200000000001,
+      "loss": 1.1994,
+      "step": 6022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075908,
+      "loss": 1.2908,
+      "step": 6023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007590400000000001,
+      "loss": 1.2387,
+      "step": 6024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000759,
+      "loss": 1.1262,
+      "step": 6025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007589599999999999,
+      "loss": 1.2665,
+      "step": 6026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007589200000000001,
+      "loss": 1.2515,
+      "step": 6027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075888,
+      "loss": 1.1871,
+      "step": 6028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075884,
+      "loss": 1.2474,
+      "step": 6029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007588,
+      "loss": 1.2013,
+      "step": 6030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075876,
+      "loss": 1.1634,
+      "step": 6031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007587199999999999,
+      "loss": 1.2912,
+      "step": 6032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007586800000000001,
+      "loss": 1.0933,
+      "step": 6033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075864,
+      "loss": 1.1772,
+      "step": 6034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007586000000000001,
+      "loss": 1.222,
+      "step": 6035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075856,
+      "loss": 1.2152,
+      "step": 6036
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007585199999999999,
+      "loss": 1.1851,
+      "step": 6037
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007584800000000001,
+      "loss": 1.1656,
+      "step": 6038
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075844,
+      "loss": 1.2361,
+      "step": 6039
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007584,
+      "loss": 1.1949,
+      "step": 6040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075836,
+      "loss": 1.2064,
+      "step": 6041
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075832,
+      "loss": 1.1841,
+      "step": 6042
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075828,
+      "loss": 1.2387,
+      "step": 6043
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007582400000000001,
+      "loss": 1.2569,
+      "step": 6044
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007582,
+      "loss": 1.1657,
+      "step": 6045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007581599999999999,
+      "loss": 1.1646,
+      "step": 6046
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075812,
+      "loss": 1.1729,
+      "step": 6047
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075808,
+      "loss": 1.1145,
+      "step": 6048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007580400000000001,
+      "loss": 1.1654,
+      "step": 6049
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000758,
+      "loss": 1.2656,
+      "step": 6050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075796,
+      "loss": 1.2224,
+      "step": 6051
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075792,
+      "loss": 1.1766,
+      "step": 6052
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007578800000000001,
+      "loss": 1.215,
+      "step": 6053
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075784,
+      "loss": 1.1453,
+      "step": 6054
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007578000000000001,
+      "loss": 1.267,
+      "step": 6055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075776,
+      "loss": 1.2516,
+      "step": 6056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007577199999999999,
+      "loss": 1.2229,
+      "step": 6057
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075768,
+      "loss": 1.1371,
+      "step": 6058
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075764,
+      "loss": 1.2557,
+      "step": 6059
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007576000000000001,
+      "loss": 1.1761,
+      "step": 6060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075756,
+      "loss": 1.2095,
+      "step": 6061
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075752,
+      "loss": 1.146,
+      "step": 6062
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075748,
+      "loss": 1.224,
+      "step": 6063
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007574400000000001,
+      "loss": 1.2413,
+      "step": 6064
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007574,
+      "loss": 1.1401,
+      "step": 6065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075736,
+      "loss": 1.1868,
+      "step": 6066
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075732,
+      "loss": 1.1207,
+      "step": 6067
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007572799999999999,
+      "loss": 1.2023,
+      "step": 6068
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007572400000000001,
+      "loss": 1.2425,
+      "step": 6069
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007572,
+      "loss": 1.2057,
+      "step": 6070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007571600000000001,
+      "loss": 1.2126,
+      "step": 6071
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075712,
+      "loss": 1.139,
+      "step": 6072
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075708,
+      "loss": 1.0869,
+      "step": 6073
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007570400000000001,
+      "loss": 1.2756,
+      "step": 6074
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000757,
+      "loss": 1.2368,
+      "step": 6075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075696,
+      "loss": 1.1923,
+      "step": 6076
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075692,
+      "loss": 1.2479,
+      "step": 6077
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075688,
+      "loss": 1.2827,
+      "step": 6078
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007568399999999999,
+      "loss": 1.1905,
+      "step": 6079
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007568000000000001,
+      "loss": 1.1277,
+      "step": 6080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075676,
+      "loss": 1.2186,
+      "step": 6081
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075672,
+      "loss": 1.0974,
+      "step": 6082
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075668,
+      "loss": 1.2182,
+      "step": 6083
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075664,
+      "loss": 1.1369,
+      "step": 6084
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007566000000000001,
+      "loss": 1.2454,
+      "step": 6085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075656,
+      "loss": 1.1357,
+      "step": 6086
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075652,
+      "loss": 1.1331,
+      "step": 6087
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075648,
+      "loss": 1.1803,
+      "step": 6088
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075644,
+      "loss": 1.1149,
+      "step": 6089
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007564,
+      "loss": 1.1605,
+      "step": 6090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007563600000000001,
+      "loss": 1.1666,
+      "step": 6091
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075632,
+      "loss": 1.2563,
+      "step": 6092
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075628,
+      "loss": 1.1103,
+      "step": 6093
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075624,
+      "loss": 1.2689,
+      "step": 6094
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007562,
+      "loss": 1.2154,
+      "step": 6095
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007561600000000001,
+      "loss": 1.1631,
+      "step": 6096
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075612,
+      "loss": 1.2271,
+      "step": 6097
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075608,
+      "loss": 1.2058,
+      "step": 6098
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075604,
+      "loss": 1.3657,
+      "step": 6099
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000756,
+      "loss": 1.1941,
+      "step": 6100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075596,
+      "loss": 1.2035,
+      "step": 6101
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007559200000000001,
+      "loss": 1.2152,
+      "step": 6102
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075588,
+      "loss": 1.146,
+      "step": 6103
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007558399999999999,
+      "loss": 1.3145,
+      "step": 6104
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007558,
+      "loss": 1.2917,
+      "step": 6105
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075576,
+      "loss": 1.1886,
+      "step": 6106
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075572,
+      "loss": 1.257,
+      "step": 6107
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075568,
+      "loss": 1.2258,
+      "step": 6108
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075564,
+      "loss": 1.1409,
+      "step": 6109
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007556,
+      "loss": 1.125,
+      "step": 6110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007555600000000001,
+      "loss": 1.1672,
+      "step": 6111
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075552,
+      "loss": 1.2014,
+      "step": 6112
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007554800000000001,
+      "loss": 1.2132,
+      "step": 6113
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075544,
+      "loss": 1.2295,
+      "step": 6114
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007553999999999999,
+      "loss": 1.2249,
+      "step": 6115
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007553600000000001,
+      "loss": 1.1175,
+      "step": 6116
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075532,
+      "loss": 1.1444,
+      "step": 6117
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075528,
+      "loss": 1.1818,
+      "step": 6118
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075524,
+      "loss": 1.2008,
+      "step": 6119
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007552,
+      "loss": 1.2668,
+      "step": 6120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007551600000000001,
+      "loss": 1.2095,
+      "step": 6121
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007551200000000001,
+      "loss": 1.1917,
+      "step": 6122
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075508,
+      "loss": 1.1439,
+      "step": 6123
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075504,
+      "loss": 1.2084,
+      "step": 6124
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000755,
+      "loss": 1.2535,
+      "step": 6125
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007549599999999999,
+      "loss": 1.1945,
+      "step": 6126
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007549200000000001,
+      "loss": 1.2316,
+      "step": 6127
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075488,
+      "loss": 1.2292,
+      "step": 6128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075484,
+      "loss": 1.2156,
+      "step": 6129
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007548,
+      "loss": 1.2472,
+      "step": 6130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075476,
+      "loss": 1.2446,
+      "step": 6131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075472,
+      "loss": 1.2494,
+      "step": 6132
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007546800000000001,
+      "loss": 1.1783,
+      "step": 6133
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075464,
+      "loss": 1.1019,
+      "step": 6134
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007546,
+      "loss": 1.1763,
+      "step": 6135
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075456,
+      "loss": 1.2584,
+      "step": 6136
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075452,
+      "loss": 1.1753,
+      "step": 6137
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007544800000000001,
+      "loss": 1.1041,
+      "step": 6138
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075444,
+      "loss": 1.2507,
+      "step": 6139
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007544,
+      "loss": 1.2451,
+      "step": 6140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075436,
+      "loss": 1.1449,
+      "step": 6141
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007543200000000001,
+      "loss": 1.173,
+      "step": 6142
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075428,
+      "loss": 1.1987,
+      "step": 6143
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075424,
+      "loss": 1.1062,
+      "step": 6144
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007542,
+      "loss": 1.194,
+      "step": 6145
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075416,
+      "loss": 1.2961,
+      "step": 6146
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075412,
+      "loss": 1.244,
+      "step": 6147
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075408,
+      "loss": 1.2265,
+      "step": 6148
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007540400000000001,
+      "loss": 1.2294,
+      "step": 6149
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000754,
+      "loss": 1.2321,
+      "step": 6150
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075396,
+      "loss": 1.2661,
+      "step": 6151
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075392,
+      "loss": 1.1635,
+      "step": 6152
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075388,
+      "loss": 1.1605,
+      "step": 6153
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075384,
+      "loss": 1.1982,
+      "step": 6154
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007538,
+      "loss": 1.1637,
+      "step": 6155
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075376,
+      "loss": 1.2026,
+      "step": 6156
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007537199999999999,
+      "loss": 1.2669,
+      "step": 6157
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007536800000000001,
+      "loss": 1.1263,
+      "step": 6158
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075364,
+      "loss": 1.123,
+      "step": 6159
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007536000000000001,
+      "loss": 1.1758,
+      "step": 6160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075356,
+      "loss": 1.2008,
+      "step": 6161
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075352,
+      "loss": 1.2419,
+      "step": 6162
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007534800000000001,
+      "loss": 1.1175,
+      "step": 6163
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075344,
+      "loss": 1.1614,
+      "step": 6164
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007534,
+      "loss": 1.1701,
+      "step": 6165
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075336,
+      "loss": 1.2466,
+      "step": 6166
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075332,
+      "loss": 1.2,
+      "step": 6167
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007532799999999999,
+      "loss": 1.1304,
+      "step": 6168
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007532400000000001,
+      "loss": 1.1955,
+      "step": 6169
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007532,
+      "loss": 1.1912,
+      "step": 6170
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007531600000000001,
+      "loss": 1.2323,
+      "step": 6171
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075312,
+      "loss": 1.2231,
+      "step": 6172
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007530799999999999,
+      "loss": 1.2357,
+      "step": 6173
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007530400000000001,
+      "loss": 1.1662,
+      "step": 6174
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000753,
+      "loss": 1.1221,
+      "step": 6175
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075296,
+      "loss": 1.1631,
+      "step": 6176
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075292,
+      "loss": 1.1765,
+      "step": 6177
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075288,
+      "loss": 1.1496,
+      "step": 6178
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075284,
+      "loss": 1.2002,
+      "step": 6179
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007528000000000001,
+      "loss": 1.2085,
+      "step": 6180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075276,
+      "loss": 1.2164,
+      "step": 6181
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007527199999999999,
+      "loss": 1.2068,
+      "step": 6182
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075268,
+      "loss": 1.1846,
+      "step": 6183
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075264,
+      "loss": 1.1703,
+      "step": 6184
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007526000000000001,
+      "loss": 1.2216,
+      "step": 6185
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075256,
+      "loss": 1.2421,
+      "step": 6186
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075252,
+      "loss": 1.2138,
+      "step": 6187
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075248,
+      "loss": 1.2429,
+      "step": 6188
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075244,
+      "loss": 1.1543,
+      "step": 6189
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007524,
+      "loss": 1.2445,
+      "step": 6190
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007523600000000001,
+      "loss": 1.227,
+      "step": 6191
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075232,
+      "loss": 1.2724,
+      "step": 6192
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007522799999999999,
+      "loss": 1.1628,
+      "step": 6193
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075224,
+      "loss": 1.2105,
+      "step": 6194
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007522,
+      "loss": 1.1258,
+      "step": 6195
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007521600000000001,
+      "loss": 1.248,
+      "step": 6196
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075212,
+      "loss": 1.2101,
+      "step": 6197
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075208,
+      "loss": 1.18,
+      "step": 6198
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075204,
+      "loss": 1.1766,
+      "step": 6199
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007520000000000001,
+      "loss": 1.2577,
+      "step": 6200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075196,
+      "loss": 1.2032,
+      "step": 6201
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075192,
+      "loss": 1.2895,
+      "step": 6202
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075188,
+      "loss": 1.1542,
+      "step": 6203
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007518399999999999,
+      "loss": 1.1671,
+      "step": 6204
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007518000000000001,
+      "loss": 1.2427,
+      "step": 6205
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075176,
+      "loss": 1.2041,
+      "step": 6206
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0007517200000000001,
+      "loss": 1.3174,
+      "step": 6207
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00075168,
+      "loss": 1.1165,
+      "step": 6208
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075164,
+      "loss": 1.209,
+      "step": 6209
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007516000000000001,
+      "loss": 1.1849,
+      "step": 6210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007515600000000001,
+      "loss": 1.0832,
+      "step": 6211
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075152,
+      "loss": 1.1346,
+      "step": 6212
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075148,
+      "loss": 1.2349,
+      "step": 6213
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075144,
+      "loss": 1.139,
+      "step": 6214
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007513999999999999,
+      "loss": 1.3034,
+      "step": 6215
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007513600000000001,
+      "loss": 1.1879,
+      "step": 6216
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075132,
+      "loss": 1.1455,
+      "step": 6217
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075128,
+      "loss": 1.1711,
+      "step": 6218
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075124,
+      "loss": 1.1633,
+      "step": 6219
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007512,
+      "loss": 1.1803,
+      "step": 6220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007511600000000001,
+      "loss": 1.287,
+      "step": 6221
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075112,
+      "loss": 1.2158,
+      "step": 6222
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075108,
+      "loss": 1.2495,
+      "step": 6223
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075104,
+      "loss": 1.2268,
+      "step": 6224
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000751,
+      "loss": 1.2203,
+      "step": 6225
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075096,
+      "loss": 1.2052,
+      "step": 6226
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007509200000000001,
+      "loss": 1.1603,
+      "step": 6227
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075088,
+      "loss": 1.1986,
+      "step": 6228
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075084,
+      "loss": 1.1591,
+      "step": 6229
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007508,
+      "loss": 1.1928,
+      "step": 6230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075076,
+      "loss": 1.2086,
+      "step": 6231
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007507200000000001,
+      "loss": 1.2425,
+      "step": 6232
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075068,
+      "loss": 1.1408,
+      "step": 6233
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075064,
+      "loss": 1.1725,
+      "step": 6234
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007506,
+      "loss": 1.16,
+      "step": 6235
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075056,
+      "loss": 1.3084,
+      "step": 6236
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075052,
+      "loss": 1.1919,
+      "step": 6237
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007504800000000001,
+      "loss": 1.1993,
+      "step": 6238
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075044,
+      "loss": 1.1451,
+      "step": 6239
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007504,
+      "loss": 1.1279,
+      "step": 6240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075036,
+      "loss": 1.1607,
+      "step": 6241
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075032,
+      "loss": 1.2898,
+      "step": 6242
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075028,
+      "loss": 1.2139,
+      "step": 6243
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075024,
+      "loss": 1.2926,
+      "step": 6244
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007502,
+      "loss": 1.2037,
+      "step": 6245
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075016,
+      "loss": 1.2281,
+      "step": 6246
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007501200000000001,
+      "loss": 1.2445,
+      "step": 6247
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075008,
+      "loss": 1.2326,
+      "step": 6248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007500400000000001,
+      "loss": 1.2134,
+      "step": 6249
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00075,
+      "loss": 1.2399,
+      "step": 6250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007499599999999999,
+      "loss": 1.1892,
+      "step": 6251
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007499200000000001,
+      "loss": 1.27,
+      "step": 6252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074988,
+      "loss": 1.1108,
+      "step": 6253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074984,
+      "loss": 1.2154,
+      "step": 6254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007498,
+      "loss": 1.2092,
+      "step": 6255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074976,
+      "loss": 1.1637,
+      "step": 6256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074972,
+      "loss": 1.1461,
+      "step": 6257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007496800000000001,
+      "loss": 1.1897,
+      "step": 6258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074964,
+      "loss": 1.3218,
+      "step": 6259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007496,
+      "loss": 1.2495,
+      "step": 6260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074956,
+      "loss": 1.2643,
+      "step": 6261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007495199999999999,
+      "loss": 1.2055,
+      "step": 6262
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007494800000000001,
+      "loss": 1.2608,
+      "step": 6263
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074944,
+      "loss": 1.2134,
+      "step": 6264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007494,
+      "loss": 1.3034,
+      "step": 6265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074936,
+      "loss": 1.1235,
+      "step": 6266
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074932,
+      "loss": 1.2549,
+      "step": 6267
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074928,
+      "loss": 1.2275,
+      "step": 6268
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007492400000000001,
+      "loss": 1.1472,
+      "step": 6269
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007492,
+      "loss": 1.209,
+      "step": 6270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074916,
+      "loss": 1.2685,
+      "step": 6271
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074912,
+      "loss": 1.2708,
+      "step": 6272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074908,
+      "loss": 1.1827,
+      "step": 6273
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007490400000000001,
+      "loss": 1.1433,
+      "step": 6274
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000749,
+      "loss": 1.2884,
+      "step": 6275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074896,
+      "loss": 1.1951,
+      "step": 6276
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074892,
+      "loss": 1.1592,
+      "step": 6277
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074888,
+      "loss": 1.1406,
+      "step": 6278
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074884,
+      "loss": 1.243,
+      "step": 6279
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007488,
+      "loss": 1.1203,
+      "step": 6280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074876,
+      "loss": 1.1348,
+      "step": 6281
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074872,
+      "loss": 1.1978,
+      "step": 6282
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074868,
+      "loss": 1.075,
+      "step": 6283
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074864,
+      "loss": 1.164,
+      "step": 6284
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007486000000000001,
+      "loss": 1.2254,
+      "step": 6285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074856,
+      "loss": 1.1593,
+      "step": 6286
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074852,
+      "loss": 1.3214,
+      "step": 6287
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074848,
+      "loss": 1.1885,
+      "step": 6288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074844,
+      "loss": 1.2296,
+      "step": 6289
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007484,
+      "loss": 1.1604,
+      "step": 6290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074836,
+      "loss": 1.1536,
+      "step": 6291
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074832,
+      "loss": 1.2675,
+      "step": 6292
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007482799999999999,
+      "loss": 1.219,
+      "step": 6293
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007482400000000001,
+      "loss": 1.2089,
+      "step": 6294
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007482,
+      "loss": 1.1603,
+      "step": 6295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007481600000000001,
+      "loss": 1.2889,
+      "step": 6296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074812,
+      "loss": 1.2263,
+      "step": 6297
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074808,
+      "loss": 1.1112,
+      "step": 6298
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007480400000000001,
+      "loss": 1.2179,
+      "step": 6299
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000748,
+      "loss": 1.1043,
+      "step": 6300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074796,
+      "loss": 1.1211,
+      "step": 6301
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074792,
+      "loss": 1.1855,
+      "step": 6302
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074788,
+      "loss": 1.1574,
+      "step": 6303
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007478399999999999,
+      "loss": 1.1846,
+      "step": 6304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007478000000000001,
+      "loss": 1.2037,
+      "step": 6305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074776,
+      "loss": 1.1981,
+      "step": 6306
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007477200000000001,
+      "loss": 1.1421,
+      "step": 6307
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074768,
+      "loss": 1.112,
+      "step": 6308
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007476399999999999,
+      "loss": 1.2351,
+      "step": 6309
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007476000000000001,
+      "loss": 1.22,
+      "step": 6310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074756,
+      "loss": 1.2003,
+      "step": 6311
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074752,
+      "loss": 1.1666,
+      "step": 6312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074748,
+      "loss": 1.2076,
+      "step": 6313
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074744,
+      "loss": 1.2329,
+      "step": 6314
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007474,
+      "loss": 1.2017,
+      "step": 6315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007473600000000001,
+      "loss": 1.1959,
+      "step": 6316
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074732,
+      "loss": 1.1746,
+      "step": 6317
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007472800000000001,
+      "loss": 1.2611,
+      "step": 6318
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074724,
+      "loss": 1.1167,
+      "step": 6319
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007472,
+      "loss": 1.2108,
+      "step": 6320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007471600000000001,
+      "loss": 1.1859,
+      "step": 6321
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074712,
+      "loss": 1.1825,
+      "step": 6322
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074708,
+      "loss": 1.2574,
+      "step": 6323
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074704,
+      "loss": 1.2259,
+      "step": 6324
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000747,
+      "loss": 1.2544,
+      "step": 6325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074696,
+      "loss": 1.1624,
+      "step": 6326
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007469200000000001,
+      "loss": 1.2076,
+      "step": 6327
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074688,
+      "loss": 1.1775,
+      "step": 6328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007468399999999999,
+      "loss": 1.2497,
+      "step": 6329
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007468,
+      "loss": 1.1668,
+      "step": 6330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074676,
+      "loss": 1.1522,
+      "step": 6331
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007467200000000001,
+      "loss": 1.148,
+      "step": 6332
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074668,
+      "loss": 1.1376,
+      "step": 6333
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074664,
+      "loss": 1.177,
+      "step": 6334
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007466,
+      "loss": 1.222,
+      "step": 6335
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007465600000000001,
+      "loss": 1.1721,
+      "step": 6336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074652,
+      "loss": 1.1871,
+      "step": 6337
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074648,
+      "loss": 1.2319,
+      "step": 6338
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074644,
+      "loss": 1.3216,
+      "step": 6339
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007463999999999999,
+      "loss": 1.2095,
+      "step": 6340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007463600000000001,
+      "loss": 1.2019,
+      "step": 6341
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074632,
+      "loss": 1.1501,
+      "step": 6342
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007462800000000001,
+      "loss": 1.3146,
+      "step": 6343
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074624,
+      "loss": 1.1966,
+      "step": 6344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007462,
+      "loss": 1.3361,
+      "step": 6345
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074616,
+      "loss": 1.136,
+      "step": 6346
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007461200000000001,
+      "loss": 1.1336,
+      "step": 6347
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074608,
+      "loss": 1.1493,
+      "step": 6348
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074604,
+      "loss": 1.2298,
+      "step": 6349
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000746,
+      "loss": 1.2409,
+      "step": 6350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007459599999999999,
+      "loss": 1.2077,
+      "step": 6351
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007459200000000001,
+      "loss": 1.1767,
+      "step": 6352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074588,
+      "loss": 1.228,
+      "step": 6353
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074584,
+      "loss": 1.1638,
+      "step": 6354
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007458,
+      "loss": 1.1582,
+      "step": 6355
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074576,
+      "loss": 1.1859,
+      "step": 6356
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007457200000000001,
+      "loss": 1.2265,
+      "step": 6357
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074568,
+      "loss": 1.1382,
+      "step": 6358
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074564,
+      "loss": 1.136,
+      "step": 6359
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007456,
+      "loss": 1.2754,
+      "step": 6360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074556,
+      "loss": 1.1213,
+      "step": 6361
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074552,
+      "loss": 1.2409,
+      "step": 6362
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007454800000000001,
+      "loss": 1.2433,
+      "step": 6363
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074544,
+      "loss": 1.1374,
+      "step": 6364
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007454,
+      "loss": 1.1855,
+      "step": 6365
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074536,
+      "loss": 1.1924,
+      "step": 6366
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007453199999999999,
+      "loss": 1.223,
+      "step": 6367
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007452800000000001,
+      "loss": 1.1775,
+      "step": 6368
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074524,
+      "loss": 1.1274,
+      "step": 6369
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007452,
+      "loss": 1.1742,
+      "step": 6370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074516,
+      "loss": 1.1574,
+      "step": 6371
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074512,
+      "loss": 1.2204,
+      "step": 6372
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074508,
+      "loss": 1.2129,
+      "step": 6373
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007450400000000001,
+      "loss": 1.2397,
+      "step": 6374
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000745,
+      "loss": 1.1277,
+      "step": 6375
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074496,
+      "loss": 1.1819,
+      "step": 6376
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074492,
+      "loss": 1.1162,
+      "step": 6377
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074488,
+      "loss": 1.1905,
+      "step": 6378
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074484,
+      "loss": 1.1218,
+      "step": 6379
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007448,
+      "loss": 1.2234,
+      "step": 6380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074476,
+      "loss": 1.1743,
+      "step": 6381
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074472,
+      "loss": 1.1433,
+      "step": 6382
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007446800000000001,
+      "loss": 1.2025,
+      "step": 6383
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074464,
+      "loss": 1.1377,
+      "step": 6384
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007446000000000001,
+      "loss": 1.2226,
+      "step": 6385
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074456,
+      "loss": 1.1394,
+      "step": 6386
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007445199999999999,
+      "loss": 1.1652,
+      "step": 6387
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007444800000000001,
+      "loss": 1.2805,
+      "step": 6388
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074444,
+      "loss": 1.1817,
+      "step": 6389
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007444,
+      "loss": 1.2251,
+      "step": 6390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074436,
+      "loss": 1.1713,
+      "step": 6391
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074432,
+      "loss": 1.1739,
+      "step": 6392
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074428,
+      "loss": 1.2195,
+      "step": 6393
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007442400000000001,
+      "loss": 1.1223,
+      "step": 6394
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007442,
+      "loss": 1.2393,
+      "step": 6395
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007441600000000001,
+      "loss": 1.1928,
+      "step": 6396
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074412,
+      "loss": 1.2474,
+      "step": 6397
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007440799999999999,
+      "loss": 1.2586,
+      "step": 6398
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007440400000000001,
+      "loss": 1.1836,
+      "step": 6399
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000744,
+      "loss": 1.2001,
+      "step": 6400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074396,
+      "loss": 1.2471,
+      "step": 6401
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074392,
+      "loss": 1.1981,
+      "step": 6402
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074388,
+      "loss": 1.19,
+      "step": 6403
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074384,
+      "loss": 1.2619,
+      "step": 6404
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007438000000000001,
+      "loss": 1.2515,
+      "step": 6405
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074376,
+      "loss": 1.1772,
+      "step": 6406
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074372,
+      "loss": 1.1396,
+      "step": 6407
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074368,
+      "loss": 1.2442,
+      "step": 6408
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074364,
+      "loss": 1.2372,
+      "step": 6409
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007436000000000001,
+      "loss": 1.1655,
+      "step": 6410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074356,
+      "loss": 1.1979,
+      "step": 6411
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074352,
+      "loss": 1.2467,
+      "step": 6412
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074348,
+      "loss": 1.2439,
+      "step": 6413
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074344,
+      "loss": 1.1531,
+      "step": 6414
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007434,
+      "loss": 1.2316,
+      "step": 6415
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074336,
+      "loss": 1.0961,
+      "step": 6416
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074332,
+      "loss": 1.2437,
+      "step": 6417
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074328,
+      "loss": 1.2402,
+      "step": 6418
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074324,
+      "loss": 1.2369,
+      "step": 6419
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007432,
+      "loss": 1.1607,
+      "step": 6420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007431600000000001,
+      "loss": 1.1297,
+      "step": 6421
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074312,
+      "loss": 1.1423,
+      "step": 6422
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074308,
+      "loss": 1.1089,
+      "step": 6423
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074304,
+      "loss": 1.1996,
+      "step": 6424
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007430000000000001,
+      "loss": 1.0785,
+      "step": 6425
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074296,
+      "loss": 1.1676,
+      "step": 6426
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074292,
+      "loss": 1.1645,
+      "step": 6427
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074288,
+      "loss": 1.0798,
+      "step": 6428
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007428399999999999,
+      "loss": 1.1074,
+      "step": 6429
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007428000000000001,
+      "loss": 1.2425,
+      "step": 6430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074276,
+      "loss": 1.1346,
+      "step": 6431
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007427200000000001,
+      "loss": 1.1743,
+      "step": 6432
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074268,
+      "loss": 1.1239,
+      "step": 6433
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074264,
+      "loss": 1.1648,
+      "step": 6434
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007426,
+      "loss": 1.1962,
+      "step": 6435
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074256,
+      "loss": 1.1475,
+      "step": 6436
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074252,
+      "loss": 1.1634,
+      "step": 6437
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074248,
+      "loss": 1.0961,
+      "step": 6438
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074244,
+      "loss": 1.2201,
+      "step": 6439
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007423999999999999,
+      "loss": 1.1757,
+      "step": 6440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007423600000000001,
+      "loss": 1.1315,
+      "step": 6441
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074232,
+      "loss": 1.204,
+      "step": 6442
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007422800000000001,
+      "loss": 1.2169,
+      "step": 6443
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074224,
+      "loss": 1.2632,
+      "step": 6444
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007421999999999999,
+      "loss": 1.1694,
+      "step": 6445
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007421600000000001,
+      "loss": 1.0587,
+      "step": 6446
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074212,
+      "loss": 1.1932,
+      "step": 6447
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074208,
+      "loss": 1.1512,
+      "step": 6448
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074204,
+      "loss": 1.2888,
+      "step": 6449
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000742,
+      "loss": 1.2216,
+      "step": 6450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074196,
+      "loss": 1.0688,
+      "step": 6451
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007419200000000001,
+      "loss": 1.1976,
+      "step": 6452
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074188,
+      "loss": 1.1866,
+      "step": 6453
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007418400000000001,
+      "loss": 1.2198,
+      "step": 6454
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007418,
+      "loss": 1.2172,
+      "step": 6455
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007417599999999999,
+      "loss": 1.1144,
+      "step": 6456
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007417200000000001,
+      "loss": 1.2028,
+      "step": 6457
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074168,
+      "loss": 1.1474,
+      "step": 6458
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074164,
+      "loss": 1.1873,
+      "step": 6459
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0007416,
+      "loss": 1.2728,
+      "step": 6460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00074156,
+      "loss": 1.1823,
+      "step": 6461
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074152,
+      "loss": 1.1409,
+      "step": 6462
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007414800000000001,
+      "loss": 1.1508,
+      "step": 6463
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074144,
+      "loss": 1.1776,
+      "step": 6464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007413999999999999,
+      "loss": 1.1653,
+      "step": 6465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074136,
+      "loss": 1.1262,
+      "step": 6466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074132,
+      "loss": 1.1985,
+      "step": 6467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007412800000000001,
+      "loss": 1.1662,
+      "step": 6468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074124,
+      "loss": 1.1423,
+      "step": 6469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007412,
+      "loss": 1.1879,
+      "step": 6470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074116,
+      "loss": 1.1818,
+      "step": 6471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007411200000000001,
+      "loss": 1.1991,
+      "step": 6472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074108,
+      "loss": 1.2074,
+      "step": 6473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007410400000000001,
+      "loss": 1.1146,
+      "step": 6474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000741,
+      "loss": 1.2155,
+      "step": 6475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007409599999999999,
+      "loss": 1.1524,
+      "step": 6476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007409200000000001,
+      "loss": 1.1896,
+      "step": 6477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074088,
+      "loss": 1.0594,
+      "step": 6478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007408400000000001,
+      "loss": 1.2007,
+      "step": 6479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007408,
+      "loss": 1.1865,
+      "step": 6480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074076,
+      "loss": 1.2059,
+      "step": 6481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074072,
+      "loss": 1.1289,
+      "step": 6482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007406800000000001,
+      "loss": 1.1926,
+      "step": 6483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074064,
+      "loss": 1.184,
+      "step": 6484
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007406,
+      "loss": 1.1318,
+      "step": 6485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074056,
+      "loss": 1.1345,
+      "step": 6486
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007405199999999999,
+      "loss": 1.134,
+      "step": 6487
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007404800000000001,
+      "loss": 1.141,
+      "step": 6488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074044,
+      "loss": 1.2663,
+      "step": 6489
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007404,
+      "loss": 1.2078,
+      "step": 6490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074036,
+      "loss": 1.1861,
+      "step": 6491
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074032,
+      "loss": 1.1112,
+      "step": 6492
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007402800000000001,
+      "loss": 1.1763,
+      "step": 6493
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074024,
+      "loss": 1.3032,
+      "step": 6494
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007402,
+      "loss": 1.1171,
+      "step": 6495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074016,
+      "loss": 1.0718,
+      "step": 6496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074012,
+      "loss": 1.2749,
+      "step": 6497
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074008,
+      "loss": 1.1807,
+      "step": 6498
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007400400000000001,
+      "loss": 1.2412,
+      "step": 6499
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00074,
+      "loss": 1.2733,
+      "step": 6500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073996,
+      "loss": 1.0651,
+      "step": 6501
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073992,
+      "loss": 1.159,
+      "step": 6502
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073988,
+      "loss": 1.1628,
+      "step": 6503
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007398400000000001,
+      "loss": 1.2184,
+      "step": 6504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007398,
+      "loss": 1.26,
+      "step": 6505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073976,
+      "loss": 1.219,
+      "step": 6506
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073972,
+      "loss": 1.1693,
+      "step": 6507
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073968,
+      "loss": 1.1504,
+      "step": 6508
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073964,
+      "loss": 1.1069,
+      "step": 6509
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007396000000000001,
+      "loss": 1.2046,
+      "step": 6510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073956,
+      "loss": 1.193,
+      "step": 6511
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073952,
+      "loss": 1.1436,
+      "step": 6512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073948,
+      "loss": 1.1615,
+      "step": 6513
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073944,
+      "loss": 1.2949,
+      "step": 6514
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007394,
+      "loss": 1.2881,
+      "step": 6515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073936,
+      "loss": 1.1422,
+      "step": 6516
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073932,
+      "loss": 1.1658,
+      "step": 6517
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073928,
+      "loss": 1.2595,
+      "step": 6518
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007392400000000001,
+      "loss": 1.151,
+      "step": 6519
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007392,
+      "loss": 1.1575,
+      "step": 6520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007391600000000001,
+      "loss": 1.0908,
+      "step": 6521
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073912,
+      "loss": 1.2018,
+      "step": 6522
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007390799999999999,
+      "loss": 1.1406,
+      "step": 6523
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073904,
+      "loss": 1.1856,
+      "step": 6524
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000739,
+      "loss": 1.1328,
+      "step": 6525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073896,
+      "loss": 1.2593,
+      "step": 6526
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073892,
+      "loss": 1.2445,
+      "step": 6527
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073888,
+      "loss": 1.1409,
+      "step": 6528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073884,
+      "loss": 1.1513,
+      "step": 6529
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007388000000000001,
+      "loss": 1.1559,
+      "step": 6530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073876,
+      "loss": 1.161,
+      "step": 6531
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007387200000000001,
+      "loss": 1.1674,
+      "step": 6532
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073868,
+      "loss": 1.1557,
+      "step": 6533
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007386399999999999,
+      "loss": 1.1497,
+      "step": 6534
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007386000000000001,
+      "loss": 1.2386,
+      "step": 6535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073856,
+      "loss": 1.086,
+      "step": 6536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073852,
+      "loss": 1.2354,
+      "step": 6537
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073848,
+      "loss": 1.1336,
+      "step": 6538
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073844,
+      "loss": 1.138,
+      "step": 6539
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007384,
+      "loss": 1.2018,
+      "step": 6540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007383600000000001,
+      "loss": 1.0956,
+      "step": 6541
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073832,
+      "loss": 1.2036,
+      "step": 6542
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073828,
+      "loss": 1.1689,
+      "step": 6543
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073824,
+      "loss": 1.1809,
+      "step": 6544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007381999999999999,
+      "loss": 1.1556,
+      "step": 6545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007381600000000001,
+      "loss": 1.2917,
+      "step": 6546
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073812,
+      "loss": 1.1383,
+      "step": 6547
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073808,
+      "loss": 1.1629,
+      "step": 6548
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073804,
+      "loss": 1.1124,
+      "step": 6549
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000738,
+      "loss": 1.146,
+      "step": 6550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073796,
+      "loss": 1.1995,
+      "step": 6551
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073792,
+      "loss": 1.2082,
+      "step": 6552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073788,
+      "loss": 1.2149,
+      "step": 6553
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073784,
+      "loss": 1.0889,
+      "step": 6554
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007378,
+      "loss": 1.2582,
+      "step": 6555
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073776,
+      "loss": 1.1594,
+      "step": 6556
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007377200000000001,
+      "loss": 1.1381,
+      "step": 6557
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073768,
+      "loss": 1.2288,
+      "step": 6558
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073764,
+      "loss": 1.1564,
+      "step": 6559
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007376,
+      "loss": 1.1618,
+      "step": 6560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007375600000000001,
+      "loss": 1.259,
+      "step": 6561
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073752,
+      "loss": 1.2217,
+      "step": 6562
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073748,
+      "loss": 1.2459,
+      "step": 6563
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073744,
+      "loss": 1.2512,
+      "step": 6564
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007374,
+      "loss": 1.255,
+      "step": 6565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007373600000000001,
+      "loss": 1.1305,
+      "step": 6566
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073732,
+      "loss": 1.2207,
+      "step": 6567
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007372800000000001,
+      "loss": 1.1754,
+      "step": 6568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073724,
+      "loss": 1.1228,
+      "step": 6569
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007372,
+      "loss": 1.2245,
+      "step": 6570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073716,
+      "loss": 1.074,
+      "step": 6571
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073712,
+      "loss": 1.1254,
+      "step": 6572
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073708,
+      "loss": 1.1622,
+      "step": 6573
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073704,
+      "loss": 1.2519,
+      "step": 6574
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000737,
+      "loss": 1.1977,
+      "step": 6575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007369599999999999,
+      "loss": 1.2168,
+      "step": 6576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007369200000000001,
+      "loss": 1.2182,
+      "step": 6577
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073688,
+      "loss": 1.1868,
+      "step": 6578
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007368400000000001,
+      "loss": 1.1945,
+      "step": 6579
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007368,
+      "loss": 1.1817,
+      "step": 6580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007367599999999999,
+      "loss": 1.1903,
+      "step": 6581
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007367200000000001,
+      "loss": 1.2221,
+      "step": 6582
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073668,
+      "loss": 1.267,
+      "step": 6583
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073664,
+      "loss": 1.2303,
+      "step": 6584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007366,
+      "loss": 1.1232,
+      "step": 6585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073656,
+      "loss": 1.1867,
+      "step": 6586
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073652,
+      "loss": 1.1714,
+      "step": 6587
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007364800000000001,
+      "loss": 1.1597,
+      "step": 6588
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073644,
+      "loss": 1.2147,
+      "step": 6589
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007364000000000001,
+      "loss": 1.1596,
+      "step": 6590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073636,
+      "loss": 1.1708,
+      "step": 6591
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007363199999999999,
+      "loss": 1.1785,
+      "step": 6592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007362800000000001,
+      "loss": 1.2513,
+      "step": 6593
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073624,
+      "loss": 1.1136,
+      "step": 6594
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007362,
+      "loss": 1.1771,
+      "step": 6595
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073616,
+      "loss": 1.1905,
+      "step": 6596
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073612,
+      "loss": 1.2114,
+      "step": 6597
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073608,
+      "loss": 1.1942,
+      "step": 6598
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007360400000000001,
+      "loss": 1.1385,
+      "step": 6599
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000736,
+      "loss": 1.1905,
+      "step": 6600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007359599999999999,
+      "loss": 1.1347,
+      "step": 6601
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073592,
+      "loss": 1.2137,
+      "step": 6602
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073588,
+      "loss": 1.2414,
+      "step": 6603
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007358400000000001,
+      "loss": 1.2161,
+      "step": 6604
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007358,
+      "loss": 1.0792,
+      "step": 6605
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073576,
+      "loss": 1.2061,
+      "step": 6606
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073572,
+      "loss": 1.2068,
+      "step": 6607
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007356800000000001,
+      "loss": 1.0982,
+      "step": 6608
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073564,
+      "loss": 1.2625,
+      "step": 6609
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007356000000000001,
+      "loss": 1.1988,
+      "step": 6610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073556,
+      "loss": 1.22,
+      "step": 6611
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007355199999999999,
+      "loss": 1.1572,
+      "step": 6612
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073548,
+      "loss": 1.2072,
+      "step": 6613
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073544,
+      "loss": 1.1925,
+      "step": 6614
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007354000000000001,
+      "loss": 1.3218,
+      "step": 6615
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073536,
+      "loss": 1.1127,
+      "step": 6616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073532,
+      "loss": 1.1729,
+      "step": 6617
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073528,
+      "loss": 1.1959,
+      "step": 6618
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007352400000000001,
+      "loss": 1.15,
+      "step": 6619
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007352,
+      "loss": 1.2127,
+      "step": 6620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073516,
+      "loss": 1.2109,
+      "step": 6621
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073512,
+      "loss": 1.1878,
+      "step": 6622
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007350799999999999,
+      "loss": 1.2135,
+      "step": 6623
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007350400000000001,
+      "loss": 1.1451,
+      "step": 6624
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000735,
+      "loss": 1.1805,
+      "step": 6625
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073496,
+      "loss": 1.1718,
+      "step": 6626
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073492,
+      "loss": 1.1462,
+      "step": 6627
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073488,
+      "loss": 1.1732,
+      "step": 6628
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007348400000000001,
+      "loss": 1.1172,
+      "step": 6629
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007348,
+      "loss": 1.1391,
+      "step": 6630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073476,
+      "loss": 1.1619,
+      "step": 6631
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073472,
+      "loss": 1.0934,
+      "step": 6632
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073468,
+      "loss": 1.1519,
+      "step": 6633
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073464,
+      "loss": 1.0771,
+      "step": 6634
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007346000000000001,
+      "loss": 1.2191,
+      "step": 6635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073456,
+      "loss": 1.193,
+      "step": 6636
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073452,
+      "loss": 1.1347,
+      "step": 6637
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073448,
+      "loss": 1.1145,
+      "step": 6638
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073444,
+      "loss": 1.092,
+      "step": 6639
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007344000000000001,
+      "loss": 1.1994,
+      "step": 6640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073436,
+      "loss": 1.1569,
+      "step": 6641
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073432,
+      "loss": 1.0591,
+      "step": 6642
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073428,
+      "loss": 1.2373,
+      "step": 6643
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073424,
+      "loss": 1.141,
+      "step": 6644
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007342,
+      "loss": 1.0814,
+      "step": 6645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007341600000000001,
+      "loss": 1.1719,
+      "step": 6646
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073412,
+      "loss": 1.2272,
+      "step": 6647
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073408,
+      "loss": 1.1573,
+      "step": 6648
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073404,
+      "loss": 1.0958,
+      "step": 6649
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000734,
+      "loss": 1.1475,
+      "step": 6650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073396,
+      "loss": 1.2018,
+      "step": 6651
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073392,
+      "loss": 1.1967,
+      "step": 6652
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073388,
+      "loss": 1.1475,
+      "step": 6653
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073384,
+      "loss": 1.1471,
+      "step": 6654
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007338000000000001,
+      "loss": 1.1357,
+      "step": 6655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073376,
+      "loss": 1.1512,
+      "step": 6656
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007337200000000001,
+      "loss": 1.174,
+      "step": 6657
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073368,
+      "loss": 1.2527,
+      "step": 6658
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007336399999999999,
+      "loss": 1.2387,
+      "step": 6659
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007336,
+      "loss": 1.2761,
+      "step": 6660
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073356,
+      "loss": 1.1986,
+      "step": 6661
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073352,
+      "loss": 1.2526,
+      "step": 6662
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073348,
+      "loss": 1.1723,
+      "step": 6663
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073344,
+      "loss": 1.1351,
+      "step": 6664
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007334,
+      "loss": 1.2103,
+      "step": 6665
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007333600000000001,
+      "loss": 1.2065,
+      "step": 6666
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073332,
+      "loss": 1.2423,
+      "step": 6667
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007332800000000001,
+      "loss": 1.1885,
+      "step": 6668
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073324,
+      "loss": 1.1142,
+      "step": 6669
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007331999999999999,
+      "loss": 1.1923,
+      "step": 6670
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007331600000000001,
+      "loss": 1.2299,
+      "step": 6671
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073312,
+      "loss": 1.1832,
+      "step": 6672
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073308,
+      "loss": 1.1399,
+      "step": 6673
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073304,
+      "loss": 1.0958,
+      "step": 6674
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000733,
+      "loss": 1.2315,
+      "step": 6675
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073296,
+      "loss": 1.1684,
+      "step": 6676
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007329200000000001,
+      "loss": 1.2725,
+      "step": 6677
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073288,
+      "loss": 1.1252,
+      "step": 6678
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073284,
+      "loss": 1.1535,
+      "step": 6679
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007328,
+      "loss": 1.2612,
+      "step": 6680
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007327599999999999,
+      "loss": 1.1229,
+      "step": 6681
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007327200000000001,
+      "loss": 1.2046,
+      "step": 6682
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073268,
+      "loss": 1.1423,
+      "step": 6683
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073264,
+      "loss": 1.1143,
+      "step": 6684
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007326,
+      "loss": 1.0986,
+      "step": 6685
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073256,
+      "loss": 1.1621,
+      "step": 6686
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073252,
+      "loss": 1.2188,
+      "step": 6687
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007324800000000001,
+      "loss": 1.2353,
+      "step": 6688
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073244,
+      "loss": 1.1818,
+      "step": 6689
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007324,
+      "loss": 1.0834,
+      "step": 6690
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073236,
+      "loss": 1.2094,
+      "step": 6691
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073232,
+      "loss": 1.1283,
+      "step": 6692
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007322800000000001,
+      "loss": 1.1702,
+      "step": 6693
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073224,
+      "loss": 1.1475,
+      "step": 6694
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007322,
+      "loss": 1.139,
+      "step": 6695
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073216,
+      "loss": 1.092,
+      "step": 6696
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007321200000000001,
+      "loss": 1.1953,
+      "step": 6697
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073208,
+      "loss": 1.1752,
+      "step": 6698
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073204,
+      "loss": 1.1235,
+      "step": 6699
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000732,
+      "loss": 1.0955,
+      "step": 6700
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073196,
+      "loss": 1.1811,
+      "step": 6701
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073192,
+      "loss": 1.1428,
+      "step": 6702
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073188,
+      "loss": 1.2127,
+      "step": 6703
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007318400000000001,
+      "loss": 1.2042,
+      "step": 6704
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007318,
+      "loss": 1.1839,
+      "step": 6705
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073176,
+      "loss": 1.1096,
+      "step": 6706
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073172,
+      "loss": 1.1653,
+      "step": 6707
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073168,
+      "loss": 1.261,
+      "step": 6708
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073164,
+      "loss": 1.2092,
+      "step": 6709
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007316,
+      "loss": 1.2248,
+      "step": 6710
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073156,
+      "loss": 1.1976,
+      "step": 6711
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007315199999999999,
+      "loss": 1.1824,
+      "step": 6712
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007314800000000001,
+      "loss": 1.167,
+      "step": 6713
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00073144,
+      "loss": 1.2572,
+      "step": 6714
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0007314000000000001,
+      "loss": 1.1992,
+      "step": 6715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073136,
+      "loss": 1.2124,
+      "step": 6716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073132,
+      "loss": 1.2171,
+      "step": 6717
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007312800000000001,
+      "loss": 1.2421,
+      "step": 6718
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073124,
+      "loss": 1.2098,
+      "step": 6719
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007312,
+      "loss": 1.1763,
+      "step": 6720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073116,
+      "loss": 1.2447,
+      "step": 6721
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073112,
+      "loss": 1.2722,
+      "step": 6722
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073108,
+      "loss": 1.1543,
+      "step": 6723
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007310400000000001,
+      "loss": 1.1427,
+      "step": 6724
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000731,
+      "loss": 1.1181,
+      "step": 6725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007309600000000001,
+      "loss": 1.1181,
+      "step": 6726
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073092,
+      "loss": 1.1562,
+      "step": 6727
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007308799999999999,
+      "loss": 1.1963,
+      "step": 6728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007308400000000001,
+      "loss": 1.1433,
+      "step": 6729
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007308,
+      "loss": 1.1067,
+      "step": 6730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073076,
+      "loss": 1.2058,
+      "step": 6731
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073072,
+      "loss": 1.1639,
+      "step": 6732
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073068,
+      "loss": 1.206,
+      "step": 6733
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073064,
+      "loss": 1.2332,
+      "step": 6734
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007306000000000001,
+      "loss": 1.1128,
+      "step": 6735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073056,
+      "loss": 1.2251,
+      "step": 6736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007305199999999999,
+      "loss": 1.0504,
+      "step": 6737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073048,
+      "loss": 1.222,
+      "step": 6738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073044,
+      "loss": 1.2083,
+      "step": 6739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007304000000000001,
+      "loss": 1.1649,
+      "step": 6740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073036,
+      "loss": 1.1979,
+      "step": 6741
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073032,
+      "loss": 1.1383,
+      "step": 6742
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073028,
+      "loss": 1.2274,
+      "step": 6743
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007302400000000001,
+      "loss": 1.0581,
+      "step": 6744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007302,
+      "loss": 1.1553,
+      "step": 6745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007301600000000001,
+      "loss": 1.1833,
+      "step": 6746
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073012,
+      "loss": 1.2398,
+      "step": 6747
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007300799999999999,
+      "loss": 1.1667,
+      "step": 6748
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073004,
+      "loss": 1.1236,
+      "step": 6749
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00073,
+      "loss": 1.111,
+      "step": 6750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007299600000000001,
+      "loss": 1.1797,
+      "step": 6751
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072992,
+      "loss": 1.2164,
+      "step": 6752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072988,
+      "loss": 1.1584,
+      "step": 6753
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072984,
+      "loss": 1.26,
+      "step": 6754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007298000000000001,
+      "loss": 1.0739,
+      "step": 6755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072976,
+      "loss": 1.2266,
+      "step": 6756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072972,
+      "loss": 1.121,
+      "step": 6757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072968,
+      "loss": 1.0899,
+      "step": 6758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007296399999999999,
+      "loss": 1.1654,
+      "step": 6759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007296000000000001,
+      "loss": 1.1869,
+      "step": 6760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072956,
+      "loss": 1.1808,
+      "step": 6761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072952,
+      "loss": 1.2203,
+      "step": 6762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072948,
+      "loss": 1.1701,
+      "step": 6763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072944,
+      "loss": 1.1628,
+      "step": 6764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007294000000000001,
+      "loss": 1.216,
+      "step": 6765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007293600000000001,
+      "loss": 1.241,
+      "step": 6766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072932,
+      "loss": 1.0922,
+      "step": 6767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072928,
+      "loss": 1.2072,
+      "step": 6768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072924,
+      "loss": 1.1472,
+      "step": 6769
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007291999999999999,
+      "loss": 1.1297,
+      "step": 6770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007291600000000001,
+      "loss": 1.1435,
+      "step": 6771
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072912,
+      "loss": 1.2257,
+      "step": 6772
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072908,
+      "loss": 1.2005,
+      "step": 6773
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072904,
+      "loss": 1.116,
+      "step": 6774
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000729,
+      "loss": 1.1922,
+      "step": 6775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007289600000000001,
+      "loss": 1.1826,
+      "step": 6776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072892,
+      "loss": 1.1478,
+      "step": 6777
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072888,
+      "loss": 1.1765,
+      "step": 6778
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072884,
+      "loss": 1.1968,
+      "step": 6779
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007288,
+      "loss": 1.1522,
+      "step": 6780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072876,
+      "loss": 1.1769,
+      "step": 6781
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007287200000000001,
+      "loss": 1.212,
+      "step": 6782
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072868,
+      "loss": 1.139,
+      "step": 6783
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072864,
+      "loss": 1.1536,
+      "step": 6784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007286,
+      "loss": 1.2346,
+      "step": 6785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072856,
+      "loss": 1.1481,
+      "step": 6786
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072852,
+      "loss": 1.1443,
+      "step": 6787
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072848,
+      "loss": 1.1737,
+      "step": 6788
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072844,
+      "loss": 1.1044,
+      "step": 6789
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007284,
+      "loss": 1.1912,
+      "step": 6790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072836,
+      "loss": 1.199,
+      "step": 6791
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072832,
+      "loss": 1.1791,
+      "step": 6792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007282800000000001,
+      "loss": 1.1393,
+      "step": 6793
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072824,
+      "loss": 1.2143,
+      "step": 6794
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007281999999999999,
+      "loss": 1.1723,
+      "step": 6795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072816,
+      "loss": 1.1102,
+      "step": 6796
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072812,
+      "loss": 1.2244,
+      "step": 6797
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072808,
+      "loss": 1.1973,
+      "step": 6798
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072804,
+      "loss": 1.1446,
+      "step": 6799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000728,
+      "loss": 1.1047,
+      "step": 6800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072796,
+      "loss": 1.1772,
+      "step": 6801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007279200000000001,
+      "loss": 1.1451,
+      "step": 6802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072788,
+      "loss": 1.1845,
+      "step": 6803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007278400000000001,
+      "loss": 1.2085,
+      "step": 6804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007278,
+      "loss": 1.218,
+      "step": 6805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007277599999999999,
+      "loss": 1.1827,
+      "step": 6806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007277200000000001,
+      "loss": 1.175,
+      "step": 6807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072768,
+      "loss": 1.1488,
+      "step": 6808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072764,
+      "loss": 1.1704,
+      "step": 6809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007276,
+      "loss": 1.2271,
+      "step": 6810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072756,
+      "loss": 1.2206,
+      "step": 6811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072752,
+      "loss": 1.1852,
+      "step": 6812
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007274800000000001,
+      "loss": 1.2238,
+      "step": 6813
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072744,
+      "loss": 1.1863,
+      "step": 6814
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007274,
+      "loss": 1.1678,
+      "step": 6815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072736,
+      "loss": 1.0778,
+      "step": 6816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007273199999999999,
+      "loss": 1.1891,
+      "step": 6817
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007272800000000001,
+      "loss": 1.139,
+      "step": 6818
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072724,
+      "loss": 1.1301,
+      "step": 6819
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007272,
+      "loss": 1.1996,
+      "step": 6820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072716,
+      "loss": 1.2075,
+      "step": 6821
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072712,
+      "loss": 1.1826,
+      "step": 6822
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072708,
+      "loss": 1.1881,
+      "step": 6823
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007270400000000001,
+      "loss": 1.2417,
+      "step": 6824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000727,
+      "loss": 1.092,
+      "step": 6825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072696,
+      "loss": 1.1501,
+      "step": 6826
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072692,
+      "loss": 1.1994,
+      "step": 6827
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072688,
+      "loss": 1.1946,
+      "step": 6828
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007268400000000001,
+      "loss": 1.1814,
+      "step": 6829
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007268,
+      "loss": 1.0863,
+      "step": 6830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072676,
+      "loss": 1.0916,
+      "step": 6831
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072672,
+      "loss": 1.1984,
+      "step": 6832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007266800000000001,
+      "loss": 1.1019,
+      "step": 6833
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072664,
+      "loss": 1.2177,
+      "step": 6834
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007266,
+      "loss": 1.251,
+      "step": 6835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072656,
+      "loss": 1.166,
+      "step": 6836
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072652,
+      "loss": 1.2037,
+      "step": 6837
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072648,
+      "loss": 1.1958,
+      "step": 6838
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072644,
+      "loss": 1.2187,
+      "step": 6839
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007264000000000001,
+      "loss": 1.0863,
+      "step": 6840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072636,
+      "loss": 1.2283,
+      "step": 6841
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072632,
+      "loss": 1.1056,
+      "step": 6842
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072628,
+      "loss": 1.1813,
+      "step": 6843
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072624,
+      "loss": 1.193,
+      "step": 6844
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007262,
+      "loss": 1.1772,
+      "step": 6845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072616,
+      "loss": 1.2036,
+      "step": 6846
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072612,
+      "loss": 1.1572,
+      "step": 6847
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007260799999999999,
+      "loss": 1.2012,
+      "step": 6848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007260400000000001,
+      "loss": 1.2457,
+      "step": 6849
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000726,
+      "loss": 1.2068,
+      "step": 6850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007259600000000001,
+      "loss": 1.1338,
+      "step": 6851
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072592,
+      "loss": 1.1126,
+      "step": 6852
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072588,
+      "loss": 1.2098,
+      "step": 6853
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007258400000000001,
+      "loss": 1.1278,
+      "step": 6854
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007258,
+      "loss": 1.0763,
+      "step": 6855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072576,
+      "loss": 1.1019,
+      "step": 6856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072572,
+      "loss": 1.1991,
+      "step": 6857
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072568,
+      "loss": 1.1371,
+      "step": 6858
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007256399999999999,
+      "loss": 1.1892,
+      "step": 6859
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007256000000000001,
+      "loss": 1.1682,
+      "step": 6860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072556,
+      "loss": 1.2076,
+      "step": 6861
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007255200000000001,
+      "loss": 1.1746,
+      "step": 6862
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072548,
+      "loss": 1.1471,
+      "step": 6863
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007254399999999999,
+      "loss": 1.2562,
+      "step": 6864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007254000000000001,
+      "loss": 1.2191,
+      "step": 6865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072536,
+      "loss": 1.1803,
+      "step": 6866
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072532,
+      "loss": 1.1589,
+      "step": 6867
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072528,
+      "loss": 1.1274,
+      "step": 6868
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072524,
+      "loss": 1.2565,
+      "step": 6869
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007252,
+      "loss": 1.1574,
+      "step": 6870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007251600000000001,
+      "loss": 1.2898,
+      "step": 6871
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072512,
+      "loss": 1.1021,
+      "step": 6872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007250799999999999,
+      "loss": 1.1719,
+      "step": 6873
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072504,
+      "loss": 1.1466,
+      "step": 6874
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000725,
+      "loss": 1.1799,
+      "step": 6875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007249600000000001,
+      "loss": 1.2369,
+      "step": 6876
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072492,
+      "loss": 1.2305,
+      "step": 6877
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072488,
+      "loss": 1.1902,
+      "step": 6878
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072484,
+      "loss": 1.1752,
+      "step": 6879
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007248,
+      "loss": 1.09,
+      "step": 6880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072476,
+      "loss": 1.147,
+      "step": 6881
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007247200000000001,
+      "loss": 1.092,
+      "step": 6882
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072468,
+      "loss": 1.1731,
+      "step": 6883
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007246399999999999,
+      "loss": 1.1712,
+      "step": 6884
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007246,
+      "loss": 1.183,
+      "step": 6885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072456,
+      "loss": 1.1758,
+      "step": 6886
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007245200000000001,
+      "loss": 1.2032,
+      "step": 6887
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072448,
+      "loss": 1.1485,
+      "step": 6888
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072444,
+      "loss": 1.2159,
+      "step": 6889
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007244,
+      "loss": 1.2155,
+      "step": 6890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007243600000000001,
+      "loss": 1.1859,
+      "step": 6891
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072432,
+      "loss": 1.2291,
+      "step": 6892
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072428,
+      "loss": 1.2008,
+      "step": 6893
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072424,
+      "loss": 1.1427,
+      "step": 6894
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007241999999999999,
+      "loss": 1.1361,
+      "step": 6895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007241600000000001,
+      "loss": 1.1326,
+      "step": 6896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072412,
+      "loss": 1.1494,
+      "step": 6897
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072408,
+      "loss": 1.1611,
+      "step": 6898
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072404,
+      "loss": 1.2037,
+      "step": 6899
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000724,
+      "loss": 1.203,
+      "step": 6900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007239600000000001,
+      "loss": 1.1903,
+      "step": 6901
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007239200000000001,
+      "loss": 1.1741,
+      "step": 6902
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072388,
+      "loss": 1.1973,
+      "step": 6903
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072384,
+      "loss": 1.0954,
+      "step": 6904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007238,
+      "loss": 1.1834,
+      "step": 6905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007237599999999999,
+      "loss": 1.09,
+      "step": 6906
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007237200000000001,
+      "loss": 1.1128,
+      "step": 6907
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072368,
+      "loss": 1.1095,
+      "step": 6908
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072364,
+      "loss": 1.2286,
+      "step": 6909
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007236,
+      "loss": 1.2062,
+      "step": 6910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072356,
+      "loss": 1.1849,
+      "step": 6911
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007235200000000001,
+      "loss": 1.046,
+      "step": 6912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072348,
+      "loss": 1.1455,
+      "step": 6913
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072344,
+      "loss": 1.1421,
+      "step": 6914
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007234,
+      "loss": 1.0848,
+      "step": 6915
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072336,
+      "loss": 1.2382,
+      "step": 6916
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072332,
+      "loss": 1.1388,
+      "step": 6917
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007232800000000001,
+      "loss": 1.1758,
+      "step": 6918
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072324,
+      "loss": 1.1324,
+      "step": 6919
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007232,
+      "loss": 1.0853,
+      "step": 6920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072316,
+      "loss": 1.1618,
+      "step": 6921
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072312,
+      "loss": 1.1497,
+      "step": 6922
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072308,
+      "loss": 1.1307,
+      "step": 6923
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072304,
+      "loss": 1.179,
+      "step": 6924
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000723,
+      "loss": 1.2319,
+      "step": 6925
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072296,
+      "loss": 1.1422,
+      "step": 6926
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072292,
+      "loss": 1.2352,
+      "step": 6927
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072288,
+      "loss": 1.2236,
+      "step": 6928
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007228400000000001,
+      "loss": 1.1559,
+      "step": 6929
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007228,
+      "loss": 1.1852,
+      "step": 6930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072276,
+      "loss": 1.2198,
+      "step": 6931
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072272,
+      "loss": 1.163,
+      "step": 6932
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072268,
+      "loss": 1.1868,
+      "step": 6933
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072264,
+      "loss": 1.1777,
+      "step": 6934
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007226,
+      "loss": 1.1456,
+      "step": 6935
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072256,
+      "loss": 1.2138,
+      "step": 6936
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072252,
+      "loss": 1.2181,
+      "step": 6937
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007224800000000001,
+      "loss": 1.2614,
+      "step": 6938
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072244,
+      "loss": 1.1059,
+      "step": 6939
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007224000000000001,
+      "loss": 1.2631,
+      "step": 6940
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072236,
+      "loss": 1.1132,
+      "step": 6941
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007223199999999999,
+      "loss": 1.1081,
+      "step": 6942
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007222800000000001,
+      "loss": 1.2873,
+      "step": 6943
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072224,
+      "loss": 1.2036,
+      "step": 6944
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007222,
+      "loss": 1.1974,
+      "step": 6945
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072216,
+      "loss": 1.1649,
+      "step": 6946
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072212,
+      "loss": 1.1995,
+      "step": 6947
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072208,
+      "loss": 1.0915,
+      "step": 6948
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007220400000000001,
+      "loss": 1.1356,
+      "step": 6949
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000722,
+      "loss": 1.1625,
+      "step": 6950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007219600000000001,
+      "loss": 1.0647,
+      "step": 6951
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072192,
+      "loss": 1.2066,
+      "step": 6952
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007218799999999999,
+      "loss": 1.1539,
+      "step": 6953
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007218400000000001,
+      "loss": 1.1715,
+      "step": 6954
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007218,
+      "loss": 1.2343,
+      "step": 6955
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072176,
+      "loss": 1.1728,
+      "step": 6956
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072172,
+      "loss": 1.2251,
+      "step": 6957
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072168,
+      "loss": 1.1634,
+      "step": 6958
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072164,
+      "loss": 1.0396,
+      "step": 6959
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007216000000000001,
+      "loss": 1.1454,
+      "step": 6960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072156,
+      "loss": 1.2405,
+      "step": 6961
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072152,
+      "loss": 1.1808,
+      "step": 6962
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072148,
+      "loss": 1.1497,
+      "step": 6963
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072144,
+      "loss": 1.2874,
+      "step": 6964
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0007214000000000001,
+      "loss": 1.1045,
+      "step": 6965
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072136,
+      "loss": 1.1088,
+      "step": 6966
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072132,
+      "loss": 1.1066,
+      "step": 6967
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00072128,
+      "loss": 1.1797,
+      "step": 6968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072124,
+      "loss": 1.3237,
+      "step": 6969
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007212,
+      "loss": 1.1702,
+      "step": 6970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072116,
+      "loss": 1.1341,
+      "step": 6971
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072112,
+      "loss": 1.225,
+      "step": 6972
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072108,
+      "loss": 1.1671,
+      "step": 6973
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072104,
+      "loss": 1.1064,
+      "step": 6974
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000721,
+      "loss": 1.1808,
+      "step": 6975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007209600000000001,
+      "loss": 1.2148,
+      "step": 6976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072092,
+      "loss": 1.2301,
+      "step": 6977
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072088,
+      "loss": 1.0839,
+      "step": 6978
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072084,
+      "loss": 1.1386,
+      "step": 6979
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007208000000000001,
+      "loss": 1.2427,
+      "step": 6980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072076,
+      "loss": 1.2398,
+      "step": 6981
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072072,
+      "loss": 1.1653,
+      "step": 6982
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072068,
+      "loss": 1.1706,
+      "step": 6983
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007206399999999999,
+      "loss": 1.2554,
+      "step": 6984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007206000000000001,
+      "loss": 1.2247,
+      "step": 6985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072056,
+      "loss": 1.1984,
+      "step": 6986
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007205200000000001,
+      "loss": 1.0732,
+      "step": 6987
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072048,
+      "loss": 1.1527,
+      "step": 6988
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072044,
+      "loss": 1.2365,
+      "step": 6989
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007204000000000001,
+      "loss": 1.1491,
+      "step": 6990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072036,
+      "loss": 1.2397,
+      "step": 6991
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072032,
+      "loss": 1.148,
+      "step": 6992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072028,
+      "loss": 1.138,
+      "step": 6993
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072024,
+      "loss": 1.1913,
+      "step": 6994
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007201999999999999,
+      "loss": 1.173,
+      "step": 6995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007201600000000001,
+      "loss": 1.1604,
+      "step": 6996
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072012,
+      "loss": 1.1471,
+      "step": 6997
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007200800000000001,
+      "loss": 1.1616,
+      "step": 6998
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00072004,
+      "loss": 1.1925,
+      "step": 6999
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007199999999999999,
+      "loss": 1.1269,
+      "step": 7000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007199600000000001,
+      "loss": 1.203,
+      "step": 7001
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071992,
+      "loss": 1.2118,
+      "step": 7002
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071988,
+      "loss": 1.2081,
+      "step": 7003
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071984,
+      "loss": 1.0779,
+      "step": 7004
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007198,
+      "loss": 1.1247,
+      "step": 7005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071976,
+      "loss": 1.1488,
+      "step": 7006
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007197200000000001,
+      "loss": 1.2819,
+      "step": 7007
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071968,
+      "loss": 1.0989,
+      "step": 7008
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071964,
+      "loss": 1.2389,
+      "step": 7009
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007196,
+      "loss": 1.232,
+      "step": 7010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071956,
+      "loss": 1.2117,
+      "step": 7011
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007195200000000001,
+      "loss": 1.1505,
+      "step": 7012
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071948,
+      "loss": 1.2435,
+      "step": 7013
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071944,
+      "loss": 1.1341,
+      "step": 7014
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007194,
+      "loss": 1.2469,
+      "step": 7015
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071936,
+      "loss": 1.0948,
+      "step": 7016
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071932,
+      "loss": 1.1623,
+      "step": 7017
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007192800000000001,
+      "loss": 1.1687,
+      "step": 7018
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071924,
+      "loss": 1.2568,
+      "step": 7019
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007191999999999999,
+      "loss": 1.0954,
+      "step": 7020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071916,
+      "loss": 1.1301,
+      "step": 7021
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071912,
+      "loss": 1.2241,
+      "step": 7022
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007190800000000001,
+      "loss": 1.1252,
+      "step": 7023
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071904,
+      "loss": 1.1757,
+      "step": 7024
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000719,
+      "loss": 1.1602,
+      "step": 7025
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071896,
+      "loss": 1.1776,
+      "step": 7026
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007189200000000001,
+      "loss": 1.1364,
+      "step": 7027
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071888,
+      "loss": 1.0532,
+      "step": 7028
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071884,
+      "loss": 1.18,
+      "step": 7029
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007188,
+      "loss": 1.1905,
+      "step": 7030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007187599999999999,
+      "loss": 1.1671,
+      "step": 7031
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007187200000000001,
+      "loss": 1.1877,
+      "step": 7032
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071868,
+      "loss": 1.06,
+      "step": 7033
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071864,
+      "loss": 1.2246,
+      "step": 7034
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007186,
+      "loss": 1.1632,
+      "step": 7035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071856,
+      "loss": 1.2093,
+      "step": 7036
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071852,
+      "loss": 1.0983,
+      "step": 7037
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007184800000000001,
+      "loss": 1.1188,
+      "step": 7038
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071844,
+      "loss": 1.1701,
+      "step": 7039
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007184,
+      "loss": 1.0853,
+      "step": 7040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071836,
+      "loss": 1.0767,
+      "step": 7041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007183199999999999,
+      "loss": 1.0987,
+      "step": 7042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007182800000000001,
+      "loss": 1.2388,
+      "step": 7043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071824,
+      "loss": 1.1684,
+      "step": 7044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007182,
+      "loss": 1.1733,
+      "step": 7045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071816,
+      "loss": 1.1446,
+      "step": 7046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071812,
+      "loss": 1.0915,
+      "step": 7047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007180800000000001,
+      "loss": 1.1026,
+      "step": 7048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071804,
+      "loss": 1.1483,
+      "step": 7049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000718,
+      "loss": 1.1709,
+      "step": 7050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071796,
+      "loss": 1.1767,
+      "step": 7051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071792,
+      "loss": 1.1347,
+      "step": 7052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071788,
+      "loss": 1.2643,
+      "step": 7053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007178400000000001,
+      "loss": 1.1676,
+      "step": 7054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007178,
+      "loss": 1.1611,
+      "step": 7055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071776,
+      "loss": 1.1087,
+      "step": 7056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071772,
+      "loss": 1.1623,
+      "step": 7057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007176799999999999,
+      "loss": 1.2038,
+      "step": 7058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071764,
+      "loss": 1.1617,
+      "step": 7059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007176,
+      "loss": 1.2124,
+      "step": 7060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071756,
+      "loss": 1.133,
+      "step": 7061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071752,
+      "loss": 1.1517,
+      "step": 7062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071748,
+      "loss": 1.2454,
+      "step": 7063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071744,
+      "loss": 1.1633,
+      "step": 7064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007174000000000001,
+      "loss": 1.1497,
+      "step": 7065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071736,
+      "loss": 1.1695,
+      "step": 7066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071732,
+      "loss": 1.159,
+      "step": 7067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071728,
+      "loss": 1.1331,
+      "step": 7068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071724,
+      "loss": 1.2049,
+      "step": 7069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007172,
+      "loss": 1.2017,
+      "step": 7070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071716,
+      "loss": 1.1072,
+      "step": 7071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071712,
+      "loss": 1.1864,
+      "step": 7072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071708,
+      "loss": 1.1442,
+      "step": 7073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007170400000000001,
+      "loss": 1.1688,
+      "step": 7074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000717,
+      "loss": 1.1194,
+      "step": 7075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007169600000000001,
+      "loss": 1.0746,
+      "step": 7076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071692,
+      "loss": 1.2301,
+      "step": 7077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007168799999999999,
+      "loss": 1.1615,
+      "step": 7078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007168400000000001,
+      "loss": 1.212,
+      "step": 7079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007168,
+      "loss": 1.1173,
+      "step": 7080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071676,
+      "loss": 1.1592,
+      "step": 7081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071672,
+      "loss": 1.1681,
+      "step": 7082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071668,
+      "loss": 1.0771,
+      "step": 7083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071664,
+      "loss": 1.2003,
+      "step": 7084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007166000000000001,
+      "loss": 1.1824,
+      "step": 7085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071656,
+      "loss": 1.0714,
+      "step": 7086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007165200000000001,
+      "loss": 1.2075,
+      "step": 7087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071648,
+      "loss": 1.11,
+      "step": 7088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007164399999999999,
+      "loss": 1.2142,
+      "step": 7089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007164000000000001,
+      "loss": 1.1518,
+      "step": 7090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071636,
+      "loss": 1.2025,
+      "step": 7091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071632,
+      "loss": 1.153,
+      "step": 7092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071628,
+      "loss": 1.2313,
+      "step": 7093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071624,
+      "loss": 1.2366,
+      "step": 7094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007162,
+      "loss": 1.1901,
+      "step": 7095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007161600000000001,
+      "loss": 1.1818,
+      "step": 7096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071612,
+      "loss": 1.1663,
+      "step": 7097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071608,
+      "loss": 1.1469,
+      "step": 7098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071604,
+      "loss": 1.2476,
+      "step": 7099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000716,
+      "loss": 1.2188,
+      "step": 7100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007159600000000001,
+      "loss": 1.1467,
+      "step": 7101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071592,
+      "loss": 1.1833,
+      "step": 7102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071588,
+      "loss": 1.1487,
+      "step": 7103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071584,
+      "loss": 1.1315,
+      "step": 7104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007158,
+      "loss": 1.227,
+      "step": 7105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071576,
+      "loss": 1.114,
+      "step": 7106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071572,
+      "loss": 1.167,
+      "step": 7107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071568,
+      "loss": 1.1774,
+      "step": 7108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071564,
+      "loss": 1.1286,
+      "step": 7109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007156,
+      "loss": 1.1716,
+      "step": 7110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071556,
+      "loss": 1.1419,
+      "step": 7111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007155200000000001,
+      "loss": 1.1526,
+      "step": 7112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071548,
+      "loss": 1.2258,
+      "step": 7113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071544,
+      "loss": 1.1887,
+      "step": 7114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007154,
+      "loss": 1.0916,
+      "step": 7115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007153600000000001,
+      "loss": 1.1552,
+      "step": 7116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071532,
+      "loss": 1.2306,
+      "step": 7117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071528,
+      "loss": 1.2159,
+      "step": 7118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071524,
+      "loss": 1.1509,
+      "step": 7119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007151999999999999,
+      "loss": 1.2229,
+      "step": 7120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007151600000000001,
+      "loss": 1.2609,
+      "step": 7121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071512,
+      "loss": 1.1929,
+      "step": 7122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007150800000000001,
+      "loss": 1.2776,
+      "step": 7123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071504,
+      "loss": 1.0905,
+      "step": 7124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000715,
+      "loss": 1.1256,
+      "step": 7125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071496,
+      "loss": 1.1178,
+      "step": 7126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071492,
+      "loss": 1.2301,
+      "step": 7127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071488,
+      "loss": 1.2174,
+      "step": 7128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071484,
+      "loss": 1.1014,
+      "step": 7129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007148,
+      "loss": 1.0861,
+      "step": 7130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007147599999999999,
+      "loss": 1.1398,
+      "step": 7131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007147200000000001,
+      "loss": 1.1572,
+      "step": 7132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071468,
+      "loss": 1.0798,
+      "step": 7133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007146400000000001,
+      "loss": 1.1779,
+      "step": 7134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007146,
+      "loss": 1.177,
+      "step": 7135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007145599999999999,
+      "loss": 1.1604,
+      "step": 7136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007145200000000001,
+      "loss": 1.1087,
+      "step": 7137
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071448,
+      "loss": 1.1852,
+      "step": 7138
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071444,
+      "loss": 1.2234,
+      "step": 7139
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007144,
+      "loss": 1.1921,
+      "step": 7140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071436,
+      "loss": 1.1902,
+      "step": 7141
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071432,
+      "loss": 1.2082,
+      "step": 7142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007142800000000001,
+      "loss": 1.1392,
+      "step": 7143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071424,
+      "loss": 1.1637,
+      "step": 7144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007142,
+      "loss": 1.2631,
+      "step": 7145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071416,
+      "loss": 1.101,
+      "step": 7146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007141199999999999,
+      "loss": 1.1714,
+      "step": 7147
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007140800000000001,
+      "loss": 1.2015,
+      "step": 7148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071404,
+      "loss": 1.1752,
+      "step": 7149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000714,
+      "loss": 1.1697,
+      "step": 7150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071396,
+      "loss": 1.1297,
+      "step": 7151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071392,
+      "loss": 1.1854,
+      "step": 7152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071388,
+      "loss": 1.1879,
+      "step": 7153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007138400000000001,
+      "loss": 1.1366,
+      "step": 7154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007138,
+      "loss": 1.1671,
+      "step": 7155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007137599999999999,
+      "loss": 1.075,
+      "step": 7156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071372,
+      "loss": 1.1572,
+      "step": 7157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071368,
+      "loss": 1.151,
+      "step": 7158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007136400000000001,
+      "loss": 1.1607,
+      "step": 7159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007136,
+      "loss": 1.2576,
+      "step": 7160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071356,
+      "loss": 1.1345,
+      "step": 7161
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071352,
+      "loss": 1.1224,
+      "step": 7162
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007134800000000001,
+      "loss": 1.1872,
+      "step": 7163
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071344,
+      "loss": 1.1752,
+      "step": 7164
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007134000000000001,
+      "loss": 1.1605,
+      "step": 7165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071336,
+      "loss": 1.2115,
+      "step": 7166
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007133199999999999,
+      "loss": 1.1817,
+      "step": 7167
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007132800000000001,
+      "loss": 1.2163,
+      "step": 7168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071324,
+      "loss": 1.1478,
+      "step": 7169
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007132,
+      "loss": 1.0452,
+      "step": 7170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071316,
+      "loss": 1.1319,
+      "step": 7171
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071312,
+      "loss": 1.1563,
+      "step": 7172
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071308,
+      "loss": 1.1848,
+      "step": 7173
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007130400000000001,
+      "loss": 1.1902,
+      "step": 7174
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000713,
+      "loss": 1.1022,
+      "step": 7175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071296,
+      "loss": 1.1493,
+      "step": 7176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071292,
+      "loss": 1.1953,
+      "step": 7177
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007128799999999999,
+      "loss": 1.0996,
+      "step": 7178
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007128400000000001,
+      "loss": 1.1637,
+      "step": 7179
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007128,
+      "loss": 1.2046,
+      "step": 7180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071276,
+      "loss": 1.2585,
+      "step": 7181
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071272,
+      "loss": 1.212,
+      "step": 7182
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071268,
+      "loss": 1.1315,
+      "step": 7183
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007126400000000001,
+      "loss": 1.1132,
+      "step": 7184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007126,
+      "loss": 1.2405,
+      "step": 7185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071256,
+      "loss": 1.1695,
+      "step": 7186
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071252,
+      "loss": 1.171,
+      "step": 7187
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071248,
+      "loss": 1.201,
+      "step": 7188
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071244,
+      "loss": 1.1684,
+      "step": 7189
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007124000000000001,
+      "loss": 1.2529,
+      "step": 7190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071236,
+      "loss": 1.1536,
+      "step": 7191
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071232,
+      "loss": 1.1488,
+      "step": 7192
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071228,
+      "loss": 1.1081,
+      "step": 7193
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071224,
+      "loss": 1.2038,
+      "step": 7194
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007122000000000001,
+      "loss": 1.1606,
+      "step": 7195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071216,
+      "loss": 1.0845,
+      "step": 7196
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071212,
+      "loss": 1.105,
+      "step": 7197
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071208,
+      "loss": 1.1644,
+      "step": 7198
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071204,
+      "loss": 1.1602,
+      "step": 7199
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000712,
+      "loss": 1.1347,
+      "step": 7200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007119600000000001,
+      "loss": 1.1813,
+      "step": 7201
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071192,
+      "loss": 1.1027,
+      "step": 7202
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071188,
+      "loss": 1.1129,
+      "step": 7203
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071184,
+      "loss": 1.1748,
+      "step": 7204
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007118,
+      "loss": 1.1283,
+      "step": 7205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071176,
+      "loss": 1.1853,
+      "step": 7206
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071172,
+      "loss": 1.0835,
+      "step": 7207
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071168,
+      "loss": 1.1722,
+      "step": 7208
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071164,
+      "loss": 1.1324,
+      "step": 7209
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007116000000000001,
+      "loss": 1.1558,
+      "step": 7210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071156,
+      "loss": 1.1462,
+      "step": 7211
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007115200000000001,
+      "loss": 1.0847,
+      "step": 7212
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071148,
+      "loss": 1.1114,
+      "step": 7213
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007114399999999999,
+      "loss": 1.1452,
+      "step": 7214
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007114,
+      "loss": 1.1115,
+      "step": 7215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071136,
+      "loss": 1.3236,
+      "step": 7216
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071132,
+      "loss": 1.1613,
+      "step": 7217
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071128,
+      "loss": 1.0807,
+      "step": 7218
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00071124,
+      "loss": 1.1945,
+      "step": 7219
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007112,
+      "loss": 1.0816,
+      "step": 7220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0007111600000000001,
+      "loss": 1.2036,
+      "step": 7221
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071112,
+      "loss": 1.1727,
+      "step": 7222
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007110800000000001,
+      "loss": 1.1828,
+      "step": 7223
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071104,
+      "loss": 1.1729,
+      "step": 7224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007109999999999999,
+      "loss": 1.1008,
+      "step": 7225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007109600000000001,
+      "loss": 1.114,
+      "step": 7226
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071092,
+      "loss": 1.1284,
+      "step": 7227
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071088,
+      "loss": 1.1259,
+      "step": 7228
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071084,
+      "loss": 1.1871,
+      "step": 7229
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007108,
+      "loss": 1.1183,
+      "step": 7230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071076,
+      "loss": 1.2052,
+      "step": 7231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007107200000000001,
+      "loss": 1.2054,
+      "step": 7232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071068,
+      "loss": 1.1328,
+      "step": 7233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071064,
+      "loss": 1.1732,
+      "step": 7234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007106,
+      "loss": 1.204,
+      "step": 7235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071056,
+      "loss": 1.1281,
+      "step": 7236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007105200000000001,
+      "loss": 1.171,
+      "step": 7237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071048,
+      "loss": 1.1714,
+      "step": 7238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071044,
+      "loss": 1.1563,
+      "step": 7239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007104,
+      "loss": 1.1873,
+      "step": 7240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071036,
+      "loss": 1.1577,
+      "step": 7241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071032,
+      "loss": 1.0581,
+      "step": 7242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007102800000000001,
+      "loss": 1.1737,
+      "step": 7243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071024,
+      "loss": 1.1628,
+      "step": 7244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007102,
+      "loss": 1.1348,
+      "step": 7245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071016,
+      "loss": 1.1183,
+      "step": 7246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071012,
+      "loss": 1.1914,
+      "step": 7247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007100800000000001,
+      "loss": 1.1324,
+      "step": 7248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071004,
+      "loss": 1.2185,
+      "step": 7249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00071,
+      "loss": 1.1518,
+      "step": 7250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070996,
+      "loss": 1.1019,
+      "step": 7251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007099200000000001,
+      "loss": 1.1836,
+      "step": 7252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070988,
+      "loss": 1.1359,
+      "step": 7253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070984,
+      "loss": 1.1353,
+      "step": 7254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007098,
+      "loss": 1.2154,
+      "step": 7255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007097599999999999,
+      "loss": 1.0931,
+      "step": 7256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007097200000000001,
+      "loss": 1.1858,
+      "step": 7257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070968,
+      "loss": 1.1206,
+      "step": 7258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007096400000000001,
+      "loss": 1.2128,
+      "step": 7259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007096,
+      "loss": 1.1802,
+      "step": 7260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070956,
+      "loss": 1.1971,
+      "step": 7261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070952,
+      "loss": 1.1301,
+      "step": 7262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070948,
+      "loss": 1.1083,
+      "step": 7263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070944,
+      "loss": 1.2463,
+      "step": 7264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007094,
+      "loss": 1.1753,
+      "step": 7265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070936,
+      "loss": 1.1313,
+      "step": 7266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007093199999999999,
+      "loss": 1.1485,
+      "step": 7267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007092800000000001,
+      "loss": 1.0561,
+      "step": 7268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070924,
+      "loss": 1.1687,
+      "step": 7269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007092000000000001,
+      "loss": 1.1805,
+      "step": 7270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070916,
+      "loss": 1.2746,
+      "step": 7271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070912,
+      "loss": 1.1137,
+      "step": 7272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007090800000000001,
+      "loss": 1.1434,
+      "step": 7273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070904,
+      "loss": 1.1731,
+      "step": 7274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000709,
+      "loss": 1.3178,
+      "step": 7275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070896,
+      "loss": 1.2105,
+      "step": 7276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070892,
+      "loss": 1.175,
+      "step": 7277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070888,
+      "loss": 1.1816,
+      "step": 7278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007088400000000001,
+      "loss": 1.1208,
+      "step": 7279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007088,
+      "loss": 1.0927,
+      "step": 7280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070876,
+      "loss": 1.1519,
+      "step": 7281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070872,
+      "loss": 1.2243,
+      "step": 7282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007086799999999999,
+      "loss": 1.107,
+      "step": 7283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007086400000000001,
+      "loss": 1.0783,
+      "step": 7284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007086,
+      "loss": 1.1126,
+      "step": 7285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070856,
+      "loss": 1.2132,
+      "step": 7286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070852,
+      "loss": 1.1483,
+      "step": 7287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070848,
+      "loss": 1.1079,
+      "step": 7288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070844,
+      "loss": 1.1074,
+      "step": 7289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007084000000000001,
+      "loss": 1.213,
+      "step": 7290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070836,
+      "loss": 1.1741,
+      "step": 7291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007083199999999999,
+      "loss": 1.1604,
+      "step": 7292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070828,
+      "loss": 1.1332,
+      "step": 7293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070824,
+      "loss": 1.1506,
+      "step": 7294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007082000000000001,
+      "loss": 1.2038,
+      "step": 7295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070816,
+      "loss": 1.2002,
+      "step": 7296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070812,
+      "loss": 1.0894,
+      "step": 7297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070808,
+      "loss": 1.1214,
+      "step": 7298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007080400000000001,
+      "loss": 1.1846,
+      "step": 7299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000708,
+      "loss": 1.1173,
+      "step": 7300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007079600000000001,
+      "loss": 1.2154,
+      "step": 7301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070792,
+      "loss": 1.1754,
+      "step": 7302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007078799999999999,
+      "loss": 1.1575,
+      "step": 7303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070784,
+      "loss": 1.1981,
+      "step": 7304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007078,
+      "loss": 1.2606,
+      "step": 7305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070776,
+      "loss": 1.0916,
+      "step": 7306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070772,
+      "loss": 1.1839,
+      "step": 7307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070768,
+      "loss": 1.1633,
+      "step": 7308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070764,
+      "loss": 1.1089,
+      "step": 7309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007076000000000001,
+      "loss": 1.2276,
+      "step": 7310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070756,
+      "loss": 1.128,
+      "step": 7311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070752,
+      "loss": 1.1768,
+      "step": 7312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070748,
+      "loss": 1.2607,
+      "step": 7313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007074399999999999,
+      "loss": 1.1956,
+      "step": 7314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007074000000000001,
+      "loss": 1.152,
+      "step": 7315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070736,
+      "loss": 1.1136,
+      "step": 7316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070732,
+      "loss": 1.2807,
+      "step": 7317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070728,
+      "loss": 1.109,
+      "step": 7318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070724,
+      "loss": 1.0872,
+      "step": 7319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007072000000000001,
+      "loss": 1.1598,
+      "step": 7320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070716,
+      "loss": 1.1313,
+      "step": 7321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070712,
+      "loss": 1.1651,
+      "step": 7322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070708,
+      "loss": 1.2028,
+      "step": 7323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070704,
+      "loss": 1.1761,
+      "step": 7324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000707,
+      "loss": 1.1933,
+      "step": 7325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007069600000000001,
+      "loss": 1.2343,
+      "step": 7326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070692,
+      "loss": 1.1875,
+      "step": 7327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070688,
+      "loss": 1.0796,
+      "step": 7328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070684,
+      "loss": 1.2278,
+      "step": 7329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007068,
+      "loss": 1.1261,
+      "step": 7330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007067600000000001,
+      "loss": 1.2186,
+      "step": 7331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070672,
+      "loss": 1.0807,
+      "step": 7332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070668,
+      "loss": 1.1009,
+      "step": 7333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070664,
+      "loss": 1.2751,
+      "step": 7334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007066,
+      "loss": 1.1006,
+      "step": 7335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070656,
+      "loss": 1.114,
+      "step": 7336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007065200000000001,
+      "loss": 1.1657,
+      "step": 7337
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070648,
+      "loss": 1.1689,
+      "step": 7338
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070644,
+      "loss": 1.2465,
+      "step": 7339
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007064,
+      "loss": 1.1126,
+      "step": 7340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070636,
+      "loss": 1.2889,
+      "step": 7341
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070632,
+      "loss": 1.0978,
+      "step": 7342
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070628,
+      "loss": 1.1469,
+      "step": 7343
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070624,
+      "loss": 1.1122,
+      "step": 7344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007062,
+      "loss": 1.1223,
+      "step": 7345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007061600000000001,
+      "loss": 1.1306,
+      "step": 7346
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070612,
+      "loss": 1.2449,
+      "step": 7347
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007060800000000001,
+      "loss": 1.1731,
+      "step": 7348
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070604,
+      "loss": 1.1555,
+      "step": 7349
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007059999999999999,
+      "loss": 1.2259,
+      "step": 7350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070596,
+      "loss": 0.9883,
+      "step": 7351
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070592,
+      "loss": 1.2212,
+      "step": 7352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070588,
+      "loss": 1.1911,
+      "step": 7353
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070584,
+      "loss": 1.1303,
+      "step": 7354
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007058,
+      "loss": 1.2529,
+      "step": 7355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070576,
+      "loss": 1.1436,
+      "step": 7356
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007057200000000001,
+      "loss": 1.1494,
+      "step": 7357
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070568,
+      "loss": 1.1477,
+      "step": 7358
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007056400000000001,
+      "loss": 1.1138,
+      "step": 7359
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007056,
+      "loss": 1.1279,
+      "step": 7360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007055599999999999,
+      "loss": 1.2029,
+      "step": 7361
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007055200000000001,
+      "loss": 1.1279,
+      "step": 7362
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070548,
+      "loss": 1.1758,
+      "step": 7363
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070544,
+      "loss": 1.2595,
+      "step": 7364
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007054,
+      "loss": 1.1324,
+      "step": 7365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070536,
+      "loss": 1.2041,
+      "step": 7366
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070532,
+      "loss": 1.2025,
+      "step": 7367
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007052800000000001,
+      "loss": 1.1428,
+      "step": 7368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070524,
+      "loss": 1.1859,
+      "step": 7369
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007052,
+      "loss": 1.222,
+      "step": 7370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070516,
+      "loss": 1.024,
+      "step": 7371
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007051199999999999,
+      "loss": 1.1523,
+      "step": 7372
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007050800000000001,
+      "loss": 1.231,
+      "step": 7373
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070504,
+      "loss": 1.2357,
+      "step": 7374
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000705,
+      "loss": 1.2084,
+      "step": 7375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070496,
+      "loss": 1.0775,
+      "step": 7376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070492,
+      "loss": 1.2035,
+      "step": 7377
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070488,
+      "loss": 1.0781,
+      "step": 7378
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007048400000000001,
+      "loss": 1.1809,
+      "step": 7379
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007048,
+      "loss": 1.1233,
+      "step": 7380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070476,
+      "loss": 1.1068,
+      "step": 7381
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070472,
+      "loss": 1.1928,
+      "step": 7382
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070468,
+      "loss": 1.1296,
+      "step": 7383
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007046400000000001,
+      "loss": 1.1932,
+      "step": 7384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007046,
+      "loss": 1.1265,
+      "step": 7385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070456,
+      "loss": 1.0507,
+      "step": 7386
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070452,
+      "loss": 1.0446,
+      "step": 7387
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007044800000000001,
+      "loss": 1.1995,
+      "step": 7388
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070444,
+      "loss": 1.1633,
+      "step": 7389
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007044,
+      "loss": 1.1905,
+      "step": 7390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070436,
+      "loss": 1.1826,
+      "step": 7391
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007043199999999999,
+      "loss": 1.1793,
+      "step": 7392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070428,
+      "loss": 1.1039,
+      "step": 7393
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070424,
+      "loss": 1.1482,
+      "step": 7394
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007042000000000001,
+      "loss": 1.1291,
+      "step": 7395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070416,
+      "loss": 1.1206,
+      "step": 7396
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070412,
+      "loss": 1.1963,
+      "step": 7397
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070408,
+      "loss": 1.172,
+      "step": 7398
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070404,
+      "loss": 1.2312,
+      "step": 7399
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000704,
+      "loss": 1.122,
+      "step": 7400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070396,
+      "loss": 1.1954,
+      "step": 7401
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070392,
+      "loss": 1.1473,
+      "step": 7402
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007038799999999999,
+      "loss": 1.1518,
+      "step": 7403
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007038400000000001,
+      "loss": 1.24,
+      "step": 7404
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007038,
+      "loss": 1.0877,
+      "step": 7405
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007037600000000001,
+      "loss": 1.1745,
+      "step": 7406
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070372,
+      "loss": 1.0924,
+      "step": 7407
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070368,
+      "loss": 1.1869,
+      "step": 7408
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007036400000000001,
+      "loss": 1.1112,
+      "step": 7409
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007036,
+      "loss": 1.2068,
+      "step": 7410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070356,
+      "loss": 1.1233,
+      "step": 7411
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070352,
+      "loss": 1.226,
+      "step": 7412
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070348,
+      "loss": 1.2007,
+      "step": 7413
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070344,
+      "loss": 1.0908,
+      "step": 7414
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007034000000000001,
+      "loss": 1.1008,
+      "step": 7415
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070336,
+      "loss": 1.1285,
+      "step": 7416
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070332,
+      "loss": 1.0954,
+      "step": 7417
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070328,
+      "loss": 1.1479,
+      "step": 7418
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007032399999999999,
+      "loss": 1.1225,
+      "step": 7419
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007032000000000001,
+      "loss": 1.1812,
+      "step": 7420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070316,
+      "loss": 1.1839,
+      "step": 7421
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070312,
+      "loss": 1.2221,
+      "step": 7422
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070308,
+      "loss": 1.2153,
+      "step": 7423
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070304,
+      "loss": 1.1685,
+      "step": 7424
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000703,
+      "loss": 1.1234,
+      "step": 7425
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007029600000000001,
+      "loss": 1.0972,
+      "step": 7426
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070292,
+      "loss": 1.1892,
+      "step": 7427
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007028799999999999,
+      "loss": 1.2124,
+      "step": 7428
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070284,
+      "loss": 1.0671,
+      "step": 7429
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007028,
+      "loss": 1.115,
+      "step": 7430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007027600000000001,
+      "loss": 1.2588,
+      "step": 7431
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070272,
+      "loss": 1.1935,
+      "step": 7432
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070268,
+      "loss": 1.1261,
+      "step": 7433
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070264,
+      "loss": 1.1559,
+      "step": 7434
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007026000000000001,
+      "loss": 1.0628,
+      "step": 7435
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070256,
+      "loss": 1.2051,
+      "step": 7436
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007025200000000001,
+      "loss": 1.224,
+      "step": 7437
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070248,
+      "loss": 1.1785,
+      "step": 7438
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007024399999999999,
+      "loss": 1.0651,
+      "step": 7439
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007024,
+      "loss": 1.0987,
+      "step": 7440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070236,
+      "loss": 1.1498,
+      "step": 7441
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007023200000000001,
+      "loss": 1.1978,
+      "step": 7442
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070228,
+      "loss": 1.1184,
+      "step": 7443
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070224,
+      "loss": 1.1216,
+      "step": 7444
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007022,
+      "loss": 1.2093,
+      "step": 7445
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007021600000000001,
+      "loss": 1.1443,
+      "step": 7446
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070212,
+      "loss": 1.11,
+      "step": 7447
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070208,
+      "loss": 1.1193,
+      "step": 7448
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070204,
+      "loss": 1.0669,
+      "step": 7449
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007019999999999999,
+      "loss": 1.135,
+      "step": 7450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007019600000000001,
+      "loss": 1.2258,
+      "step": 7451
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070192,
+      "loss": 1.1155,
+      "step": 7452
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070188,
+      "loss": 1.1895,
+      "step": 7453
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070184,
+      "loss": 1.1762,
+      "step": 7454
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007018,
+      "loss": 1.1558,
+      "step": 7455
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007017600000000001,
+      "loss": 1.2366,
+      "step": 7456
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007017200000000001,
+      "loss": 1.1282,
+      "step": 7457
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070168,
+      "loss": 1.0524,
+      "step": 7458
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070164,
+      "loss": 1.1448,
+      "step": 7459
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007016,
+      "loss": 1.1586,
+      "step": 7460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007015599999999999,
+      "loss": 1.1043,
+      "step": 7461
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007015200000000001,
+      "loss": 1.1117,
+      "step": 7462
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070148,
+      "loss": 1.1957,
+      "step": 7463
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070144,
+      "loss": 1.1998,
+      "step": 7464
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007014,
+      "loss": 1.181,
+      "step": 7465
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070136,
+      "loss": 1.2226,
+      "step": 7466
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007013200000000001,
+      "loss": 1.1591,
+      "step": 7467
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070128,
+      "loss": 1.2341,
+      "step": 7468
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070124,
+      "loss": 1.2027,
+      "step": 7469
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007012,
+      "loss": 1.1468,
+      "step": 7470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070116,
+      "loss": 1.0706,
+      "step": 7471
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070112,
+      "loss": 1.1054,
+      "step": 7472
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0007010800000000001,
+      "loss": 1.1487,
+      "step": 7473
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00070104,
+      "loss": 1.1021,
+      "step": 7474
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000701,
+      "loss": 1.0539,
+      "step": 7475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070096,
+      "loss": 1.0864,
+      "step": 7476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070092,
+      "loss": 1.2341,
+      "step": 7477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070088,
+      "loss": 1.1027,
+      "step": 7478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070084,
+      "loss": 1.1888,
+      "step": 7479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007008,
+      "loss": 1.1118,
+      "step": 7480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070076,
+      "loss": 1.0848,
+      "step": 7481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070072,
+      "loss": 1.165,
+      "step": 7482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070068,
+      "loss": 1.1245,
+      "step": 7483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007006400000000001,
+      "loss": 1.236,
+      "step": 7484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007006,
+      "loss": 1.1631,
+      "step": 7485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070056,
+      "loss": 1.1836,
+      "step": 7486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070052,
+      "loss": 1.0165,
+      "step": 7487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070048,
+      "loss": 1.1833,
+      "step": 7488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070044,
+      "loss": 1.1543,
+      "step": 7489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007004,
+      "loss": 1.1791,
+      "step": 7490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070036,
+      "loss": 1.1682,
+      "step": 7491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070032,
+      "loss": 1.1578,
+      "step": 7492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007002800000000001,
+      "loss": 1.1574,
+      "step": 7493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070024,
+      "loss": 1.2239,
+      "step": 7494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007002000000000001,
+      "loss": 1.205,
+      "step": 7495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070016,
+      "loss": 1.2171,
+      "step": 7496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007001199999999999,
+      "loss": 1.069,
+      "step": 7497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007000800000000001,
+      "loss": 1.1605,
+      "step": 7498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00070004,
+      "loss": 1.2054,
+      "step": 7499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0007,
+      "loss": 1.2504,
+      "step": 7500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069996,
+      "loss": 1.1954,
+      "step": 7501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069992,
+      "loss": 1.1178,
+      "step": 7502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069988,
+      "loss": 1.1887,
+      "step": 7503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006998400000000001,
+      "loss": 1.1375,
+      "step": 7504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006998,
+      "loss": 1.1742,
+      "step": 7505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006997600000000001,
+      "loss": 1.2311,
+      "step": 7506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069972,
+      "loss": 1.1219,
+      "step": 7507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006996799999999999,
+      "loss": 1.1113,
+      "step": 7508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006996400000000001,
+      "loss": 1.1063,
+      "step": 7509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006996,
+      "loss": 1.2277,
+      "step": 7510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069956,
+      "loss": 1.0975,
+      "step": 7511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069952,
+      "loss": 1.1205,
+      "step": 7512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069948,
+      "loss": 1.1482,
+      "step": 7513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069944,
+      "loss": 1.1324,
+      "step": 7514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006994000000000001,
+      "loss": 1.2042,
+      "step": 7515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069936,
+      "loss": 1.2237,
+      "step": 7516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069932,
+      "loss": 1.2061,
+      "step": 7517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069928,
+      "loss": 1.0805,
+      "step": 7518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069924,
+      "loss": 1.1741,
+      "step": 7519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006992000000000001,
+      "loss": 1.162,
+      "step": 7520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069916,
+      "loss": 1.1352,
+      "step": 7521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069912,
+      "loss": 1.1284,
+      "step": 7522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069908,
+      "loss": 1.1588,
+      "step": 7523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006990400000000001,
+      "loss": 1.164,
+      "step": 7524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000699,
+      "loss": 1.0976,
+      "step": 7525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069896,
+      "loss": 1.1843,
+      "step": 7526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069892,
+      "loss": 1.1066,
+      "step": 7527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006988799999999999,
+      "loss": 1.1787,
+      "step": 7528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069884,
+      "loss": 1.1139,
+      "step": 7529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006988,
+      "loss": 1.1133,
+      "step": 7530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006987600000000001,
+      "loss": 1.1525,
+      "step": 7531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069872,
+      "loss": 1.0785,
+      "step": 7532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069868,
+      "loss": 1.169,
+      "step": 7533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069864,
+      "loss": 1.2104,
+      "step": 7534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006986000000000001,
+      "loss": 1.158,
+      "step": 7535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069856,
+      "loss": 1.1001,
+      "step": 7536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069852,
+      "loss": 1.1805,
+      "step": 7537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069848,
+      "loss": 1.1767,
+      "step": 7538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006984399999999999,
+      "loss": 1.2121,
+      "step": 7539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006984000000000001,
+      "loss": 1.1203,
+      "step": 7540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069836,
+      "loss": 1.1239,
+      "step": 7541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006983200000000001,
+      "loss": 1.1348,
+      "step": 7542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069828,
+      "loss": 1.1975,
+      "step": 7543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069824,
+      "loss": 1.1256,
+      "step": 7544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006982000000000001,
+      "loss": 1.1111,
+      "step": 7545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069816,
+      "loss": 1.141,
+      "step": 7546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069812,
+      "loss": 1.067,
+      "step": 7547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069808,
+      "loss": 1.1014,
+      "step": 7548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069804,
+      "loss": 1.2244,
+      "step": 7549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006979999999999999,
+      "loss": 1.2043,
+      "step": 7550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006979600000000001,
+      "loss": 1.2295,
+      "step": 7551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069792,
+      "loss": 1.1038,
+      "step": 7552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069788,
+      "loss": 1.1542,
+      "step": 7553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069784,
+      "loss": 1.1689,
+      "step": 7554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006977999999999999,
+      "loss": 1.1065,
+      "step": 7555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006977600000000001,
+      "loss": 1.0863,
+      "step": 7556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069772,
+      "loss": 1.188,
+      "step": 7557
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069768,
+      "loss": 1.1657,
+      "step": 7558
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069764,
+      "loss": 1.1437,
+      "step": 7559
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006976,
+      "loss": 1.2001,
+      "step": 7560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069756,
+      "loss": 1.1628,
+      "step": 7561
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006975200000000001,
+      "loss": 1.0926,
+      "step": 7562
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069748,
+      "loss": 1.1495,
+      "step": 7563
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069744,
+      "loss": 1.1597,
+      "step": 7564
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006974,
+      "loss": 1.2247,
+      "step": 7565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069736,
+      "loss": 1.1564,
+      "step": 7566
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006973200000000001,
+      "loss": 1.1615,
+      "step": 7567
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069728,
+      "loss": 1.0605,
+      "step": 7568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069724,
+      "loss": 1.1228,
+      "step": 7569
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006972,
+      "loss": 1.1366,
+      "step": 7570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069716,
+      "loss": 1.1593,
+      "step": 7571
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069712,
+      "loss": 1.1919,
+      "step": 7572
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006970800000000001,
+      "loss": 1.1212,
+      "step": 7573
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069704,
+      "loss": 1.2002,
+      "step": 7574
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006969999999999999,
+      "loss": 1.1681,
+      "step": 7575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069696,
+      "loss": 1.113,
+      "step": 7576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069692,
+      "loss": 1.2512,
+      "step": 7577
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006968800000000001,
+      "loss": 1.0945,
+      "step": 7578
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069684,
+      "loss": 1.1312,
+      "step": 7579
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006968,
+      "loss": 1.1427,
+      "step": 7580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069676,
+      "loss": 1.1012,
+      "step": 7581
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006967200000000001,
+      "loss": 1.1408,
+      "step": 7582
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069668,
+      "loss": 1.1308,
+      "step": 7583
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069664,
+      "loss": 1.1537,
+      "step": 7584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006966,
+      "loss": 1.1097,
+      "step": 7585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006965599999999999,
+      "loss": 1.1401,
+      "step": 7586
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006965200000000001,
+      "loss": 1.114,
+      "step": 7587
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069648,
+      "loss": 1.0455,
+      "step": 7588
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069644,
+      "loss": 1.1317,
+      "step": 7589
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006964,
+      "loss": 1.1877,
+      "step": 7590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069636,
+      "loss": 1.1728,
+      "step": 7591
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006963200000000001,
+      "loss": 1.1044,
+      "step": 7592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006962800000000001,
+      "loss": 1.2569,
+      "step": 7593
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069624,
+      "loss": 1.1239,
+      "step": 7594
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006962,
+      "loss": 1.2016,
+      "step": 7595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069616,
+      "loss": 1.1078,
+      "step": 7596
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006961199999999999,
+      "loss": 1.2089,
+      "step": 7597
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006960800000000001,
+      "loss": 1.114,
+      "step": 7598
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069604,
+      "loss": 1.1458,
+      "step": 7599
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000696,
+      "loss": 1.0841,
+      "step": 7600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069596,
+      "loss": 1.1024,
+      "step": 7601
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069592,
+      "loss": 1.2087,
+      "step": 7602
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006958800000000001,
+      "loss": 1.1485,
+      "step": 7603
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069584,
+      "loss": 1.187,
+      "step": 7604
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006958,
+      "loss": 1.2088,
+      "step": 7605
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069576,
+      "loss": 1.203,
+      "step": 7606
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069572,
+      "loss": 1.1396,
+      "step": 7607
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069568,
+      "loss": 1.1046,
+      "step": 7608
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006956400000000001,
+      "loss": 1.0808,
+      "step": 7609
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006956,
+      "loss": 1.1159,
+      "step": 7610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069556,
+      "loss": 1.1108,
+      "step": 7611
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069552,
+      "loss": 1.1353,
+      "step": 7612
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069548,
+      "loss": 1.1874,
+      "step": 7613
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069544,
+      "loss": 1.1842,
+      "step": 7614
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006954,
+      "loss": 1.1319,
+      "step": 7615
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069536,
+      "loss": 1.1722,
+      "step": 7616
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069532,
+      "loss": 1.2658,
+      "step": 7617
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069528,
+      "loss": 1.1459,
+      "step": 7618
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069524,
+      "loss": 1.0516,
+      "step": 7619
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006952000000000001,
+      "loss": 1.1742,
+      "step": 7620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069516,
+      "loss": 1.185,
+      "step": 7621
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069512,
+      "loss": 1.0567,
+      "step": 7622
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069508,
+      "loss": 1.175,
+      "step": 7623
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069504,
+      "loss": 1.1718,
+      "step": 7624
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000695,
+      "loss": 1.1146,
+      "step": 7625
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069496,
+      "loss": 1.1499,
+      "step": 7626
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069492,
+      "loss": 1.075,
+      "step": 7627
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069488,
+      "loss": 1.2068,
+      "step": 7628
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006948400000000001,
+      "loss": 1.1272,
+      "step": 7629
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006948,
+      "loss": 1.0908,
+      "step": 7630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006947600000000001,
+      "loss": 1.1201,
+      "step": 7631
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069472,
+      "loss": 1.2397,
+      "step": 7632
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006946799999999999,
+      "loss": 1.0894,
+      "step": 7633
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006946400000000001,
+      "loss": 1.1234,
+      "step": 7634
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006946,
+      "loss": 1.1735,
+      "step": 7635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069456,
+      "loss": 1.2412,
+      "step": 7636
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069452,
+      "loss": 1.0844,
+      "step": 7637
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069448,
+      "loss": 1.0962,
+      "step": 7638
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006944399999999999,
+      "loss": 1.1452,
+      "step": 7639
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006944000000000001,
+      "loss": 1.0941,
+      "step": 7640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069436,
+      "loss": 1.1169,
+      "step": 7641
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006943200000000001,
+      "loss": 1.1272,
+      "step": 7642
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069428,
+      "loss": 1.1997,
+      "step": 7643
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006942399999999999,
+      "loss": 1.1677,
+      "step": 7644
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006942000000000001,
+      "loss": 1.1484,
+      "step": 7645
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069416,
+      "loss": 1.2321,
+      "step": 7646
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069412,
+      "loss": 1.2091,
+      "step": 7647
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069408,
+      "loss": 1.1815,
+      "step": 7648
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069404,
+      "loss": 1.12,
+      "step": 7649
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000694,
+      "loss": 1.1449,
+      "step": 7650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006939600000000001,
+      "loss": 1.1118,
+      "step": 7651
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069392,
+      "loss": 1.0811,
+      "step": 7652
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069388,
+      "loss": 1.1531,
+      "step": 7653
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069384,
+      "loss": 1.0984,
+      "step": 7654
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006938,
+      "loss": 1.1682,
+      "step": 7655
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006937600000000001,
+      "loss": 1.1528,
+      "step": 7656
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069372,
+      "loss": 1.1767,
+      "step": 7657
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069368,
+      "loss": 1.1336,
+      "step": 7658
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069364,
+      "loss": 1.1061,
+      "step": 7659
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006936,
+      "loss": 1.1617,
+      "step": 7660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069356,
+      "loss": 1.0578,
+      "step": 7661
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069352,
+      "loss": 1.1027,
+      "step": 7662
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069348,
+      "loss": 1.1782,
+      "step": 7663
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006934399999999999,
+      "loss": 1.0774,
+      "step": 7664
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006934,
+      "loss": 1.0632,
+      "step": 7665
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069336,
+      "loss": 1.2447,
+      "step": 7666
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006933200000000001,
+      "loss": 1.1022,
+      "step": 7667
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069328,
+      "loss": 1.1377,
+      "step": 7668
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069324,
+      "loss": 1.1444,
+      "step": 7669
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006932,
+      "loss": 1.1307,
+      "step": 7670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006931600000000001,
+      "loss": 1.1148,
+      "step": 7671
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069312,
+      "loss": 1.2166,
+      "step": 7672
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069308,
+      "loss": 1.209,
+      "step": 7673
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069304,
+      "loss": 1.079,
+      "step": 7674
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006929999999999999,
+      "loss": 1.1551,
+      "step": 7675
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006929600000000001,
+      "loss": 1.1628,
+      "step": 7676
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069292,
+      "loss": 1.1259,
+      "step": 7677
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006928800000000001,
+      "loss": 1.0571,
+      "step": 7678
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069284,
+      "loss": 1.1562,
+      "step": 7679
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006928,
+      "loss": 1.1644,
+      "step": 7680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006927600000000001,
+      "loss": 1.0617,
+      "step": 7681
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069272,
+      "loss": 1.1729,
+      "step": 7682
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069268,
+      "loss": 1.0945,
+      "step": 7683
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069264,
+      "loss": 1.1909,
+      "step": 7684
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006926,
+      "loss": 1.1479,
+      "step": 7685
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006925599999999999,
+      "loss": 1.1761,
+      "step": 7686
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006925200000000001,
+      "loss": 1.1799,
+      "step": 7687
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069248,
+      "loss": 1.1761,
+      "step": 7688
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006924400000000001,
+      "loss": 1.1698,
+      "step": 7689
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006924,
+      "loss": 1.1539,
+      "step": 7690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006923599999999999,
+      "loss": 1.0983,
+      "step": 7691
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006923200000000001,
+      "loss": 1.0932,
+      "step": 7692
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069228,
+      "loss": 1.0726,
+      "step": 7693
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069224,
+      "loss": 1.0712,
+      "step": 7694
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006922,
+      "loss": 1.1307,
+      "step": 7695
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069216,
+      "loss": 1.0884,
+      "step": 7696
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069212,
+      "loss": 1.1764,
+      "step": 7697
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006920800000000001,
+      "loss": 1.1514,
+      "step": 7698
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069204,
+      "loss": 1.0608,
+      "step": 7699
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000692,
+      "loss": 1.1924,
+      "step": 7700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069196,
+      "loss": 1.073,
+      "step": 7701
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069192,
+      "loss": 1.0945,
+      "step": 7702
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006918800000000001,
+      "loss": 1.1389,
+      "step": 7703
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069184,
+      "loss": 1.0883,
+      "step": 7704
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006918,
+      "loss": 1.1113,
+      "step": 7705
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069176,
+      "loss": 1.1002,
+      "step": 7706
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069172,
+      "loss": 1.1343,
+      "step": 7707
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069168,
+      "loss": 1.1654,
+      "step": 7708
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006916400000000001,
+      "loss": 1.1383,
+      "step": 7709
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006916,
+      "loss": 1.1401,
+      "step": 7710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006915599999999999,
+      "loss": 1.0745,
+      "step": 7711
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069152,
+      "loss": 1.1478,
+      "step": 7712
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069148,
+      "loss": 1.1422,
+      "step": 7713
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006914400000000001,
+      "loss": 1.1891,
+      "step": 7714
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006914,
+      "loss": 1.1146,
+      "step": 7715
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069136,
+      "loss": 1.2346,
+      "step": 7716
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069132,
+      "loss": 1.1533,
+      "step": 7717
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006912800000000001,
+      "loss": 1.1466,
+      "step": 7718
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069124,
+      "loss": 1.1255,
+      "step": 7719
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006912000000000001,
+      "loss": 1.089,
+      "step": 7720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069116,
+      "loss": 1.1861,
+      "step": 7721
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006911199999999999,
+      "loss": 1.1927,
+      "step": 7722
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0006910800000000001,
+      "loss": 1.1517,
+      "step": 7723
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069104,
+      "loss": 1.1822,
+      "step": 7724
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000691,
+      "loss": 1.0947,
+      "step": 7725
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069096,
+      "loss": 1.1011,
+      "step": 7726
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069092,
+      "loss": 1.1937,
+      "step": 7727
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00069088,
+      "loss": 1.1303,
+      "step": 7728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006908400000000001,
+      "loss": 1.1289,
+      "step": 7729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006908,
+      "loss": 1.0884,
+      "step": 7730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069076,
+      "loss": 1.0942,
+      "step": 7731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069072,
+      "loss": 1.2093,
+      "step": 7732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006906799999999999,
+      "loss": 1.2485,
+      "step": 7733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006906400000000001,
+      "loss": 1.0485,
+      "step": 7734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006906,
+      "loss": 1.0691,
+      "step": 7735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069056,
+      "loss": 1.1177,
+      "step": 7736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069052,
+      "loss": 1.1065,
+      "step": 7737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069048,
+      "loss": 1.1197,
+      "step": 7738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006904400000000001,
+      "loss": 1.0884,
+      "step": 7739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006904,
+      "loss": 1.1534,
+      "step": 7740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069036,
+      "loss": 1.2579,
+      "step": 7741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069032,
+      "loss": 1.1753,
+      "step": 7742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069028,
+      "loss": 1.1823,
+      "step": 7743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069024,
+      "loss": 1.1112,
+      "step": 7744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006902000000000001,
+      "loss": 1.085,
+      "step": 7745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069016,
+      "loss": 1.2453,
+      "step": 7746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069012,
+      "loss": 1.1725,
+      "step": 7747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069008,
+      "loss": 1.0584,
+      "step": 7748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006900400000000001,
+      "loss": 1.1696,
+      "step": 7749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00069,
+      "loss": 1.2064,
+      "step": 7750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068996,
+      "loss": 1.1508,
+      "step": 7751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068992,
+      "loss": 1.192,
+      "step": 7752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068988,
+      "loss": 1.1507,
+      "step": 7753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068984,
+      "loss": 1.1556,
+      "step": 7754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006898,
+      "loss": 1.0898,
+      "step": 7755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006897600000000001,
+      "loss": 1.1591,
+      "step": 7756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068972,
+      "loss": 1.1223,
+      "step": 7757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068968,
+      "loss": 1.2069,
+      "step": 7758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068964,
+      "loss": 1.1323,
+      "step": 7759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006896,
+      "loss": 1.1928,
+      "step": 7760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068956,
+      "loss": 1.1672,
+      "step": 7761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068952,
+      "loss": 1.1976,
+      "step": 7762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068948,
+      "loss": 1.1524,
+      "step": 7763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068944,
+      "loss": 1.1859,
+      "step": 7764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006894000000000001,
+      "loss": 1.1287,
+      "step": 7765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068936,
+      "loss": 1.2027,
+      "step": 7766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006893200000000001,
+      "loss": 1.2051,
+      "step": 7767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068928,
+      "loss": 1.0925,
+      "step": 7768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006892399999999999,
+      "loss": 1.1439,
+      "step": 7769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006892000000000001,
+      "loss": 1.1759,
+      "step": 7770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068916,
+      "loss": 1.1859,
+      "step": 7771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068912,
+      "loss": 1.1004,
+      "step": 7772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068908,
+      "loss": 1.1198,
+      "step": 7773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068904,
+      "loss": 1.2647,
+      "step": 7774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006889999999999999,
+      "loss": 1.1019,
+      "step": 7775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006889600000000001,
+      "loss": 1.1795,
+      "step": 7776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068892,
+      "loss": 1.0961,
+      "step": 7777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006888800000000001,
+      "loss": 1.0703,
+      "step": 7778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068884,
+      "loss": 1.0984,
+      "step": 7779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006887999999999999,
+      "loss": 1.1422,
+      "step": 7780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006887600000000001,
+      "loss": 1.095,
+      "step": 7781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068872,
+      "loss": 1.1384,
+      "step": 7782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068868,
+      "loss": 1.1787,
+      "step": 7783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068864,
+      "loss": 1.2058,
+      "step": 7784
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006886,
+      "loss": 1.1955,
+      "step": 7785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068856,
+      "loss": 1.0934,
+      "step": 7786
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006885200000000001,
+      "loss": 1.1473,
+      "step": 7787
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068848,
+      "loss": 1.2316,
+      "step": 7788
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068844,
+      "loss": 1.1263,
+      "step": 7789
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006884,
+      "loss": 1.193,
+      "step": 7790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068836,
+      "loss": 1.1627,
+      "step": 7791
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006883200000000001,
+      "loss": 1.144,
+      "step": 7792
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068828,
+      "loss": 1.1247,
+      "step": 7793
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068824,
+      "loss": 1.0585,
+      "step": 7794
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006882,
+      "loss": 1.1914,
+      "step": 7795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068816,
+      "loss": 1.1727,
+      "step": 7796
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068812,
+      "loss": 1.1846,
+      "step": 7797
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006880800000000001,
+      "loss": 1.1215,
+      "step": 7798
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068804,
+      "loss": 1.1048,
+      "step": 7799
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006879999999999999,
+      "loss": 1.2211,
+      "step": 7800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068796,
+      "loss": 1.2293,
+      "step": 7801
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068792,
+      "loss": 1.2187,
+      "step": 7802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006878800000000001,
+      "loss": 1.0891,
+      "step": 7803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068784,
+      "loss": 1.1464,
+      "step": 7804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006878,
+      "loss": 1.1965,
+      "step": 7805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068776,
+      "loss": 1.1394,
+      "step": 7806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006877200000000001,
+      "loss": 1.1338,
+      "step": 7807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068768,
+      "loss": 1.1533,
+      "step": 7808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068764,
+      "loss": 1.2682,
+      "step": 7809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006876,
+      "loss": 1.2051,
+      "step": 7810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006875599999999999,
+      "loss": 1.2132,
+      "step": 7811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006875200000000001,
+      "loss": 1.0202,
+      "step": 7812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068748,
+      "loss": 1.0907,
+      "step": 7813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006874400000000001,
+      "loss": 1.186,
+      "step": 7814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006874,
+      "loss": 1.1443,
+      "step": 7815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068736,
+      "loss": 1.1263,
+      "step": 7816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068732,
+      "loss": 1.1955,
+      "step": 7817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068728,
+      "loss": 1.162,
+      "step": 7818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068724,
+      "loss": 1.1409,
+      "step": 7819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006872,
+      "loss": 1.1486,
+      "step": 7820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068716,
+      "loss": 1.1302,
+      "step": 7821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006871199999999999,
+      "loss": 1.1338,
+      "step": 7822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006870800000000001,
+      "loss": 1.1003,
+      "step": 7823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068704,
+      "loss": 1.0604,
+      "step": 7824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006870000000000001,
+      "loss": 1.2807,
+      "step": 7825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068696,
+      "loss": 1.0951,
+      "step": 7826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006869199999999999,
+      "loss": 1.128,
+      "step": 7827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006868800000000001,
+      "loss": 1.1128,
+      "step": 7828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068684,
+      "loss": 1.1623,
+      "step": 7829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006868,
+      "loss": 1.0531,
+      "step": 7830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068676,
+      "loss": 1.1937,
+      "step": 7831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068672,
+      "loss": 1.132,
+      "step": 7832
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068668,
+      "loss": 1.16,
+      "step": 7833
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006866400000000001,
+      "loss": 1.1669,
+      "step": 7834
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006866,
+      "loss": 1.1068,
+      "step": 7835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068656,
+      "loss": 1.1682,
+      "step": 7836
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068652,
+      "loss": 1.0817,
+      "step": 7837
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068648,
+      "loss": 1.1232,
+      "step": 7838
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006864400000000001,
+      "loss": 1.1321,
+      "step": 7839
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006864,
+      "loss": 1.1682,
+      "step": 7840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068636,
+      "loss": 1.1484,
+      "step": 7841
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068632,
+      "loss": 1.1642,
+      "step": 7842
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068628,
+      "loss": 1.1475,
+      "step": 7843
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068624,
+      "loss": 1.1565,
+      "step": 7844
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006862000000000001,
+      "loss": 1.1489,
+      "step": 7845
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068616,
+      "loss": 1.081,
+      "step": 7846
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006861199999999999,
+      "loss": 1.0797,
+      "step": 7847
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068608,
+      "loss": 1.1129,
+      "step": 7848
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068604,
+      "loss": 1.1348,
+      "step": 7849
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006860000000000001,
+      "loss": 1.148,
+      "step": 7850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068596,
+      "loss": 1.2342,
+      "step": 7851
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068592,
+      "loss": 1.1265,
+      "step": 7852
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068588,
+      "loss": 1.0565,
+      "step": 7853
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006858400000000001,
+      "loss": 1.1205,
+      "step": 7854
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006858,
+      "loss": 1.1951,
+      "step": 7855
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006857600000000001,
+      "loss": 1.133,
+      "step": 7856
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068572,
+      "loss": 1.1706,
+      "step": 7857
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006856799999999999,
+      "loss": 1.1056,
+      "step": 7858
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006856400000000001,
+      "loss": 1.1616,
+      "step": 7859
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006856,
+      "loss": 1.1347,
+      "step": 7860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068556,
+      "loss": 1.2407,
+      "step": 7861
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068552,
+      "loss": 1.2268,
+      "step": 7862
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068548,
+      "loss": 1.0538,
+      "step": 7863
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068544,
+      "loss": 1.1653,
+      "step": 7864
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006854000000000001,
+      "loss": 1.0997,
+      "step": 7865
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068536,
+      "loss": 1.0776,
+      "step": 7866
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068532,
+      "loss": 1.2206,
+      "step": 7867
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068528,
+      "loss": 1.1441,
+      "step": 7868
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006852399999999999,
+      "loss": 1.0797,
+      "step": 7869
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006852000000000001,
+      "loss": 1.1755,
+      "step": 7870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068516,
+      "loss": 1.1214,
+      "step": 7871
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068512,
+      "loss": 1.1579,
+      "step": 7872
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068508,
+      "loss": 1.1503,
+      "step": 7873
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068504,
+      "loss": 1.1216,
+      "step": 7874
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006850000000000001,
+      "loss": 1.1648,
+      "step": 7875
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068496,
+      "loss": 1.1399,
+      "step": 7876
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068492,
+      "loss": 1.1191,
+      "step": 7877
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068488,
+      "loss": 1.1296,
+      "step": 7878
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068484,
+      "loss": 1.0864,
+      "step": 7879
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006848,
+      "loss": 1.1155,
+      "step": 7880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006847600000000001,
+      "loss": 1.2213,
+      "step": 7881
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068472,
+      "loss": 1.1121,
+      "step": 7882
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068468,
+      "loss": 1.1174,
+      "step": 7883
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068464,
+      "loss": 1.1723,
+      "step": 7884
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006846,
+      "loss": 1.1483,
+      "step": 7885
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068456,
+      "loss": 1.1685,
+      "step": 7886
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068452,
+      "loss": 1.1387,
+      "step": 7887
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068448,
+      "loss": 1.157,
+      "step": 7888
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068444,
+      "loss": 1.1709,
+      "step": 7889
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006844,
+      "loss": 1.038,
+      "step": 7890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068436,
+      "loss": 1.1397,
+      "step": 7891
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006843200000000001,
+      "loss": 1.1651,
+      "step": 7892
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068428,
+      "loss": 1.1003,
+      "step": 7893
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068424,
+      "loss": 1.1897,
+      "step": 7894
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006842,
+      "loss": 1.1893,
+      "step": 7895
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068416,
+      "loss": 1.0754,
+      "step": 7896
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068412,
+      "loss": 1.1809,
+      "step": 7897
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068408,
+      "loss": 1.1466,
+      "step": 7898
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068404,
+      "loss": 1.2108,
+      "step": 7899
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000684,
+      "loss": 1.1323,
+      "step": 7900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006839600000000001,
+      "loss": 1.0987,
+      "step": 7901
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068392,
+      "loss": 1.1754,
+      "step": 7902
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006838800000000001,
+      "loss": 1.1733,
+      "step": 7903
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068384,
+      "loss": 1.155,
+      "step": 7904
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006837999999999999,
+      "loss": 1.1626,
+      "step": 7905
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068376,
+      "loss": 1.1475,
+      "step": 7906
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068372,
+      "loss": 1.1818,
+      "step": 7907
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068368,
+      "loss": 1.1199,
+      "step": 7908
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068364,
+      "loss": 1.204,
+      "step": 7909
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006836,
+      "loss": 1.198,
+      "step": 7910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006835599999999999,
+      "loss": 1.1735,
+      "step": 7911
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006835200000000001,
+      "loss": 1.1736,
+      "step": 7912
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068348,
+      "loss": 1.1459,
+      "step": 7913
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006834400000000001,
+      "loss": 1.1203,
+      "step": 7914
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006834,
+      "loss": 1.1712,
+      "step": 7915
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006833599999999999,
+      "loss": 1.1255,
+      "step": 7916
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006833200000000001,
+      "loss": 1.1392,
+      "step": 7917
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068328,
+      "loss": 1.1645,
+      "step": 7918
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068324,
+      "loss": 1.2535,
+      "step": 7919
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006832,
+      "loss": 1.1044,
+      "step": 7920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068316,
+      "loss": 1.1358,
+      "step": 7921
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068312,
+      "loss": 1.1198,
+      "step": 7922
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006830800000000001,
+      "loss": 1.1426,
+      "step": 7923
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068304,
+      "loss": 1.1023,
+      "step": 7924
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000683,
+      "loss": 1.144,
+      "step": 7925
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068296,
+      "loss": 1.1373,
+      "step": 7926
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068292,
+      "loss": 1.1048,
+      "step": 7927
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006828800000000001,
+      "loss": 1.1704,
+      "step": 7928
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068284,
+      "loss": 1.1759,
+      "step": 7929
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006828,
+      "loss": 1.1271,
+      "step": 7930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068276,
+      "loss": 1.2444,
+      "step": 7931
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068272,
+      "loss": 1.0439,
+      "step": 7932
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068268,
+      "loss": 1.1156,
+      "step": 7933
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006826400000000001,
+      "loss": 1.1124,
+      "step": 7934
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006826,
+      "loss": 1.1548,
+      "step": 7935
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006825599999999999,
+      "loss": 1.0538,
+      "step": 7936
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068252,
+      "loss": 1.1132,
+      "step": 7937
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068248,
+      "loss": 1.0657,
+      "step": 7938
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006824400000000001,
+      "loss": 1.1975,
+      "step": 7939
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006824,
+      "loss": 1.0683,
+      "step": 7940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068236,
+      "loss": 1.1494,
+      "step": 7941
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068232,
+      "loss": 1.1615,
+      "step": 7942
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006822800000000001,
+      "loss": 1.1289,
+      "step": 7943
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068224,
+      "loss": 1.2462,
+      "step": 7944
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006822,
+      "loss": 1.2039,
+      "step": 7945
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068216,
+      "loss": 1.1276,
+      "step": 7946
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006821199999999999,
+      "loss": 1.0951,
+      "step": 7947
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006820800000000001,
+      "loss": 1.1742,
+      "step": 7948
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068204,
+      "loss": 1.1156,
+      "step": 7949
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006820000000000001,
+      "loss": 1.0672,
+      "step": 7950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068196,
+      "loss": 1.0948,
+      "step": 7951
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068192,
+      "loss": 1.1508,
+      "step": 7952
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068188,
+      "loss": 1.1598,
+      "step": 7953
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068184,
+      "loss": 1.1871,
+      "step": 7954
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006818,
+      "loss": 1.1038,
+      "step": 7955
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068176,
+      "loss": 1.095,
+      "step": 7956
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068172,
+      "loss": 1.1763,
+      "step": 7957
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006816799999999999,
+      "loss": 1.2029,
+      "step": 7958
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006816400000000001,
+      "loss": 1.1645,
+      "step": 7959
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006816,
+      "loss": 1.1478,
+      "step": 7960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006815600000000001,
+      "loss": 1.1116,
+      "step": 7961
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068152,
+      "loss": 1.0744,
+      "step": 7962
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068148,
+      "loss": 1.123,
+      "step": 7963
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006814400000000001,
+      "loss": 1.1613,
+      "step": 7964
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006814,
+      "loss": 1.0535,
+      "step": 7965
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068136,
+      "loss": 1.166,
+      "step": 7966
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068132,
+      "loss": 1.1299,
+      "step": 7967
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068128,
+      "loss": 0.9989,
+      "step": 7968
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068124,
+      "loss": 1.1855,
+      "step": 7969
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006812000000000001,
+      "loss": 1.1102,
+      "step": 7970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068116,
+      "loss": 1.0941,
+      "step": 7971
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068112,
+      "loss": 1.1827,
+      "step": 7972
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068108,
+      "loss": 1.0946,
+      "step": 7973
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006810399999999999,
+      "loss": 1.1967,
+      "step": 7974
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006810000000000001,
+      "loss": 1.073,
+      "step": 7975
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068096,
+      "loss": 1.2164,
+      "step": 7976
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068092,
+      "loss": 1.1991,
+      "step": 7977
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068088,
+      "loss": 1.0825,
+      "step": 7978
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068084,
+      "loss": 1.0674,
+      "step": 7979
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006808,
+      "loss": 1.1591,
+      "step": 7980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0006807600000000001,
+      "loss": 1.1518,
+      "step": 7981
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00068072,
+      "loss": 1.0906,
+      "step": 7982
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006806799999999999,
+      "loss": 1.1568,
+      "step": 7983
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068064,
+      "loss": 1.1203,
+      "step": 7984
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006806,
+      "loss": 1.0626,
+      "step": 7985
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006805600000000001,
+      "loss": 1.132,
+      "step": 7986
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068052,
+      "loss": 1.1097,
+      "step": 7987
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068048,
+      "loss": 1.1034,
+      "step": 7988
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068044,
+      "loss": 1.1458,
+      "step": 7989
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006804000000000001,
+      "loss": 1.0464,
+      "step": 7990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068036,
+      "loss": 1.1427,
+      "step": 7991
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006803200000000001,
+      "loss": 1.1427,
+      "step": 7992
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068028,
+      "loss": 1.0518,
+      "step": 7993
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006802399999999999,
+      "loss": 1.1007,
+      "step": 7994
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006802,
+      "loss": 1.1365,
+      "step": 7995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068016,
+      "loss": 1.1622,
+      "step": 7996
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068012,
+      "loss": 1.1721,
+      "step": 7997
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068008,
+      "loss": 1.1602,
+      "step": 7998
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068004,
+      "loss": 1.1959,
+      "step": 7999
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00068,
+      "loss": 1.0727,
+      "step": 8000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006799600000000001,
+      "loss": 1.1153,
+      "step": 8001
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067992,
+      "loss": 1.0588,
+      "step": 8002
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067988,
+      "loss": 1.1643,
+      "step": 8003
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067984,
+      "loss": 1.2034,
+      "step": 8004
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006797999999999999,
+      "loss": 1.093,
+      "step": 8005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006797600000000001,
+      "loss": 1.1378,
+      "step": 8006
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067972,
+      "loss": 1.1001,
+      "step": 8007
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067968,
+      "loss": 1.1447,
+      "step": 8008
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067964,
+      "loss": 1.1398,
+      "step": 8009
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006796,
+      "loss": 0.9893,
+      "step": 8010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006795600000000001,
+      "loss": 1.2386,
+      "step": 8011
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006795200000000001,
+      "loss": 1.2091,
+      "step": 8012
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067948,
+      "loss": 1.0238,
+      "step": 8013
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067944,
+      "loss": 1.1077,
+      "step": 8014
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006794,
+      "loss": 1.1847,
+      "step": 8015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067936,
+      "loss": 1.0599,
+      "step": 8016
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006793200000000001,
+      "loss": 1.1476,
+      "step": 8017
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067928,
+      "loss": 1.0947,
+      "step": 8018
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067924,
+      "loss": 1.058,
+      "step": 8019
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006792,
+      "loss": 1.2121,
+      "step": 8020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067916,
+      "loss": 1.1548,
+      "step": 8021
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067912,
+      "loss": 1.144,
+      "step": 8022
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067908,
+      "loss": 1.0805,
+      "step": 8023
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067904,
+      "loss": 1.1706,
+      "step": 8024
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000679,
+      "loss": 1.1092,
+      "step": 8025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067896,
+      "loss": 1.1989,
+      "step": 8026
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067892,
+      "loss": 1.2158,
+      "step": 8027
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006788800000000001,
+      "loss": 1.2376,
+      "step": 8028
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067884,
+      "loss": 1.18,
+      "step": 8029
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006788,
+      "loss": 1.1388,
+      "step": 8030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067876,
+      "loss": 1.1671,
+      "step": 8031
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067872,
+      "loss": 1.1435,
+      "step": 8032
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067868,
+      "loss": 1.1947,
+      "step": 8033
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067864,
+      "loss": 1.1681,
+      "step": 8034
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006786,
+      "loss": 1.1507,
+      "step": 8035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067856,
+      "loss": 1.117,
+      "step": 8036
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006785200000000001,
+      "loss": 1.1249,
+      "step": 8037
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067848,
+      "loss": 1.143,
+      "step": 8038
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006784400000000001,
+      "loss": 1.153,
+      "step": 8039
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006784,
+      "loss": 1.0776,
+      "step": 8040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067836,
+      "loss": 1.1682,
+      "step": 8041
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067832,
+      "loss": 1.0966,
+      "step": 8042
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067828,
+      "loss": 1.1213,
+      "step": 8043
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067824,
+      "loss": 1.0961,
+      "step": 8044
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006782,
+      "loss": 1.1623,
+      "step": 8045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067816,
+      "loss": 1.1685,
+      "step": 8046
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006781199999999999,
+      "loss": 1.1611,
+      "step": 8047
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006780800000000001,
+      "loss": 1.1505,
+      "step": 8048
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067804,
+      "loss": 1.1128,
+      "step": 8049
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006780000000000001,
+      "loss": 1.1907,
+      "step": 8050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067796,
+      "loss": 1.0992,
+      "step": 8051
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006779199999999999,
+      "loss": 1.1826,
+      "step": 8052
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006778800000000001,
+      "loss": 1.1449,
+      "step": 8053
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067784,
+      "loss": 1.1811,
+      "step": 8054
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006778,
+      "loss": 1.1782,
+      "step": 8055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067776,
+      "loss": 1.1038,
+      "step": 8056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067772,
+      "loss": 1.1311,
+      "step": 8057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067768,
+      "loss": 1.1524,
+      "step": 8058
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006776400000000001,
+      "loss": 1.1069,
+      "step": 8059
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006776,
+      "loss": 1.1232,
+      "step": 8060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067756,
+      "loss": 1.1333,
+      "step": 8061
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067752,
+      "loss": 1.1849,
+      "step": 8062
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006774799999999999,
+      "loss": 1.1993,
+      "step": 8063
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006774400000000001,
+      "loss": 1.1183,
+      "step": 8064
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006774,
+      "loss": 1.1843,
+      "step": 8065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067736,
+      "loss": 1.1725,
+      "step": 8066
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067732,
+      "loss": 1.1441,
+      "step": 8067
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067728,
+      "loss": 1.118,
+      "step": 8068
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067724,
+      "loss": 1.1574,
+      "step": 8069
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006772000000000001,
+      "loss": 1.1217,
+      "step": 8070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067716,
+      "loss": 1.131,
+      "step": 8071
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067712,
+      "loss": 1.1324,
+      "step": 8072
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067708,
+      "loss": 1.1021,
+      "step": 8073
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067704,
+      "loss": 1.1619,
+      "step": 8074
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006770000000000001,
+      "loss": 1.193,
+      "step": 8075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067696,
+      "loss": 1.1766,
+      "step": 8076
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067692,
+      "loss": 1.1798,
+      "step": 8077
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067688,
+      "loss": 1.1226,
+      "step": 8078
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006768400000000001,
+      "loss": 1.12,
+      "step": 8079
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006768,
+      "loss": 1.058,
+      "step": 8080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067676,
+      "loss": 1.1466,
+      "step": 8081
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067672,
+      "loss": 1.116,
+      "step": 8082
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006766799999999999,
+      "loss": 1.2295,
+      "step": 8083
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067664,
+      "loss": 1.0673,
+      "step": 8084
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006766,
+      "loss": 1.1321,
+      "step": 8085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006765600000000001,
+      "loss": 1.1472,
+      "step": 8086
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067652,
+      "loss": 1.1066,
+      "step": 8087
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067648,
+      "loss": 1.1043,
+      "step": 8088
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067644,
+      "loss": 1.0929,
+      "step": 8089
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006764,
+      "loss": 1.1567,
+      "step": 8090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067636,
+      "loss": 1.1438,
+      "step": 8091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067632,
+      "loss": 1.053,
+      "step": 8092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067628,
+      "loss": 1.1049,
+      "step": 8093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006762399999999999,
+      "loss": 1.1661,
+      "step": 8094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006762000000000001,
+      "loss": 1.1064,
+      "step": 8095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067616,
+      "loss": 1.1243,
+      "step": 8096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006761200000000001,
+      "loss": 1.2406,
+      "step": 8097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067608,
+      "loss": 1.1729,
+      "step": 8098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067604,
+      "loss": 1.1822,
+      "step": 8099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006760000000000001,
+      "loss": 1.1954,
+      "step": 8100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067596,
+      "loss": 1.1758,
+      "step": 8101
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067592,
+      "loss": 1.1818,
+      "step": 8102
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067588,
+      "loss": 1.1717,
+      "step": 8103
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067584,
+      "loss": 1.031,
+      "step": 8104
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006758,
+      "loss": 1.1994,
+      "step": 8105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006757600000000001,
+      "loss": 1.1235,
+      "step": 8106
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067572,
+      "loss": 1.1437,
+      "step": 8107
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067568,
+      "loss": 1.1213,
+      "step": 8108
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067564,
+      "loss": 1.1501,
+      "step": 8109
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006755999999999999,
+      "loss": 1.1812,
+      "step": 8110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006755600000000001,
+      "loss": 1.145,
+      "step": 8111
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067552,
+      "loss": 1.2146,
+      "step": 8112
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067548,
+      "loss": 1.1017,
+      "step": 8113
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067544,
+      "loss": 1.1209,
+      "step": 8114
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006754,
+      "loss": 1.1032,
+      "step": 8115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067536,
+      "loss": 1.0971,
+      "step": 8116
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006753200000000001,
+      "loss": 1.1315,
+      "step": 8117
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067528,
+      "loss": 1.1296,
+      "step": 8118
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006752399999999999,
+      "loss": 1.1359,
+      "step": 8119
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006752,
+      "loss": 1.1517,
+      "step": 8120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067516,
+      "loss": 1.1224,
+      "step": 8121
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006751200000000001,
+      "loss": 1.2114,
+      "step": 8122
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067508,
+      "loss": 1.1085,
+      "step": 8123
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067504,
+      "loss": 1.0889,
+      "step": 8124
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000675,
+      "loss": 1.1697,
+      "step": 8125
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006749600000000001,
+      "loss": 1.1096,
+      "step": 8126
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067492,
+      "loss": 1.1771,
+      "step": 8127
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006748800000000001,
+      "loss": 1.1675,
+      "step": 8128
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067484,
+      "loss": 1.1393,
+      "step": 8129
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006747999999999999,
+      "loss": 1.2182,
+      "step": 8130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067476,
+      "loss": 1.1293,
+      "step": 8131
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067472,
+      "loss": 1.12,
+      "step": 8132
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067468,
+      "loss": 1.0387,
+      "step": 8133
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067464,
+      "loss": 1.2053,
+      "step": 8134
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006746,
+      "loss": 1.1258,
+      "step": 8135
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067456,
+      "loss": 1.1583,
+      "step": 8136
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006745200000000001,
+      "loss": 1.0562,
+      "step": 8137
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067448,
+      "loss": 1.0714,
+      "step": 8138
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067444,
+      "loss": 1.1793,
+      "step": 8139
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006744,
+      "loss": 1.1561,
+      "step": 8140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006743599999999999,
+      "loss": 1.1359,
+      "step": 8141
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006743200000000001,
+      "loss": 1.226,
+      "step": 8142
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067428,
+      "loss": 1.163,
+      "step": 8143
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067424,
+      "loss": 1.1655,
+      "step": 8144
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006742,
+      "loss": 1.0715,
+      "step": 8145
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067416,
+      "loss": 1.1729,
+      "step": 8146
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006741200000000001,
+      "loss": 1.1295,
+      "step": 8147
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006740800000000001,
+      "loss": 1.0582,
+      "step": 8148
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067404,
+      "loss": 1.1503,
+      "step": 8149
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000674,
+      "loss": 1.1985,
+      "step": 8150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067396,
+      "loss": 1.2002,
+      "step": 8151
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006739199999999999,
+      "loss": 1.1395,
+      "step": 8152
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006738800000000001,
+      "loss": 1.1742,
+      "step": 8153
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067384,
+      "loss": 1.1005,
+      "step": 8154
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006738,
+      "loss": 1.1647,
+      "step": 8155
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067376,
+      "loss": 1.1655,
+      "step": 8156
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067372,
+      "loss": 1.1508,
+      "step": 8157
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067368,
+      "loss": 1.145,
+      "step": 8158
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067364,
+      "loss": 1.1169,
+      "step": 8159
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006736,
+      "loss": 1.1711,
+      "step": 8160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067356,
+      "loss": 1.075,
+      "step": 8161
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067352,
+      "loss": 1.1268,
+      "step": 8162
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067348,
+      "loss": 1.1965,
+      "step": 8163
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006734400000000001,
+      "loss": 1.1504,
+      "step": 8164
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006734,
+      "loss": 1.1306,
+      "step": 8165
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067336,
+      "loss": 1.2019,
+      "step": 8166
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067332,
+      "loss": 1.1666,
+      "step": 8167
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067328,
+      "loss": 1.1249,
+      "step": 8168
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067324,
+      "loss": 1.2044,
+      "step": 8169
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006732,
+      "loss": 1.0356,
+      "step": 8170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067316,
+      "loss": 1.0914,
+      "step": 8171
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067312,
+      "loss": 1.0863,
+      "step": 8172
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067308,
+      "loss": 1.1429,
+      "step": 8173
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067304,
+      "loss": 1.2193,
+      "step": 8174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006730000000000001,
+      "loss": 1.1265,
+      "step": 8175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067296,
+      "loss": 1.1551,
+      "step": 8176
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067292,
+      "loss": 1.2032,
+      "step": 8177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067288,
+      "loss": 1.1358,
+      "step": 8178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067284,
+      "loss": 1.1389,
+      "step": 8179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006728,
+      "loss": 1.1698,
+      "step": 8180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067276,
+      "loss": 1.2016,
+      "step": 8181
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067272,
+      "loss": 1.1346,
+      "step": 8182
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006726799999999999,
+      "loss": 1.0654,
+      "step": 8183
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006726400000000001,
+      "loss": 1.1537,
+      "step": 8184
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006726,
+      "loss": 1.1372,
+      "step": 8185
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006725600000000001,
+      "loss": 1.1154,
+      "step": 8186
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067252,
+      "loss": 1.094,
+      "step": 8187
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006724799999999999,
+      "loss": 1.1408,
+      "step": 8188
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006724400000000001,
+      "loss": 1.1094,
+      "step": 8189
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006724,
+      "loss": 1.1615,
+      "step": 8190
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067236,
+      "loss": 1.0786,
+      "step": 8191
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067232,
+      "loss": 1.0952,
+      "step": 8192
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067228,
+      "loss": 1.1226,
+      "step": 8193
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067224,
+      "loss": 1.1597,
+      "step": 8194
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006722000000000001,
+      "loss": 1.0972,
+      "step": 8195
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067216,
+      "loss": 1.0192,
+      "step": 8196
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006721200000000001,
+      "loss": 1.2054,
+      "step": 8197
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067208,
+      "loss": 1.0868,
+      "step": 8198
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006720399999999999,
+      "loss": 1.0739,
+      "step": 8199
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006720000000000001,
+      "loss": 1.1059,
+      "step": 8200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067196,
+      "loss": 1.1253,
+      "step": 8201
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067192,
+      "loss": 1.128,
+      "step": 8202
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067188,
+      "loss": 1.1509,
+      "step": 8203
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067184,
+      "loss": 1.1527,
+      "step": 8204
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006718,
+      "loss": 1.1291,
+      "step": 8205
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006717600000000001,
+      "loss": 1.1461,
+      "step": 8206
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067172,
+      "loss": 1.2301,
+      "step": 8207
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067168,
+      "loss": 1.1473,
+      "step": 8208
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067164,
+      "loss": 1.0384,
+      "step": 8209
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006716,
+      "loss": 1.1195,
+      "step": 8210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006715600000000001,
+      "loss": 1.168,
+      "step": 8211
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067152,
+      "loss": 1.0305,
+      "step": 8212
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067148,
+      "loss": 1.1033,
+      "step": 8213
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067144,
+      "loss": 1.0234,
+      "step": 8214
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006714000000000001,
+      "loss": 1.1495,
+      "step": 8215
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067136,
+      "loss": 1.0851,
+      "step": 8216
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067132,
+      "loss": 1.0974,
+      "step": 8217
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067128,
+      "loss": 1.1088,
+      "step": 8218
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006712399999999999,
+      "loss": 1.1371,
+      "step": 8219
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006712,
+      "loss": 1.0914,
+      "step": 8220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067116,
+      "loss": 1.0913,
+      "step": 8221
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006711200000000001,
+      "loss": 1.1017,
+      "step": 8222
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067108,
+      "loss": 1.1759,
+      "step": 8223
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067104,
+      "loss": 1.1104,
+      "step": 8224
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000671,
+      "loss": 1.1577,
+      "step": 8225
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006709600000000001,
+      "loss": 1.0904,
+      "step": 8226
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067092,
+      "loss": 1.1829,
+      "step": 8227
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067088,
+      "loss": 1.1742,
+      "step": 8228
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067084,
+      "loss": 1.1552,
+      "step": 8229
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006707999999999999,
+      "loss": 1.1617,
+      "step": 8230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006707600000000001,
+      "loss": 1.1019,
+      "step": 8231
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067072,
+      "loss": 1.1827,
+      "step": 8232
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006706800000000001,
+      "loss": 1.1163,
+      "step": 8233
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00067064,
+      "loss": 1.0605,
+      "step": 8234
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0006706,
+      "loss": 1.1144,
+      "step": 8235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006705600000000001,
+      "loss": 1.1215,
+      "step": 8236
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00067052,
+      "loss": 1.1192,
+      "step": 8237
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00067048,
+      "loss": 1.0691,
+      "step": 8238
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00067044,
+      "loss": 1.0711,
+      "step": 8239
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006704,
+      "loss": 1.1359,
+      "step": 8240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006703599999999999,
+      "loss": 1.0229,
+      "step": 8241
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006703200000000001,
+      "loss": 1.1484,
+      "step": 8242
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00067028,
+      "loss": 1.1287,
+      "step": 8243
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00067024,
+      "loss": 1.1305,
+      "step": 8244
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006702,
+      "loss": 1.193,
+      "step": 8245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006701599999999999,
+      "loss": 1.1715,
+      "step": 8246
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006701200000000001,
+      "loss": 1.1149,
+      "step": 8247
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00067008,
+      "loss": 1.1112,
+      "step": 8248
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00067004,
+      "loss": 1.2418,
+      "step": 8249
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00067,
+      "loss": 1.1648,
+      "step": 8250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066996,
+      "loss": 1.1289,
+      "step": 8251
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066992,
+      "loss": 1.0448,
+      "step": 8252
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006698800000000001,
+      "loss": 1.1567,
+      "step": 8253
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066984,
+      "loss": 1.074,
+      "step": 8254
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006698,
+      "loss": 1.1822,
+      "step": 8255
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066976,
+      "loss": 1.0767,
+      "step": 8256
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066972,
+      "loss": 1.0612,
+      "step": 8257
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006696800000000001,
+      "loss": 1.1209,
+      "step": 8258
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066964,
+      "loss": 1.1475,
+      "step": 8259
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006696,
+      "loss": 1.1215,
+      "step": 8260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066956,
+      "loss": 1.1147,
+      "step": 8261
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006695200000000001,
+      "loss": 1.125,
+      "step": 8262
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066948,
+      "loss": 1.2607,
+      "step": 8263
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006694400000000001,
+      "loss": 1.087,
+      "step": 8264
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006694,
+      "loss": 1.1281,
+      "step": 8265
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006693599999999999,
+      "loss": 1.0917,
+      "step": 8266
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066932,
+      "loss": 1.1271,
+      "step": 8267
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066928,
+      "loss": 1.2038,
+      "step": 8268
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066924,
+      "loss": 1.1428,
+      "step": 8269
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006692,
+      "loss": 1.1089,
+      "step": 8270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066916,
+      "loss": 1.1728,
+      "step": 8271
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066912,
+      "loss": 1.1778,
+      "step": 8272
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006690800000000001,
+      "loss": 1.1271,
+      "step": 8273
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066904,
+      "loss": 1.122,
+      "step": 8274
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006690000000000001,
+      "loss": 1.1095,
+      "step": 8275
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066896,
+      "loss": 1.1446,
+      "step": 8276
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006689199999999999,
+      "loss": 1.0684,
+      "step": 8277
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006688800000000001,
+      "loss": 1.119,
+      "step": 8278
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066884,
+      "loss": 1.1186,
+      "step": 8279
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006688,
+      "loss": 1.1297,
+      "step": 8280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066876,
+      "loss": 1.1214,
+      "step": 8281
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066872,
+      "loss": 1.119,
+      "step": 8282
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006686800000000001,
+      "loss": 1.0665,
+      "step": 8283
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006686400000000001,
+      "loss": 1.1391,
+      "step": 8284
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006686,
+      "loss": 1.0761,
+      "step": 8285
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066856,
+      "loss": 1.1424,
+      "step": 8286
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066852,
+      "loss": 1.1097,
+      "step": 8287
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006684799999999999,
+      "loss": 1.111,
+      "step": 8288
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006684400000000001,
+      "loss": 1.1079,
+      "step": 8289
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006684,
+      "loss": 1.0559,
+      "step": 8290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066836,
+      "loss": 1.1158,
+      "step": 8291
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066832,
+      "loss": 1.0705,
+      "step": 8292
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066828,
+      "loss": 1.1758,
+      "step": 8293
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066824,
+      "loss": 1.1756,
+      "step": 8294
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006682,
+      "loss": 1.0695,
+      "step": 8295
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066816,
+      "loss": 1.147,
+      "step": 8296
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066812,
+      "loss": 1.1565,
+      "step": 8297
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066808,
+      "loss": 1.211,
+      "step": 8298
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066804,
+      "loss": 1.1773,
+      "step": 8299
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006680000000000001,
+      "loss": 0.9995,
+      "step": 8300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066796,
+      "loss": 1.0344,
+      "step": 8301
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066792,
+      "loss": 1.1308,
+      "step": 8302
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066788,
+      "loss": 1.1098,
+      "step": 8303
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006678400000000001,
+      "loss": 1.1401,
+      "step": 8304
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006678,
+      "loss": 1.0998,
+      "step": 8305
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066776,
+      "loss": 1.1681,
+      "step": 8306
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066772,
+      "loss": 1.1444,
+      "step": 8307
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066768,
+      "loss": 1.0289,
+      "step": 8308
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066764,
+      "loss": 1.176,
+      "step": 8309
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006676,
+      "loss": 1.1233,
+      "step": 8310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006675600000000001,
+      "loss": 1.2088,
+      "step": 8311
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066752,
+      "loss": 1.0949,
+      "step": 8312
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066748,
+      "loss": 1.1851,
+      "step": 8313
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066744,
+      "loss": 1.1527,
+      "step": 8314
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006674,
+      "loss": 1.1093,
+      "step": 8315
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066736,
+      "loss": 1.0929,
+      "step": 8316
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066732,
+      "loss": 1.108,
+      "step": 8317
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066728,
+      "loss": 1.0033,
+      "step": 8318
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066724,
+      "loss": 1.0521,
+      "step": 8319
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006672000000000001,
+      "loss": 1.1466,
+      "step": 8320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066716,
+      "loss": 1.1631,
+      "step": 8321
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006671200000000001,
+      "loss": 1.2762,
+      "step": 8322
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066708,
+      "loss": 1.2011,
+      "step": 8323
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006670399999999999,
+      "loss": 1.1543,
+      "step": 8324
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006670000000000001,
+      "loss": 1.0688,
+      "step": 8325
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066696,
+      "loss": 1.0759,
+      "step": 8326
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066692,
+      "loss": 1.1311,
+      "step": 8327
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066688,
+      "loss": 1.1646,
+      "step": 8328
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066684,
+      "loss": 1.128,
+      "step": 8329
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006667999999999999,
+      "loss": 1.083,
+      "step": 8330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006667600000000001,
+      "loss": 1.0817,
+      "step": 8331
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066672,
+      "loss": 1.0689,
+      "step": 8332
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006666800000000001,
+      "loss": 1.1226,
+      "step": 8333
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066664,
+      "loss": 1.1531,
+      "step": 8334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006665999999999999,
+      "loss": 1.0459,
+      "step": 8335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006665600000000001,
+      "loss": 1.063,
+      "step": 8336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066652,
+      "loss": 1.1668,
+      "step": 8337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066648,
+      "loss": 1.0952,
+      "step": 8338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066644,
+      "loss": 1.1288,
+      "step": 8339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006664,
+      "loss": 1.2191,
+      "step": 8340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066636,
+      "loss": 1.1315,
+      "step": 8341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006663200000000001,
+      "loss": 1.1787,
+      "step": 8342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066628,
+      "loss": 1.0799,
+      "step": 8343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066624,
+      "loss": 1.1114,
+      "step": 8344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006662,
+      "loss": 1.1097,
+      "step": 8345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066616,
+      "loss": 1.1171,
+      "step": 8346
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006661200000000001,
+      "loss": 1.0834,
+      "step": 8347
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066608,
+      "loss": 1.0271,
+      "step": 8348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066604,
+      "loss": 1.1883,
+      "step": 8349
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000666,
+      "loss": 1.0956,
+      "step": 8350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006659600000000001,
+      "loss": 1.097,
+      "step": 8351
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066592,
+      "loss": 1.2119,
+      "step": 8352
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066588,
+      "loss": 1.1501,
+      "step": 8353
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066584,
+      "loss": 1.1879,
+      "step": 8354
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006657999999999999,
+      "loss": 1.1815,
+      "step": 8355
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066576,
+      "loss": 1.178,
+      "step": 8356
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066572,
+      "loss": 1.2194,
+      "step": 8357
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006656800000000001,
+      "loss": 1.1253,
+      "step": 8358
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066564,
+      "loss": 1.1404,
+      "step": 8359
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006656,
+      "loss": 1.0758,
+      "step": 8360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066556,
+      "loss": 1.1714,
+      "step": 8361
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006655200000000001,
+      "loss": 1.1264,
+      "step": 8362
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066548,
+      "loss": 1.1786,
+      "step": 8363
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066544,
+      "loss": 1.0232,
+      "step": 8364
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006654,
+      "loss": 1.0858,
+      "step": 8365
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006653599999999999,
+      "loss": 1.0762,
+      "step": 8366
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006653200000000001,
+      "loss": 1.1936,
+      "step": 8367
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066528,
+      "loss": 1.1572,
+      "step": 8368
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006652400000000001,
+      "loss": 1.1686,
+      "step": 8369
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006652,
+      "loss": 1.1559,
+      "step": 8370
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066516,
+      "loss": 1.0841,
+      "step": 8371
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006651200000000001,
+      "loss": 1.1725,
+      "step": 8372
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066508,
+      "loss": 1.1112,
+      "step": 8373
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066504,
+      "loss": 1.1883,
+      "step": 8374
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000665,
+      "loss": 1.0835,
+      "step": 8375
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066496,
+      "loss": 1.1242,
+      "step": 8376
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006649199999999999,
+      "loss": 1.1423,
+      "step": 8377
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006648800000000001,
+      "loss": 1.1597,
+      "step": 8378
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066484,
+      "loss": 1.1616,
+      "step": 8379
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006648,
+      "loss": 1.136,
+      "step": 8380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066476,
+      "loss": 1.1507,
+      "step": 8381
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006647199999999999,
+      "loss": 1.1661,
+      "step": 8382
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006646800000000001,
+      "loss": 1.1489,
+      "step": 8383
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066464,
+      "loss": 1.1233,
+      "step": 8384
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006646,
+      "loss": 1.1127,
+      "step": 8385
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066456,
+      "loss": 1.1453,
+      "step": 8386
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066452,
+      "loss": 1.1249,
+      "step": 8387
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066448,
+      "loss": 1.146,
+      "step": 8388
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006644400000000001,
+      "loss": 1.1572,
+      "step": 8389
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006644,
+      "loss": 1.1184,
+      "step": 8390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066436,
+      "loss": 1.1124,
+      "step": 8391
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066432,
+      "loss": 1.1534,
+      "step": 8392
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066428,
+      "loss": 1.1093,
+      "step": 8393
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006642400000000001,
+      "loss": 1.2245,
+      "step": 8394
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006642,
+      "loss": 1.1246,
+      "step": 8395
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066416,
+      "loss": 1.1129,
+      "step": 8396
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066412,
+      "loss": 1.1277,
+      "step": 8397
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066408,
+      "loss": 1.1814,
+      "step": 8398
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066404,
+      "loss": 1.1545,
+      "step": 8399
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006640000000000001,
+      "loss": 1.0713,
+      "step": 8400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066396,
+      "loss": 1.132,
+      "step": 8401
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006639199999999999,
+      "loss": 1.0759,
+      "step": 8402
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066388,
+      "loss": 1.0923,
+      "step": 8403
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066384,
+      "loss": 1.1597,
+      "step": 8404
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006638,
+      "loss": 1.1925,
+      "step": 8405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066376,
+      "loss": 0.9868,
+      "step": 8406
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066372,
+      "loss": 1.1954,
+      "step": 8407
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066368,
+      "loss": 1.1098,
+      "step": 8408
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006636400000000001,
+      "loss": 1.0811,
+      "step": 8409
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006636,
+      "loss": 1.1525,
+      "step": 8410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006635600000000001,
+      "loss": 1.1691,
+      "step": 8411
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066352,
+      "loss": 1.0975,
+      "step": 8412
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006634799999999999,
+      "loss": 1.0496,
+      "step": 8413
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006634400000000001,
+      "loss": 1.0915,
+      "step": 8414
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006634,
+      "loss": 1.0726,
+      "step": 8415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066336,
+      "loss": 1.0591,
+      "step": 8416
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066332,
+      "loss": 1.1653,
+      "step": 8417
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066328,
+      "loss": 1.2064,
+      "step": 8418
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066324,
+      "loss": 1.1713,
+      "step": 8419
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006632000000000001,
+      "loss": 1.0699,
+      "step": 8420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066316,
+      "loss": 1.1227,
+      "step": 8421
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066312,
+      "loss": 1.0712,
+      "step": 8422
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066308,
+      "loss": 1.0585,
+      "step": 8423
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006630399999999999,
+      "loss": 1.2034,
+      "step": 8424
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006630000000000001,
+      "loss": 1.0223,
+      "step": 8425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066296,
+      "loss": 1.133,
+      "step": 8426
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066292,
+      "loss": 1.1776,
+      "step": 8427
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066288,
+      "loss": 1.1058,
+      "step": 8428
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066284,
+      "loss": 1.0916,
+      "step": 8429
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006628,
+      "loss": 1.1401,
+      "step": 8430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066276,
+      "loss": 1.0763,
+      "step": 8431
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066272,
+      "loss": 1.1509,
+      "step": 8432
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066268,
+      "loss": 1.0436,
+      "step": 8433
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066264,
+      "loss": 1.1408,
+      "step": 8434
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006626,
+      "loss": 1.1147,
+      "step": 8435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006625600000000001,
+      "loss": 1.1005,
+      "step": 8436
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066252,
+      "loss": 1.1265,
+      "step": 8437
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066248,
+      "loss": 1.155,
+      "step": 8438
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066244,
+      "loss": 1.1339,
+      "step": 8439
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006624000000000001,
+      "loss": 1.1536,
+      "step": 8440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066236,
+      "loss": 1.1397,
+      "step": 8441
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066232,
+      "loss": 1.1067,
+      "step": 8442
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066228,
+      "loss": 1.0799,
+      "step": 8443
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066224,
+      "loss": 1.1259,
+      "step": 8444
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006622,
+      "loss": 1.0928,
+      "step": 8445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066216,
+      "loss": 1.1087,
+      "step": 8446
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006621200000000001,
+      "loss": 1.1775,
+      "step": 8447
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066208,
+      "loss": 1.0482,
+      "step": 8448
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066204,
+      "loss": 1.1718,
+      "step": 8449
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000662,
+      "loss": 1.1231,
+      "step": 8450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066196,
+      "loss": 1.0695,
+      "step": 8451
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066192,
+      "loss": 1.1668,
+      "step": 8452
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066188,
+      "loss": 1.1455,
+      "step": 8453
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066184,
+      "loss": 1.1326,
+      "step": 8454
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006618,
+      "loss": 1.1012,
+      "step": 8455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006617600000000001,
+      "loss": 1.188,
+      "step": 8456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066172,
+      "loss": 1.1817,
+      "step": 8457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006616800000000001,
+      "loss": 1.1239,
+      "step": 8458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066164,
+      "loss": 1.1536,
+      "step": 8459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006615999999999999,
+      "loss": 1.0835,
+      "step": 8460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006615600000000001,
+      "loss": 1.1811,
+      "step": 8461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066152,
+      "loss": 1.1149,
+      "step": 8462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066148,
+      "loss": 1.1658,
+      "step": 8463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066144,
+      "loss": 1.0667,
+      "step": 8464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006614,
+      "loss": 1.1432,
+      "step": 8465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006613599999999999,
+      "loss": 1.154,
+      "step": 8466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006613200000000001,
+      "loss": 1.0775,
+      "step": 8467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066128,
+      "loss": 1.1985,
+      "step": 8468
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006612400000000001,
+      "loss": 1.1677,
+      "step": 8469
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006612,
+      "loss": 1.2058,
+      "step": 8470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006611599999999999,
+      "loss": 1.1182,
+      "step": 8471
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006611200000000001,
+      "loss": 1.0936,
+      "step": 8472
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066108,
+      "loss": 1.1837,
+      "step": 8473
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066104,
+      "loss": 1.1493,
+      "step": 8474
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000661,
+      "loss": 1.1275,
+      "step": 8475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066096,
+      "loss": 1.1024,
+      "step": 8476
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066092,
+      "loss": 1.1333,
+      "step": 8477
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006608800000000001,
+      "loss": 1.1912,
+      "step": 8478
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066084,
+      "loss": 1.0879,
+      "step": 8479
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006608,
+      "loss": 1.1029,
+      "step": 8480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066076,
+      "loss": 1.1467,
+      "step": 8481
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066072,
+      "loss": 1.1533,
+      "step": 8482
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006606800000000001,
+      "loss": 1.1745,
+      "step": 8483
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066064,
+      "loss": 1.1445,
+      "step": 8484
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0006606,
+      "loss": 1.2004,
+      "step": 8485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066056,
+      "loss": 1.0818,
+      "step": 8486
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066052,
+      "loss": 1.2109,
+      "step": 8487
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00066048,
+      "loss": 1.1031,
+      "step": 8488
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006604400000000001,
+      "loss": 1.1012,
+      "step": 8489
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006604,
+      "loss": 1.1435,
+      "step": 8490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006603599999999999,
+      "loss": 1.1039,
+      "step": 8491
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00066032,
+      "loss": 1.2529,
+      "step": 8492
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00066028,
+      "loss": 1.1774,
+      "step": 8493
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006602400000000001,
+      "loss": 1.0203,
+      "step": 8494
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006602,
+      "loss": 1.1987,
+      "step": 8495
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00066016,
+      "loss": 1.1233,
+      "step": 8496
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00066012,
+      "loss": 1.1074,
+      "step": 8497
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006600800000000001,
+      "loss": 1.1474,
+      "step": 8498
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00066004,
+      "loss": 1.2466,
+      "step": 8499
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00066,
+      "loss": 1.0777,
+      "step": 8500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065996,
+      "loss": 1.035,
+      "step": 8501
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006599199999999999,
+      "loss": 1.0732,
+      "step": 8502
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006598800000000001,
+      "loss": 1.1544,
+      "step": 8503
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065984,
+      "loss": 1.1433,
+      "step": 8504
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006598000000000001,
+      "loss": 1.1648,
+      "step": 8505
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065976,
+      "loss": 1.1511,
+      "step": 8506
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065972,
+      "loss": 1.0938,
+      "step": 8507
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065968,
+      "loss": 1.1721,
+      "step": 8508
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065964,
+      "loss": 1.1336,
+      "step": 8509
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006596,
+      "loss": 1.1672,
+      "step": 8510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065956,
+      "loss": 1.1315,
+      "step": 8511
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065952,
+      "loss": 1.1136,
+      "step": 8512
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006594799999999999,
+      "loss": 1.1461,
+      "step": 8513
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006594400000000001,
+      "loss": 1.1104,
+      "step": 8514
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006594,
+      "loss": 1.145,
+      "step": 8515
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065936,
+      "loss": 1.1025,
+      "step": 8516
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065932,
+      "loss": 1.1401,
+      "step": 8517
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065928,
+      "loss": 1.2485,
+      "step": 8518
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006592400000000001,
+      "loss": 1.1699,
+      "step": 8519
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006592,
+      "loss": 0.9652,
+      "step": 8520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065916,
+      "loss": 1.2238,
+      "step": 8521
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065912,
+      "loss": 1.1273,
+      "step": 8522
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065908,
+      "loss": 1.0992,
+      "step": 8523
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065904,
+      "loss": 1.1325,
+      "step": 8524
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006590000000000001,
+      "loss": 1.1114,
+      "step": 8525
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065896,
+      "loss": 1.1287,
+      "step": 8526
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065892,
+      "loss": 1.1537,
+      "step": 8527
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065888,
+      "loss": 1.1391,
+      "step": 8528
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065884,
+      "loss": 1.1665,
+      "step": 8529
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006588000000000001,
+      "loss": 1.1214,
+      "step": 8530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065876,
+      "loss": 1.0999,
+      "step": 8531
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065872,
+      "loss": 1.1178,
+      "step": 8532
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065868,
+      "loss": 1.1444,
+      "step": 8533
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065864,
+      "loss": 1.1648,
+      "step": 8534
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006586,
+      "loss": 1.1449,
+      "step": 8535
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006585600000000001,
+      "loss": 1.1229,
+      "step": 8536
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065852,
+      "loss": 1.0928,
+      "step": 8537
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006584799999999999,
+      "loss": 1.0842,
+      "step": 8538
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065844,
+      "loss": 1.1607,
+      "step": 8539
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006584,
+      "loss": 1.098,
+      "step": 8540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065836,
+      "loss": 1.1117,
+      "step": 8541
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065832,
+      "loss": 1.0761,
+      "step": 8542
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065828,
+      "loss": 1.1907,
+      "step": 8543
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065824,
+      "loss": 1.168,
+      "step": 8544
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006582000000000001,
+      "loss": 1.1362,
+      "step": 8545
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065816,
+      "loss": 1.1375,
+      "step": 8546
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006581200000000001,
+      "loss": 1.0673,
+      "step": 8547
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065808,
+      "loss": 1.1036,
+      "step": 8548
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006580399999999999,
+      "loss": 1.1016,
+      "step": 8549
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006580000000000001,
+      "loss": 1.1401,
+      "step": 8550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065796,
+      "loss": 1.1119,
+      "step": 8551
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065792,
+      "loss": 1.2034,
+      "step": 8552
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065788,
+      "loss": 1.1466,
+      "step": 8553
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065784,
+      "loss": 1.1338,
+      "step": 8554
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006578,
+      "loss": 1.1762,
+      "step": 8555
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006577600000000001,
+      "loss": 1.1316,
+      "step": 8556
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065772,
+      "loss": 1.1149,
+      "step": 8557
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065768,
+      "loss": 1.0967,
+      "step": 8558
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065764,
+      "loss": 1.094,
+      "step": 8559
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006575999999999999,
+      "loss": 1.0534,
+      "step": 8560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006575600000000001,
+      "loss": 1.1473,
+      "step": 8561
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065752,
+      "loss": 1.1283,
+      "step": 8562
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065748,
+      "loss": 1.1372,
+      "step": 8563
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065744,
+      "loss": 1.0594,
+      "step": 8564
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006574,
+      "loss": 1.0878,
+      "step": 8565
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006573600000000001,
+      "loss": 1.1723,
+      "step": 8566
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006573200000000001,
+      "loss": 1.1108,
+      "step": 8567
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065728,
+      "loss": 1.0989,
+      "step": 8568
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065724,
+      "loss": 1.1572,
+      "step": 8569
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006572,
+      "loss": 1.1448,
+      "step": 8570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065716,
+      "loss": 1.1141,
+      "step": 8571
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006571200000000001,
+      "loss": 1.0502,
+      "step": 8572
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065708,
+      "loss": 1.1693,
+      "step": 8573
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065704,
+      "loss": 1.1524,
+      "step": 8574
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000657,
+      "loss": 1.1365,
+      "step": 8575
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065696,
+      "loss": 1.1819,
+      "step": 8576
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065692,
+      "loss": 1.0817,
+      "step": 8577
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065688,
+      "loss": 1.1143,
+      "step": 8578
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065684,
+      "loss": 1.0619,
+      "step": 8579
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006568,
+      "loss": 1.1892,
+      "step": 8580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065676,
+      "loss": 1.1023,
+      "step": 8581
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065672,
+      "loss": 1.1352,
+      "step": 8582
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006566800000000001,
+      "loss": 1.0911,
+      "step": 8583
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065664,
+      "loss": 1.1195,
+      "step": 8584
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006566,
+      "loss": 1.0528,
+      "step": 8585
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065656,
+      "loss": 1.1221,
+      "step": 8586
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065652,
+      "loss": 1.1684,
+      "step": 8587
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065648,
+      "loss": 1.1482,
+      "step": 8588
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065644,
+      "loss": 1.1403,
+      "step": 8589
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006564,
+      "loss": 1.1036,
+      "step": 8590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065636,
+      "loss": 1.0167,
+      "step": 8591
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006563200000000001,
+      "loss": 1.0763,
+      "step": 8592
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065628,
+      "loss": 1.0999,
+      "step": 8593
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006562400000000001,
+      "loss": 1.1024,
+      "step": 8594
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006562,
+      "loss": 1.101,
+      "step": 8595
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065616,
+      "loss": 1.1441,
+      "step": 8596
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065612,
+      "loss": 1.1691,
+      "step": 8597
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065608,
+      "loss": 1.1408,
+      "step": 8598
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065604,
+      "loss": 1.0237,
+      "step": 8599
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000656,
+      "loss": 1.0991,
+      "step": 8600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065596,
+      "loss": 1.141,
+      "step": 8601
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006559199999999999,
+      "loss": 1.1099,
+      "step": 8602
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006558800000000001,
+      "loss": 1.121,
+      "step": 8603
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065584,
+      "loss": 1.1659,
+      "step": 8604
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006558000000000001,
+      "loss": 1.1767,
+      "step": 8605
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065576,
+      "loss": 1.1283,
+      "step": 8606
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006557199999999999,
+      "loss": 1.1271,
+      "step": 8607
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006556800000000001,
+      "loss": 1.0742,
+      "step": 8608
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065564,
+      "loss": 1.057,
+      "step": 8609
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006556,
+      "loss": 1.1106,
+      "step": 8610
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065556,
+      "loss": 1.0911,
+      "step": 8611
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065552,
+      "loss": 1.0936,
+      "step": 8612
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065548,
+      "loss": 1.1463,
+      "step": 8613
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006554400000000001,
+      "loss": 1.138,
+      "step": 8614
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006554,
+      "loss": 0.9554,
+      "step": 8615
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065536,
+      "loss": 1.0987,
+      "step": 8616
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065532,
+      "loss": 1.1858,
+      "step": 8617
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065528,
+      "loss": 1.1264,
+      "step": 8618
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006552400000000001,
+      "loss": 1.0992,
+      "step": 8619
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006552,
+      "loss": 1.1596,
+      "step": 8620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065516,
+      "loss": 1.0851,
+      "step": 8621
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065512,
+      "loss": 1.1183,
+      "step": 8622
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065508,
+      "loss": 1.0007,
+      "step": 8623
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065504,
+      "loss": 1.0814,
+      "step": 8624
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006550000000000001,
+      "loss": 1.1277,
+      "step": 8625
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065496,
+      "loss": 1.1239,
+      "step": 8626
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006549199999999999,
+      "loss": 1.1275,
+      "step": 8627
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065488,
+      "loss": 1.1586,
+      "step": 8628
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065484,
+      "loss": 1.1649,
+      "step": 8629
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006548000000000001,
+      "loss": 1.1551,
+      "step": 8630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065476,
+      "loss": 1.107,
+      "step": 8631
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065472,
+      "loss": 1.1534,
+      "step": 8632
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065468,
+      "loss": 1.057,
+      "step": 8633
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006546400000000001,
+      "loss": 1.0934,
+      "step": 8634
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006546,
+      "loss": 1.0718,
+      "step": 8635
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065456,
+      "loss": 1.1182,
+      "step": 8636
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065452,
+      "loss": 1.1468,
+      "step": 8637
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006544799999999999,
+      "loss": 1.0624,
+      "step": 8638
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006544400000000001,
+      "loss": 1.0864,
+      "step": 8639
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006544,
+      "loss": 1.014,
+      "step": 8640
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006543600000000001,
+      "loss": 0.9957,
+      "step": 8641
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065432,
+      "loss": 1.1545,
+      "step": 8642
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065428,
+      "loss": 1.0128,
+      "step": 8643
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065424,
+      "loss": 1.0333,
+      "step": 8644
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006542,
+      "loss": 1.1834,
+      "step": 8645
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065416,
+      "loss": 1.0944,
+      "step": 8646
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065412,
+      "loss": 1.0352,
+      "step": 8647
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065408,
+      "loss": 1.0779,
+      "step": 8648
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006540399999999999,
+      "loss": 1.1225,
+      "step": 8649
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006540000000000001,
+      "loss": 1.1836,
+      "step": 8650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065396,
+      "loss": 1.0956,
+      "step": 8651
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065392,
+      "loss": 0.9991,
+      "step": 8652
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065388,
+      "loss": 1.1038,
+      "step": 8653
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065384,
+      "loss": 1.1318,
+      "step": 8654
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006538000000000001,
+      "loss": 1.0903,
+      "step": 8655
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065376,
+      "loss": 1.1079,
+      "step": 8656
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065372,
+      "loss": 1.0899,
+      "step": 8657
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065368,
+      "loss": 1.122,
+      "step": 8658
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065364,
+      "loss": 1.1959,
+      "step": 8659
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006536,
+      "loss": 1.0532,
+      "step": 8660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006535600000000001,
+      "loss": 1.0176,
+      "step": 8661
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065352,
+      "loss": 1.1015,
+      "step": 8662
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065348,
+      "loss": 1.0943,
+      "step": 8663
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065344,
+      "loss": 1.1157,
+      "step": 8664
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006533999999999999,
+      "loss": 1.1309,
+      "step": 8665
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006533600000000001,
+      "loss": 1.1548,
+      "step": 8666
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065332,
+      "loss": 1.0851,
+      "step": 8667
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065328,
+      "loss": 1.1661,
+      "step": 8668
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065324,
+      "loss": 1.1524,
+      "step": 8669
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006532,
+      "loss": 1.2083,
+      "step": 8670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065316,
+      "loss": 1.2014,
+      "step": 8671
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006531200000000001,
+      "loss": 1.1542,
+      "step": 8672
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065308,
+      "loss": 1.1425,
+      "step": 8673
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006530399999999999,
+      "loss": 1.2017,
+      "step": 8674
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000653,
+      "loss": 1.079,
+      "step": 8675
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065296,
+      "loss": 1.1826,
+      "step": 8676
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065292,
+      "loss": 1.0883,
+      "step": 8677
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065288,
+      "loss": 1.1543,
+      "step": 8678
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065284,
+      "loss": 1.0416,
+      "step": 8679
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006528,
+      "loss": 1.1575,
+      "step": 8680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006527600000000001,
+      "loss": 1.0999,
+      "step": 8681
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065272,
+      "loss": 1.1279,
+      "step": 8682
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006526800000000001,
+      "loss": 1.1049,
+      "step": 8683
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065264,
+      "loss": 1.0958,
+      "step": 8684
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006525999999999999,
+      "loss": 1.1122,
+      "step": 8685
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065256,
+      "loss": 1.1372,
+      "step": 8686
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065252,
+      "loss": 1.1025,
+      "step": 8687
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065248,
+      "loss": 1.0156,
+      "step": 8688
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065244,
+      "loss": 1.1172,
+      "step": 8689
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006524,
+      "loss": 1.2026,
+      "step": 8690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065236,
+      "loss": 1.1288,
+      "step": 8691
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006523200000000001,
+      "loss": 1.189,
+      "step": 8692
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065228,
+      "loss": 1.1372,
+      "step": 8693
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065224,
+      "loss": 1.0542,
+      "step": 8694
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006522,
+      "loss": 1.059,
+      "step": 8695
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006521599999999999,
+      "loss": 1.1374,
+      "step": 8696
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006521200000000001,
+      "loss": 1.1581,
+      "step": 8697
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065208,
+      "loss": 1.2134,
+      "step": 8698
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065204,
+      "loss": 1.1213,
+      "step": 8699
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000652,
+      "loss": 0.994,
+      "step": 8700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065196,
+      "loss": 1.1024,
+      "step": 8701
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006519200000000001,
+      "loss": 1.1069,
+      "step": 8702
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006518800000000001,
+      "loss": 1.045,
+      "step": 8703
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065184,
+      "loss": 1.0792,
+      "step": 8704
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006518,
+      "loss": 1.0726,
+      "step": 8705
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065176,
+      "loss": 1.1247,
+      "step": 8706
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065172,
+      "loss": 1.1448,
+      "step": 8707
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006516800000000001,
+      "loss": 1.1909,
+      "step": 8708
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065164,
+      "loss": 1.1459,
+      "step": 8709
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006516,
+      "loss": 1.1824,
+      "step": 8710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065156,
+      "loss": 1.2129,
+      "step": 8711
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065152,
+      "loss": 1.1815,
+      "step": 8712
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065148,
+      "loss": 1.1034,
+      "step": 8713
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065144,
+      "loss": 1.0548,
+      "step": 8714
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006514,
+      "loss": 1.1063,
+      "step": 8715
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065136,
+      "loss": 1.1647,
+      "step": 8716
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065132,
+      "loss": 1.1576,
+      "step": 8717
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065128,
+      "loss": 1.077,
+      "step": 8718
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006512400000000001,
+      "loss": 1.0599,
+      "step": 8719
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006512,
+      "loss": 1.0855,
+      "step": 8720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065116,
+      "loss": 1.1334,
+      "step": 8721
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065112,
+      "loss": 1.1907,
+      "step": 8722
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065108,
+      "loss": 1.1262,
+      "step": 8723
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065104,
+      "loss": 1.0738,
+      "step": 8724
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000651,
+      "loss": 1.0985,
+      "step": 8725
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065096,
+      "loss": 1.1539,
+      "step": 8726
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065092,
+      "loss": 1.1161,
+      "step": 8727
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006508800000000001,
+      "loss": 1.105,
+      "step": 8728
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065084,
+      "loss": 0.9966,
+      "step": 8729
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006508000000000001,
+      "loss": 1.1081,
+      "step": 8730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065076,
+      "loss": 1.1189,
+      "step": 8731
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065072,
+      "loss": 1.0264,
+      "step": 8732
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065068,
+      "loss": 1.0859,
+      "step": 8733
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065064,
+      "loss": 1.0692,
+      "step": 8734
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006506,
+      "loss": 1.0885,
+      "step": 8735
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065056,
+      "loss": 1.1543,
+      "step": 8736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065052,
+      "loss": 1.2053,
+      "step": 8737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006504799999999999,
+      "loss": 1.1074,
+      "step": 8738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006504400000000001,
+      "loss": 1.1079,
+      "step": 8739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006504,
+      "loss": 1.0753,
+      "step": 8740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0006503600000000001,
+      "loss": 1.1472,
+      "step": 8741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00065032,
+      "loss": 1.115,
+      "step": 8742
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006502799999999999,
+      "loss": 1.2213,
+      "step": 8743
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006502400000000001,
+      "loss": 1.1827,
+      "step": 8744
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006502,
+      "loss": 1.0868,
+      "step": 8745
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00065016,
+      "loss": 1.1663,
+      "step": 8746
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00065012,
+      "loss": 1.0657,
+      "step": 8747
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00065008,
+      "loss": 1.118,
+      "step": 8748
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00065004,
+      "loss": 1.0946,
+      "step": 8749
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006500000000000001,
+      "loss": 1.1622,
+      "step": 8750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064996,
+      "loss": 1.0699,
+      "step": 8751
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006499200000000001,
+      "loss": 1.1491,
+      "step": 8752
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064988,
+      "loss": 1.0565,
+      "step": 8753
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006498399999999999,
+      "loss": 1.0964,
+      "step": 8754
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006498000000000001,
+      "loss": 1.095,
+      "step": 8755
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064976,
+      "loss": 1.111,
+      "step": 8756
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064972,
+      "loss": 1.1118,
+      "step": 8757
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064968,
+      "loss": 1.1346,
+      "step": 8758
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064964,
+      "loss": 1.1801,
+      "step": 8759
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006496,
+      "loss": 1.1566,
+      "step": 8760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006495600000000001,
+      "loss": 1.1485,
+      "step": 8761
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064952,
+      "loss": 1.1124,
+      "step": 8762
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006494799999999999,
+      "loss": 1.0834,
+      "step": 8763
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064944,
+      "loss": 1.1943,
+      "step": 8764
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006494,
+      "loss": 1.1859,
+      "step": 8765
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006493600000000001,
+      "loss": 1.0797,
+      "step": 8766
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064932,
+      "loss": 1.0475,
+      "step": 8767
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064928,
+      "loss": 1.0547,
+      "step": 8768
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064924,
+      "loss": 1.1802,
+      "step": 8769
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006492000000000001,
+      "loss": 1.111,
+      "step": 8770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064916,
+      "loss": 1.1315,
+      "step": 8771
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064912,
+      "loss": 1.0147,
+      "step": 8772
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064908,
+      "loss": 1.0808,
+      "step": 8773
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006490399999999999,
+      "loss": 1.1173,
+      "step": 8774
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006490000000000001,
+      "loss": 1.1471,
+      "step": 8775
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064896,
+      "loss": 1.119,
+      "step": 8776
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006489200000000001,
+      "loss": 1.0695,
+      "step": 8777
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064888,
+      "loss": 1.1364,
+      "step": 8778
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064884,
+      "loss": 1.166,
+      "step": 8779
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006488,
+      "loss": 1.1095,
+      "step": 8780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006487600000000001,
+      "loss": 1.1188,
+      "step": 8781
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064872,
+      "loss": 1.0521,
+      "step": 8782
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064868,
+      "loss": 1.0615,
+      "step": 8783
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064864,
+      "loss": 1.1024,
+      "step": 8784
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006485999999999999,
+      "loss": 1.1196,
+      "step": 8785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006485600000000001,
+      "loss": 1.0714,
+      "step": 8786
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064852,
+      "loss": 1.1327,
+      "step": 8787
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064848,
+      "loss": 1.0609,
+      "step": 8788
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064844,
+      "loss": 1.1145,
+      "step": 8789
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006484,
+      "loss": 1.1601,
+      "step": 8790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006483600000000001,
+      "loss": 1.1097,
+      "step": 8791
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064832,
+      "loss": 1.0654,
+      "step": 8792
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064828,
+      "loss": 1.1406,
+      "step": 8793
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064824,
+      "loss": 1.084,
+      "step": 8794
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006482,
+      "loss": 1.067,
+      "step": 8795
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064816,
+      "loss": 1.0486,
+      "step": 8796
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006481200000000001,
+      "loss": 1.1572,
+      "step": 8797
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064808,
+      "loss": 1.1179,
+      "step": 8798
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064804,
+      "loss": 1.0913,
+      "step": 8799
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000648,
+      "loss": 1.1485,
+      "step": 8800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006479599999999999,
+      "loss": 1.1906,
+      "step": 8801
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006479200000000001,
+      "loss": 1.0924,
+      "step": 8802
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064788,
+      "loss": 1.1647,
+      "step": 8803
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064784,
+      "loss": 1.0981,
+      "step": 8804
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006478,
+      "loss": 1.1624,
+      "step": 8805
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064776,
+      "loss": 1.1038,
+      "step": 8806
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064772,
+      "loss": 1.0852,
+      "step": 8807
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006476800000000001,
+      "loss": 1.05,
+      "step": 8808
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064764,
+      "loss": 1.1015,
+      "step": 8809
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006476,
+      "loss": 1.0365,
+      "step": 8810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064756,
+      "loss": 1.122,
+      "step": 8811
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064752,
+      "loss": 1.2044,
+      "step": 8812
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006474800000000001,
+      "loss": 1.1361,
+      "step": 8813
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064744,
+      "loss": 1.1314,
+      "step": 8814
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006474,
+      "loss": 1.1153,
+      "step": 8815
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064736,
+      "loss": 1.1796,
+      "step": 8816
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006473200000000001,
+      "loss": 1.0966,
+      "step": 8817
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064728,
+      "loss": 1.0835,
+      "step": 8818
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006472400000000001,
+      "loss": 1.1841,
+      "step": 8819
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006472,
+      "loss": 1.185,
+      "step": 8820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006471599999999999,
+      "loss": 1.1736,
+      "step": 8821
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064712,
+      "loss": 1.1504,
+      "step": 8822
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064708,
+      "loss": 1.0846,
+      "step": 8823
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064704,
+      "loss": 1.1856,
+      "step": 8824
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000647,
+      "loss": 1.1199,
+      "step": 8825
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064696,
+      "loss": 1.1069,
+      "step": 8826
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064692,
+      "loss": 1.1524,
+      "step": 8827
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006468800000000001,
+      "loss": 1.1362,
+      "step": 8828
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064684,
+      "loss": 1.0811,
+      "step": 8829
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006468000000000001,
+      "loss": 1.1133,
+      "step": 8830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064676,
+      "loss": 1.1268,
+      "step": 8831
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006467199999999999,
+      "loss": 1.0604,
+      "step": 8832
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006466800000000001,
+      "loss": 1.1126,
+      "step": 8833
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064664,
+      "loss": 1.1318,
+      "step": 8834
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006466,
+      "loss": 1.1316,
+      "step": 8835
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064656,
+      "loss": 1.1174,
+      "step": 8836
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064652,
+      "loss": 1.0517,
+      "step": 8837
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006464800000000001,
+      "loss": 1.0275,
+      "step": 8838
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006464400000000001,
+      "loss": 1.0431,
+      "step": 8839
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006464,
+      "loss": 1.089,
+      "step": 8840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064636,
+      "loss": 1.0521,
+      "step": 8841
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064632,
+      "loss": 1.1446,
+      "step": 8842
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006462799999999999,
+      "loss": 1.1992,
+      "step": 8843
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006462400000000001,
+      "loss": 1.0899,
+      "step": 8844
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006462,
+      "loss": 1.0585,
+      "step": 8845
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064616,
+      "loss": 1.0843,
+      "step": 8846
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064612,
+      "loss": 1.0665,
+      "step": 8847
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064608,
+      "loss": 1.1426,
+      "step": 8848
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064604,
+      "loss": 1.1868,
+      "step": 8849
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000646,
+      "loss": 1.1238,
+      "step": 8850
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064596,
+      "loss": 1.0222,
+      "step": 8851
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064592,
+      "loss": 1.1272,
+      "step": 8852
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064588,
+      "loss": 1.0356,
+      "step": 8853
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064584,
+      "loss": 1.1085,
+      "step": 8854
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006458000000000001,
+      "loss": 1.0699,
+      "step": 8855
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064576,
+      "loss": 1.131,
+      "step": 8856
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064572,
+      "loss": 1.1609,
+      "step": 8857
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064568,
+      "loss": 1.118,
+      "step": 8858
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064564,
+      "loss": 1.1339,
+      "step": 8859
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006456,
+      "loss": 1.0108,
+      "step": 8860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064556,
+      "loss": 1.0829,
+      "step": 8861
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064552,
+      "loss": 1.1125,
+      "step": 8862
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064548,
+      "loss": 1.0895,
+      "step": 8863
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006454400000000001,
+      "loss": 1.1103,
+      "step": 8864
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006454,
+      "loss": 1.0863,
+      "step": 8865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006453600000000001,
+      "loss": 1.0485,
+      "step": 8866
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064532,
+      "loss": 1.1273,
+      "step": 8867
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064528,
+      "loss": 1.0682,
+      "step": 8868
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064524,
+      "loss": 1.0682,
+      "step": 8869
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006452,
+      "loss": 1.1286,
+      "step": 8870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064516,
+      "loss": 1.1627,
+      "step": 8871
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064512,
+      "loss": 1.1049,
+      "step": 8872
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064508,
+      "loss": 1.0333,
+      "step": 8873
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006450399999999999,
+      "loss": 1.1976,
+      "step": 8874
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006450000000000001,
+      "loss": 1.0913,
+      "step": 8875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064496,
+      "loss": 1.1338,
+      "step": 8876
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006449200000000001,
+      "loss": 1.1336,
+      "step": 8877
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064488,
+      "loss": 1.1512,
+      "step": 8878
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006448399999999999,
+      "loss": 1.1608,
+      "step": 8879
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006448000000000001,
+      "loss": 1.0716,
+      "step": 8880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064476,
+      "loss": 1.1078,
+      "step": 8881
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064472,
+      "loss": 1.1551,
+      "step": 8882
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064468,
+      "loss": 1.0797,
+      "step": 8883
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064464,
+      "loss": 1.1299,
+      "step": 8884
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006446,
+      "loss": 1.0921,
+      "step": 8885
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006445600000000001,
+      "loss": 1.0708,
+      "step": 8886
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064452,
+      "loss": 1.2021,
+      "step": 8887
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006444800000000001,
+      "loss": 1.0609,
+      "step": 8888
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064444,
+      "loss": 1.1289,
+      "step": 8889
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006443999999999999,
+      "loss": 1.0828,
+      "step": 8890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006443600000000001,
+      "loss": 1.0964,
+      "step": 8891
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064432,
+      "loss": 1.1816,
+      "step": 8892
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064428,
+      "loss": 1.1233,
+      "step": 8893
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064424,
+      "loss": 1.0979,
+      "step": 8894
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006442,
+      "loss": 1.0763,
+      "step": 8895
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064416,
+      "loss": 1.0984,
+      "step": 8896
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006441200000000001,
+      "loss": 1.1489,
+      "step": 8897
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064408,
+      "loss": 1.1157,
+      "step": 8898
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006440399999999999,
+      "loss": 1.1951,
+      "step": 8899
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000644,
+      "loss": 1.165,
+      "step": 8900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064396,
+      "loss": 1.085,
+      "step": 8901
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006439200000000001,
+      "loss": 1.1493,
+      "step": 8902
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064388,
+      "loss": 1.0835,
+      "step": 8903
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064384,
+      "loss": 1.1827,
+      "step": 8904
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006438,
+      "loss": 1.1104,
+      "step": 8905
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006437600000000001,
+      "loss": 1.1984,
+      "step": 8906
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064372,
+      "loss": 1.1077,
+      "step": 8907
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064368,
+      "loss": 1.1497,
+      "step": 8908
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064364,
+      "loss": 1.0579,
+      "step": 8909
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006435999999999999,
+      "loss": 1.0493,
+      "step": 8910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064356,
+      "loss": 1.1727,
+      "step": 8911
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064352,
+      "loss": 1.1151,
+      "step": 8912
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006434800000000001,
+      "loss": 1.1891,
+      "step": 8913
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064344,
+      "loss": 1.2094,
+      "step": 8914
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006434,
+      "loss": 1.0996,
+      "step": 8915
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064336,
+      "loss": 0.9488,
+      "step": 8916
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006433200000000001,
+      "loss": 1.1936,
+      "step": 8917
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064328,
+      "loss": 1.1164,
+      "step": 8918
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064324,
+      "loss": 1.1399,
+      "step": 8919
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006432,
+      "loss": 1.1057,
+      "step": 8920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006431599999999999,
+      "loss": 1.1539,
+      "step": 8921
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006431200000000001,
+      "loss": 1.0255,
+      "step": 8922
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064308,
+      "loss": 1.0316,
+      "step": 8923
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064304,
+      "loss": 1.1895,
+      "step": 8924
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000643,
+      "loss": 1.1131,
+      "step": 8925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064296,
+      "loss": 1.0088,
+      "step": 8926
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006429200000000001,
+      "loss": 1.0129,
+      "step": 8927
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064288,
+      "loss": 1.0543,
+      "step": 8928
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064284,
+      "loss": 1.1575,
+      "step": 8929
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006428,
+      "loss": 1.1791,
+      "step": 8930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064276,
+      "loss": 1.1659,
+      "step": 8931
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006427199999999999,
+      "loss": 1.0272,
+      "step": 8932
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006426800000000001,
+      "loss": 1.1311,
+      "step": 8933
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064264,
+      "loss": 1.1139,
+      "step": 8934
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006426,
+      "loss": 1.1945,
+      "step": 8935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064256,
+      "loss": 1.1217,
+      "step": 8936
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006425199999999999,
+      "loss": 1.1092,
+      "step": 8937
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006424800000000001,
+      "loss": 1.0825,
+      "step": 8938
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064244,
+      "loss": 1.1334,
+      "step": 8939
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006424,
+      "loss": 1.1312,
+      "step": 8940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064236,
+      "loss": 1.116,
+      "step": 8941
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064232,
+      "loss": 1.1424,
+      "step": 8942
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064228,
+      "loss": 1.1595,
+      "step": 8943
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006422400000000001,
+      "loss": 1.1113,
+      "step": 8944
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006422,
+      "loss": 1.0801,
+      "step": 8945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064216,
+      "loss": 1.1113,
+      "step": 8946
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064212,
+      "loss": 1.1249,
+      "step": 8947
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064208,
+      "loss": 1.1743,
+      "step": 8948
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006420400000000001,
+      "loss": 1.085,
+      "step": 8949
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000642,
+      "loss": 1.1125,
+      "step": 8950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064196,
+      "loss": 1.132,
+      "step": 8951
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064192,
+      "loss": 1.1135,
+      "step": 8952
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006418800000000001,
+      "loss": 1.1547,
+      "step": 8953
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064184,
+      "loss": 1.0278,
+      "step": 8954
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006418000000000001,
+      "loss": 1.2551,
+      "step": 8955
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064176,
+      "loss": 1.1089,
+      "step": 8956
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006417199999999999,
+      "loss": 1.1528,
+      "step": 8957
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064168,
+      "loss": 1.1684,
+      "step": 8958
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064164,
+      "loss": 1.061,
+      "step": 8959
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006416,
+      "loss": 1.1166,
+      "step": 8960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064156,
+      "loss": 1.1131,
+      "step": 8961
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064152,
+      "loss": 1.1046,
+      "step": 8962
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064148,
+      "loss": 1.0032,
+      "step": 8963
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006414400000000001,
+      "loss": 1.1035,
+      "step": 8964
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006414,
+      "loss": 1.119,
+      "step": 8965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006413600000000001,
+      "loss": 1.0874,
+      "step": 8966
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064132,
+      "loss": 1.1789,
+      "step": 8967
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006412799999999999,
+      "loss": 1.113,
+      "step": 8968
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006412400000000001,
+      "loss": 1.0577,
+      "step": 8969
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006412,
+      "loss": 1.0524,
+      "step": 8970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064116,
+      "loss": 1.1179,
+      "step": 8971
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064112,
+      "loss": 1.1042,
+      "step": 8972
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064108,
+      "loss": 1.1504,
+      "step": 8973
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006410400000000001,
+      "loss": 1.0521,
+      "step": 8974
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006410000000000001,
+      "loss": 1.0526,
+      "step": 8975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064096,
+      "loss": 1.1352,
+      "step": 8976
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064092,
+      "loss": 0.9815,
+      "step": 8977
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064088,
+      "loss": 1.0828,
+      "step": 8978
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006408399999999999,
+      "loss": 1.1496,
+      "step": 8979
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006408000000000001,
+      "loss": 1.0995,
+      "step": 8980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064076,
+      "loss": 1.1807,
+      "step": 8981
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064072,
+      "loss": 0.9914,
+      "step": 8982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064068,
+      "loss": 1.0845,
+      "step": 8983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064064,
+      "loss": 1.1331,
+      "step": 8984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006406,
+      "loss": 1.0886,
+      "step": 8985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064056,
+      "loss": 1.1198,
+      "step": 8986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064052,
+      "loss": 1.1134,
+      "step": 8987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064048,
+      "loss": 1.106,
+      "step": 8988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064044,
+      "loss": 1.0337,
+      "step": 8989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006404,
+      "loss": 1.0794,
+      "step": 8990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006403600000000001,
+      "loss": 1.23,
+      "step": 8991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064032,
+      "loss": 1.0169,
+      "step": 8992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064028,
+      "loss": 0.9916,
+      "step": 8993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00064024,
+      "loss": 1.0715,
+      "step": 8994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0006402000000000001,
+      "loss": 1.0486,
+      "step": 8995
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00064016,
+      "loss": 1.0402,
+      "step": 8996
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00064012,
+      "loss": 1.1568,
+      "step": 8997
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00064008,
+      "loss": 1.1521,
+      "step": 8998
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00064004,
+      "loss": 1.0947,
+      "step": 8999
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00064,
+      "loss": 1.1247,
+      "step": 9000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063996,
+      "loss": 1.199,
+      "step": 9001
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006399200000000001,
+      "loss": 1.1339,
+      "step": 9002
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063988,
+      "loss": 1.1356,
+      "step": 9003
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063984,
+      "loss": 1.1618,
+      "step": 9004
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006398,
+      "loss": 1.0698,
+      "step": 9005
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063976,
+      "loss": 1.0428,
+      "step": 9006
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063972,
+      "loss": 1.1684,
+      "step": 9007
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063968,
+      "loss": 1.0685,
+      "step": 9008
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063964,
+      "loss": 1.0802,
+      "step": 9009
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006395999999999999,
+      "loss": 1.0747,
+      "step": 9010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006395600000000001,
+      "loss": 1.0831,
+      "step": 9011
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063952,
+      "loss": 0.9912,
+      "step": 9012
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006394800000000001,
+      "loss": 1.1354,
+      "step": 9013
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063944,
+      "loss": 1.1969,
+      "step": 9014
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006393999999999999,
+      "loss": 1.1139,
+      "step": 9015
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006393600000000001,
+      "loss": 1.1183,
+      "step": 9016
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063932,
+      "loss": 1.0847,
+      "step": 9017
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063928,
+      "loss": 1.0256,
+      "step": 9018
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063924,
+      "loss": 1.1494,
+      "step": 9019
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006392,
+      "loss": 1.0886,
+      "step": 9020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006391599999999999,
+      "loss": 1.0743,
+      "step": 9021
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006391200000000001,
+      "loss": 1.0119,
+      "step": 9022
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063908,
+      "loss": 1.1597,
+      "step": 9023
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006390400000000001,
+      "loss": 1.0863,
+      "step": 9024
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000639,
+      "loss": 1.1355,
+      "step": 9025
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006389599999999999,
+      "loss": 1.1523,
+      "step": 9026
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006389200000000001,
+      "loss": 1.024,
+      "step": 9027
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063888,
+      "loss": 1.0573,
+      "step": 9028
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063884,
+      "loss": 1.061,
+      "step": 9029
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006388,
+      "loss": 1.077,
+      "step": 9030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063876,
+      "loss": 1.093,
+      "step": 9031
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063872,
+      "loss": 1.0852,
+      "step": 9032
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006386800000000001,
+      "loss": 1.1659,
+      "step": 9033
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063864,
+      "loss": 1.099,
+      "step": 9034
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006385999999999999,
+      "loss": 1.1463,
+      "step": 9035
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063856,
+      "loss": 1.1538,
+      "step": 9036
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063852,
+      "loss": 1.1483,
+      "step": 9037
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006384800000000001,
+      "loss": 1.1686,
+      "step": 9038
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063844,
+      "loss": 1.0751,
+      "step": 9039
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006384,
+      "loss": 1.0179,
+      "step": 9040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063836,
+      "loss": 1.1576,
+      "step": 9041
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006383200000000001,
+      "loss": 1.0833,
+      "step": 9042
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063828,
+      "loss": 1.1026,
+      "step": 9043
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006382400000000001,
+      "loss": 1.1546,
+      "step": 9044
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006382,
+      "loss": 1.0662,
+      "step": 9045
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006381599999999999,
+      "loss": 1.0903,
+      "step": 9046
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063812,
+      "loss": 1.0343,
+      "step": 9047
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063808,
+      "loss": 1.1375,
+      "step": 9048
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006380400000000001,
+      "loss": 1.1115,
+      "step": 9049
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000638,
+      "loss": 1.128,
+      "step": 9050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063796,
+      "loss": 1.0768,
+      "step": 9051
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063792,
+      "loss": 1.0886,
+      "step": 9052
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006378800000000001,
+      "loss": 1.0712,
+      "step": 9053
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063784,
+      "loss": 1.1603,
+      "step": 9054
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006378,
+      "loss": 1.1412,
+      "step": 9055
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063776,
+      "loss": 1.1586,
+      "step": 9056
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006377199999999999,
+      "loss": 1.1483,
+      "step": 9057
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006376800000000001,
+      "loss": 1.1063,
+      "step": 9058
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063764,
+      "loss": 1.1304,
+      "step": 9059
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006376,
+      "loss": 1.0944,
+      "step": 9060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063756,
+      "loss": 1.1283,
+      "step": 9061
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063752,
+      "loss": 1.0614,
+      "step": 9062
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006374800000000001,
+      "loss": 1.0135,
+      "step": 9063
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063744,
+      "loss": 1.0366,
+      "step": 9064
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006374,
+      "loss": 1.0901,
+      "step": 9065
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063736,
+      "loss": 1.1327,
+      "step": 9066
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063732,
+      "loss": 1.1388,
+      "step": 9067
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006372799999999999,
+      "loss": 1.0721,
+      "step": 9068
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006372400000000001,
+      "loss": 1.1263,
+      "step": 9069
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006372,
+      "loss": 1.1811,
+      "step": 9070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063716,
+      "loss": 1.0524,
+      "step": 9071
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063712,
+      "loss": 1.0863,
+      "step": 9072
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063708,
+      "loss": 1.1178,
+      "step": 9073
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006370400000000001,
+      "loss": 1.1488,
+      "step": 9074
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000637,
+      "loss": 1.0818,
+      "step": 9075
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063696,
+      "loss": 1.1472,
+      "step": 9076
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063692,
+      "loss": 1.0927,
+      "step": 9077
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063688,
+      "loss": 1.0489,
+      "step": 9078
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063684,
+      "loss": 1.1394,
+      "step": 9079
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006368000000000001,
+      "loss": 1.0788,
+      "step": 9080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063676,
+      "loss": 1.2387,
+      "step": 9081
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063672,
+      "loss": 1.0216,
+      "step": 9082
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063668,
+      "loss": 1.1819,
+      "step": 9083
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063664,
+      "loss": 1.1005,
+      "step": 9084
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006366000000000001,
+      "loss": 1.1336,
+      "step": 9085
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063656,
+      "loss": 1.1224,
+      "step": 9086
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063652,
+      "loss": 1.172,
+      "step": 9087
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063648,
+      "loss": 1.1303,
+      "step": 9088
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063644,
+      "loss": 1.0505,
+      "step": 9089
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006364,
+      "loss": 1.0904,
+      "step": 9090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006363600000000001,
+      "loss": 1.1527,
+      "step": 9091
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063632,
+      "loss": 1.1534,
+      "step": 9092
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006362799999999999,
+      "loss": 1.1221,
+      "step": 9093
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063624,
+      "loss": 1.2181,
+      "step": 9094
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006362,
+      "loss": 1.0866,
+      "step": 9095
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063616,
+      "loss": 1.0779,
+      "step": 9096
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063612,
+      "loss": 1.0876,
+      "step": 9097
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063608,
+      "loss": 1.1038,
+      "step": 9098
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063604,
+      "loss": 1.0828,
+      "step": 9099
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006360000000000001,
+      "loss": 1.1616,
+      "step": 9100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063596,
+      "loss": 1.0794,
+      "step": 9101
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006359200000000001,
+      "loss": 1.132,
+      "step": 9102
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063588,
+      "loss": 1.0954,
+      "step": 9103
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006358399999999999,
+      "loss": 1.0725,
+      "step": 9104
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006358000000000001,
+      "loss": 1.1214,
+      "step": 9105
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063576,
+      "loss": 1.1778,
+      "step": 9106
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063572,
+      "loss": 1.0306,
+      "step": 9107
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063568,
+      "loss": 1.1703,
+      "step": 9108
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063564,
+      "loss": 1.0927,
+      "step": 9109
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006356,
+      "loss": 1.1706,
+      "step": 9110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006355600000000001,
+      "loss": 1.1445,
+      "step": 9111
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063552,
+      "loss": 1.1174,
+      "step": 9112
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063548,
+      "loss": 1.0721,
+      "step": 9113
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063544,
+      "loss": 1.0767,
+      "step": 9114
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006353999999999999,
+      "loss": 1.19,
+      "step": 9115
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006353600000000001,
+      "loss": 1.123,
+      "step": 9116
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063532,
+      "loss": 1.1236,
+      "step": 9117
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063528,
+      "loss": 1.0171,
+      "step": 9118
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063524,
+      "loss": 1.1043,
+      "step": 9119
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006352,
+      "loss": 1.1061,
+      "step": 9120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063516,
+      "loss": 1.0779,
+      "step": 9121
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063512,
+      "loss": 1.1095,
+      "step": 9122
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063508,
+      "loss": 1.1034,
+      "step": 9123
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063504,
+      "loss": 1.0733,
+      "step": 9124
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000635,
+      "loss": 1.1254,
+      "step": 9125
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063496,
+      "loss": 1.1847,
+      "step": 9126
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006349200000000001,
+      "loss": 1.064,
+      "step": 9127
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063488,
+      "loss": 1.1091,
+      "step": 9128
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063484,
+      "loss": 1.102,
+      "step": 9129
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006348,
+      "loss": 1.096,
+      "step": 9130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006347600000000001,
+      "loss": 1.1415,
+      "step": 9131
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063472,
+      "loss": 1.1405,
+      "step": 9132
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063468,
+      "loss": 1.0857,
+      "step": 9133
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063464,
+      "loss": 1.1045,
+      "step": 9134
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006346,
+      "loss": 1.1105,
+      "step": 9135
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063456,
+      "loss": 1.038,
+      "step": 9136
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063452,
+      "loss": 1.1681,
+      "step": 9137
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006344800000000001,
+      "loss": 1.1035,
+      "step": 9138
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063444,
+      "loss": 1.1477,
+      "step": 9139
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006344,
+      "loss": 1.0977,
+      "step": 9140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063436,
+      "loss": 1.1359,
+      "step": 9141
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063432,
+      "loss": 1.1216,
+      "step": 9142
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063428,
+      "loss": 1.1763,
+      "step": 9143
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063424,
+      "loss": 1.1406,
+      "step": 9144
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006342,
+      "loss": 1.094,
+      "step": 9145
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006341599999999999,
+      "loss": 1.1124,
+      "step": 9146
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006341200000000001,
+      "loss": 1.1105,
+      "step": 9147
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063408,
+      "loss": 1.1409,
+      "step": 9148
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006340400000000001,
+      "loss": 1.1004,
+      "step": 9149
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000634,
+      "loss": 1.1064,
+      "step": 9150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006339599999999999,
+      "loss": 1.1584,
+      "step": 9151
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006339200000000001,
+      "loss": 1.071,
+      "step": 9152
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063388,
+      "loss": 1.0671,
+      "step": 9153
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063384,
+      "loss": 1.0969,
+      "step": 9154
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006338,
+      "loss": 1.025,
+      "step": 9155
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063376,
+      "loss": 1.0812,
+      "step": 9156
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006337199999999999,
+      "loss": 1.0549,
+      "step": 9157
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006336800000000001,
+      "loss": 1.1071,
+      "step": 9158
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063364,
+      "loss": 1.1204,
+      "step": 9159
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006336000000000001,
+      "loss": 1.0662,
+      "step": 9160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063356,
+      "loss": 1.1303,
+      "step": 9161
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006335199999999999,
+      "loss": 1.1305,
+      "step": 9162
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006334800000000001,
+      "loss": 1.1089,
+      "step": 9163
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063344,
+      "loss": 1.1027,
+      "step": 9164
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006334,
+      "loss": 1.1151,
+      "step": 9165
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063336,
+      "loss": 1.1383,
+      "step": 9166
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063332,
+      "loss": 1.0154,
+      "step": 9167
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063328,
+      "loss": 1.1706,
+      "step": 9168
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006332400000000001,
+      "loss": 1.0954,
+      "step": 9169
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006332,
+      "loss": 1.103,
+      "step": 9170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006331599999999999,
+      "loss": 1.0874,
+      "step": 9171
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063312,
+      "loss": 1.0423,
+      "step": 9172
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063308,
+      "loss": 1.1578,
+      "step": 9173
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006330400000000001,
+      "loss": 1.1201,
+      "step": 9174
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000633,
+      "loss": 1.0721,
+      "step": 9175
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063296,
+      "loss": 1.1147,
+      "step": 9176
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063292,
+      "loss": 1.0456,
+      "step": 9177
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063288,
+      "loss": 1.0501,
+      "step": 9178
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063284,
+      "loss": 1.0803,
+      "step": 9179
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006328000000000001,
+      "loss": 1.0831,
+      "step": 9180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063276,
+      "loss": 1.1404,
+      "step": 9181
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006327199999999999,
+      "loss": 1.0623,
+      "step": 9182
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063268,
+      "loss": 1.0833,
+      "step": 9183
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063264,
+      "loss": 1.1109,
+      "step": 9184
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006326000000000001,
+      "loss": 1.1354,
+      "step": 9185
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063256,
+      "loss": 1.0822,
+      "step": 9186
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063252,
+      "loss": 1.0769,
+      "step": 9187
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063248,
+      "loss": 1.0998,
+      "step": 9188
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006324400000000001,
+      "loss": 1.1028,
+      "step": 9189
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006324,
+      "loss": 1.0757,
+      "step": 9190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063236,
+      "loss": 1.1392,
+      "step": 9191
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063232,
+      "loss": 1.1577,
+      "step": 9192
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006322799999999999,
+      "loss": 1.0095,
+      "step": 9193
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006322400000000001,
+      "loss": 1.0035,
+      "step": 9194
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006322,
+      "loss": 1.153,
+      "step": 9195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006321600000000001,
+      "loss": 1.0556,
+      "step": 9196
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063212,
+      "loss": 1.1328,
+      "step": 9197
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063208,
+      "loss": 1.0598,
+      "step": 9198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063204,
+      "loss": 1.1203,
+      "step": 9199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000632,
+      "loss": 1.0862,
+      "step": 9200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063196,
+      "loss": 1.1826,
+      "step": 9201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063192,
+      "loss": 1.1057,
+      "step": 9202
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063188,
+      "loss": 1.0462,
+      "step": 9203
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006318399999999999,
+      "loss": 1.0865,
+      "step": 9204
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006318000000000001,
+      "loss": 1.1548,
+      "step": 9205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063176,
+      "loss": 1.0653,
+      "step": 9206
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063172,
+      "loss": 1.1088,
+      "step": 9207
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063168,
+      "loss": 1.0991,
+      "step": 9208
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063164,
+      "loss": 1.0321,
+      "step": 9209
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006316000000000001,
+      "loss": 1.0676,
+      "step": 9210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063156,
+      "loss": 1.0863,
+      "step": 9211
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063152,
+      "loss": 1.168,
+      "step": 9212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063148,
+      "loss": 1.0932,
+      "step": 9213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063144,
+      "loss": 1.0687,
+      "step": 9214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006314,
+      "loss": 1.099,
+      "step": 9215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006313600000000001,
+      "loss": 1.0722,
+      "step": 9216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063132,
+      "loss": 1.079,
+      "step": 9217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063128,
+      "loss": 1.0953,
+      "step": 9218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063124,
+      "loss": 1.127,
+      "step": 9219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006312,
+      "loss": 1.0763,
+      "step": 9220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006311600000000001,
+      "loss": 1.0893,
+      "step": 9221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063112,
+      "loss": 1.1777,
+      "step": 9222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063108,
+      "loss": 1.1134,
+      "step": 9223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063104,
+      "loss": 1.137,
+      "step": 9224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000631,
+      "loss": 1.06,
+      "step": 9225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063096,
+      "loss": 1.0522,
+      "step": 9226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006309200000000001,
+      "loss": 1.1152,
+      "step": 9227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063088,
+      "loss": 1.1323,
+      "step": 9228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006308399999999999,
+      "loss": 1.0966,
+      "step": 9229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006308,
+      "loss": 1.1827,
+      "step": 9230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063076,
+      "loss": 1.0452,
+      "step": 9231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063072,
+      "loss": 1.082,
+      "step": 9232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063068,
+      "loss": 1.118,
+      "step": 9233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063064,
+      "loss": 1.08,
+      "step": 9234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006306,
+      "loss": 1.1398,
+      "step": 9235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006305600000000001,
+      "loss": 1.0962,
+      "step": 9236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063052,
+      "loss": 1.0874,
+      "step": 9237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006304800000000001,
+      "loss": 1.1634,
+      "step": 9238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063044,
+      "loss": 1.0074,
+      "step": 9239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006303999999999999,
+      "loss": 1.1304,
+      "step": 9240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006303600000000001,
+      "loss": 1.0953,
+      "step": 9241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063032,
+      "loss": 1.0994,
+      "step": 9242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063028,
+      "loss": 1.0953,
+      "step": 9243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063024,
+      "loss": 1.0916,
+      "step": 9244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006302,
+      "loss": 1.1004,
+      "step": 9245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063016,
+      "loss": 1.074,
+      "step": 9246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0006301200000000001,
+      "loss": 1.0793,
+      "step": 9247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063008,
+      "loss": 1.0662,
+      "step": 9248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00063004,
+      "loss": 1.109,
+      "step": 9249
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00063,
+      "loss": 1.1528,
+      "step": 9250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006299599999999999,
+      "loss": 1.114,
+      "step": 9251
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006299200000000001,
+      "loss": 1.0797,
+      "step": 9252
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062988,
+      "loss": 1.1326,
+      "step": 9253
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062984,
+      "loss": 1.0466,
+      "step": 9254
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006298,
+      "loss": 1.0756,
+      "step": 9255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062976,
+      "loss": 1.0504,
+      "step": 9256
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062972,
+      "loss": 1.0942,
+      "step": 9257
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006296800000000001,
+      "loss": 1.0807,
+      "step": 9258
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062964,
+      "loss": 1.1559,
+      "step": 9259
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006296,
+      "loss": 1.1382,
+      "step": 9260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062956,
+      "loss": 1.1718,
+      "step": 9261
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062952,
+      "loss": 1.1543,
+      "step": 9262
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006294800000000001,
+      "loss": 1.0269,
+      "step": 9263
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062944,
+      "loss": 1.0555,
+      "step": 9264
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006294,
+      "loss": 1.1498,
+      "step": 9265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062936,
+      "loss": 1.1016,
+      "step": 9266
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062932,
+      "loss": 1.0568,
+      "step": 9267
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062928,
+      "loss": 1.0788,
+      "step": 9268
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062924,
+      "loss": 1.1018,
+      "step": 9269
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006292,
+      "loss": 1.1367,
+      "step": 9270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062916,
+      "loss": 1.1503,
+      "step": 9271
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062912,
+      "loss": 1.0681,
+      "step": 9272
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062908,
+      "loss": 1.0956,
+      "step": 9273
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006290400000000001,
+      "loss": 1.0889,
+      "step": 9274
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000629,
+      "loss": 1.119,
+      "step": 9275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062896,
+      "loss": 1.0348,
+      "step": 9276
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062892,
+      "loss": 1.0915,
+      "step": 9277
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062888,
+      "loss": 1.0305,
+      "step": 9278
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062884,
+      "loss": 1.0879,
+      "step": 9279
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006288,
+      "loss": 1.1239,
+      "step": 9280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062876,
+      "loss": 1.0845,
+      "step": 9281
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006287199999999999,
+      "loss": 1.1355,
+      "step": 9282
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006286800000000001,
+      "loss": 1.1866,
+      "step": 9283
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062864,
+      "loss": 1.1404,
+      "step": 9284
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006286000000000001,
+      "loss": 1.1295,
+      "step": 9285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062856,
+      "loss": 1.0888,
+      "step": 9286
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062852,
+      "loss": 1.0783,
+      "step": 9287
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006284800000000001,
+      "loss": 1.1333,
+      "step": 9288
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062844,
+      "loss": 1.0413,
+      "step": 9289
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006284,
+      "loss": 1.0306,
+      "step": 9290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062836,
+      "loss": 1.1135,
+      "step": 9291
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062832,
+      "loss": 1.1527,
+      "step": 9292
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006282799999999999,
+      "loss": 1.0877,
+      "step": 9293
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006282400000000001,
+      "loss": 1.0754,
+      "step": 9294
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006282,
+      "loss": 1.0728,
+      "step": 9295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006281600000000001,
+      "loss": 1.1368,
+      "step": 9296
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062812,
+      "loss": 1.086,
+      "step": 9297
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006280799999999999,
+      "loss": 1.061,
+      "step": 9298
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006280400000000001,
+      "loss": 1.0795,
+      "step": 9299
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000628,
+      "loss": 1.1592,
+      "step": 9300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062796,
+      "loss": 1.1069,
+      "step": 9301
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062792,
+      "loss": 1.1125,
+      "step": 9302
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062788,
+      "loss": 1.0639,
+      "step": 9303
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062784,
+      "loss": 1.0318,
+      "step": 9304
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006278000000000001,
+      "loss": 1.0975,
+      "step": 9305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062776,
+      "loss": 1.1402,
+      "step": 9306
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006277199999999999,
+      "loss": 1.0921,
+      "step": 9307
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062768,
+      "loss": 1.0951,
+      "step": 9308
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062764,
+      "loss": 1.0729,
+      "step": 9309
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006276000000000001,
+      "loss": 1.1711,
+      "step": 9310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062756,
+      "loss": 1.1134,
+      "step": 9311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062752,
+      "loss": 1.1454,
+      "step": 9312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062748,
+      "loss": 1.1355,
+      "step": 9313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062744,
+      "loss": 0.995,
+      "step": 9314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006274,
+      "loss": 1.1162,
+      "step": 9315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006273600000000001,
+      "loss": 1.0863,
+      "step": 9316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062732,
+      "loss": 1.0539,
+      "step": 9317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006272799999999999,
+      "loss": 1.0294,
+      "step": 9318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062724,
+      "loss": 1.0371,
+      "step": 9319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006272,
+      "loss": 1.0816,
+      "step": 9320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006271600000000001,
+      "loss": 1.1351,
+      "step": 9321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062712,
+      "loss": 1.1209,
+      "step": 9322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062708,
+      "loss": 1.1454,
+      "step": 9323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062704,
+      "loss": 1.1102,
+      "step": 9324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006270000000000001,
+      "loss": 1.028,
+      "step": 9325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062696,
+      "loss": 1.1784,
+      "step": 9326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062692,
+      "loss": 1.067,
+      "step": 9327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062688,
+      "loss": 1.1643,
+      "step": 9328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006268399999999999,
+      "loss": 1.078,
+      "step": 9329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006268000000000001,
+      "loss": 1.1495,
+      "step": 9330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062676,
+      "loss": 1.0858,
+      "step": 9331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006267200000000001,
+      "loss": 1.1957,
+      "step": 9332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062668,
+      "loss": 1.0321,
+      "step": 9333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062664,
+      "loss": 1.048,
+      "step": 9334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006266,
+      "loss": 1.1064,
+      "step": 9335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006265600000000001,
+      "loss": 1.1151,
+      "step": 9336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062652,
+      "loss": 1.1255,
+      "step": 9337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062648,
+      "loss": 1.0699,
+      "step": 9338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062644,
+      "loss": 1.0428,
+      "step": 9339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006263999999999999,
+      "loss": 1.1629,
+      "step": 9340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006263600000000001,
+      "loss": 1.0145,
+      "step": 9341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062632,
+      "loss": 1.1256,
+      "step": 9342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062628,
+      "loss": 1.0233,
+      "step": 9343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062624,
+      "loss": 1.1423,
+      "step": 9344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006262,
+      "loss": 1.0885,
+      "step": 9345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006261600000000001,
+      "loss": 1.1498,
+      "step": 9346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062612,
+      "loss": 1.107,
+      "step": 9347
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062608,
+      "loss": 1.1926,
+      "step": 9348
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062604,
+      "loss": 1.1236,
+      "step": 9349
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000626,
+      "loss": 1.0697,
+      "step": 9350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062596,
+      "loss": 1.1242,
+      "step": 9351
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006259200000000001,
+      "loss": 1.0601,
+      "step": 9352
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062588,
+      "loss": 1.1341,
+      "step": 9353
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062584,
+      "loss": 1.0912,
+      "step": 9354
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006258,
+      "loss": 1.1205,
+      "step": 9355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006257599999999999,
+      "loss": 1.2155,
+      "step": 9356
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006257200000000001,
+      "loss": 1.1441,
+      "step": 9357
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062568,
+      "loss": 1.1302,
+      "step": 9358
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062564,
+      "loss": 1.1111,
+      "step": 9359
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006256,
+      "loss": 1.1798,
+      "step": 9360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062556,
+      "loss": 1.1046,
+      "step": 9361
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062552,
+      "loss": 1.1308,
+      "step": 9362
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006254800000000001,
+      "loss": 1.106,
+      "step": 9363
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062544,
+      "loss": 1.0916,
+      "step": 9364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006254,
+      "loss": 1.0982,
+      "step": 9365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062536,
+      "loss": 1.0565,
+      "step": 9366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062532,
+      "loss": 1.123,
+      "step": 9367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062528,
+      "loss": 1.0276,
+      "step": 9368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062524,
+      "loss": 1.1126,
+      "step": 9369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006252,
+      "loss": 1.1937,
+      "step": 9370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062516,
+      "loss": 1.1155,
+      "step": 9371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006251200000000001,
+      "loss": 1.19,
+      "step": 9372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062508,
+      "loss": 1.1082,
+      "step": 9373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006250400000000001,
+      "loss": 1.082,
+      "step": 9374
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000625,
+      "loss": 1.1457,
+      "step": 9375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006249599999999999,
+      "loss": 1.1254,
+      "step": 9376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006249200000000001,
+      "loss": 1.0713,
+      "step": 9377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062488,
+      "loss": 1.1198,
+      "step": 9378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062484,
+      "loss": 1.0354,
+      "step": 9379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006248,
+      "loss": 1.1531,
+      "step": 9380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062476,
+      "loss": 1.0974,
+      "step": 9381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062472,
+      "loss": 1.0149,
+      "step": 9382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006246800000000001,
+      "loss": 1.0858,
+      "step": 9383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062464,
+      "loss": 1.0815,
+      "step": 9384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006246,
+      "loss": 1.1284,
+      "step": 9385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062456,
+      "loss": 1.1727,
+      "step": 9386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006245199999999999,
+      "loss": 1.1023,
+      "step": 9387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006244800000000001,
+      "loss": 1.0952,
+      "step": 9388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062444,
+      "loss": 1.0639,
+      "step": 9389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006244,
+      "loss": 1.0894,
+      "step": 9390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062436,
+      "loss": 1.0734,
+      "step": 9391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062432,
+      "loss": 1.0652,
+      "step": 9392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062428,
+      "loss": 1.0631,
+      "step": 9393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006242400000000001,
+      "loss": 1.0495,
+      "step": 9394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006242,
+      "loss": 1.0458,
+      "step": 9395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062416,
+      "loss": 1.1879,
+      "step": 9396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062412,
+      "loss": 1.0946,
+      "step": 9397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062408,
+      "loss": 1.0805,
+      "step": 9398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006240400000000001,
+      "loss": 1.086,
+      "step": 9399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000624,
+      "loss": 1.0548,
+      "step": 9400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062396,
+      "loss": 1.0537,
+      "step": 9401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062392,
+      "loss": 1.0712,
+      "step": 9402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062388,
+      "loss": 1.126,
+      "step": 9403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062384,
+      "loss": 1.109,
+      "step": 9404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006238,
+      "loss": 1.1202,
+      "step": 9405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062376,
+      "loss": 1.1163,
+      "step": 9406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062372,
+      "loss": 1.1057,
+      "step": 9407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062368,
+      "loss": 1.0911,
+      "step": 9408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062364,
+      "loss": 1.0154,
+      "step": 9409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006236000000000001,
+      "loss": 1.1161,
+      "step": 9410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062356,
+      "loss": 1.1603,
+      "step": 9411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062352,
+      "loss": 1.0889,
+      "step": 9412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062348,
+      "loss": 1.1174,
+      "step": 9413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062344,
+      "loss": 1.1296,
+      "step": 9414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006234,
+      "loss": 1.1461,
+      "step": 9415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062336,
+      "loss": 1.1251,
+      "step": 9416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062332,
+      "loss": 1.1313,
+      "step": 9417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006232799999999999,
+      "loss": 1.1063,
+      "step": 9418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006232400000000001,
+      "loss": 1.1575,
+      "step": 9419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006232,
+      "loss": 1.1046,
+      "step": 9420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006231600000000001,
+      "loss": 1.148,
+      "step": 9421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062312,
+      "loss": 1.0549,
+      "step": 9422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062308,
+      "loss": 1.0549,
+      "step": 9423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062304,
+      "loss": 1.0869,
+      "step": 9424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000623,
+      "loss": 1.1759,
+      "step": 9425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062296,
+      "loss": 1.106,
+      "step": 9426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062292,
+      "loss": 1.0719,
+      "step": 9427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062288,
+      "loss": 1.0958,
+      "step": 9428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006228399999999999,
+      "loss": 1.0512,
+      "step": 9429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006228000000000001,
+      "loss": 1.1497,
+      "step": 9430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062276,
+      "loss": 1.0264,
+      "step": 9431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006227200000000001,
+      "loss": 1.1751,
+      "step": 9432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062268,
+      "loss": 1.1948,
+      "step": 9433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006226399999999999,
+      "loss": 1.0398,
+      "step": 9434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006226000000000001,
+      "loss": 1.0599,
+      "step": 9435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062256,
+      "loss": 1.1194,
+      "step": 9436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062252,
+      "loss": 1.1307,
+      "step": 9437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062248,
+      "loss": 1.1168,
+      "step": 9438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062244,
+      "loss": 1.0946,
+      "step": 9439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006224,
+      "loss": 1.155,
+      "step": 9440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006223600000000001,
+      "loss": 1.09,
+      "step": 9441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062232,
+      "loss": 1.0908,
+      "step": 9442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006222800000000001,
+      "loss": 1.0924,
+      "step": 9443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062224,
+      "loss": 1.025,
+      "step": 9444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006221999999999999,
+      "loss": 1.1744,
+      "step": 9445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006221600000000001,
+      "loss": 1.0758,
+      "step": 9446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062212,
+      "loss": 1.0421,
+      "step": 9447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062208,
+      "loss": 1.0776,
+      "step": 9448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062204,
+      "loss": 1.1036,
+      "step": 9449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000622,
+      "loss": 1.0624,
+      "step": 9450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062196,
+      "loss": 1.0976,
+      "step": 9451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006219200000000001,
+      "loss": 1.0017,
+      "step": 9452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062188,
+      "loss": 1.1434,
+      "step": 9453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006218399999999999,
+      "loss": 1.1902,
+      "step": 9454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006218,
+      "loss": 1.0862,
+      "step": 9455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062176,
+      "loss": 1.1563,
+      "step": 9456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006217200000000001,
+      "loss": 1.0166,
+      "step": 9457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062168,
+      "loss": 1.0704,
+      "step": 9458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062164,
+      "loss": 1.0628,
+      "step": 9459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006216,
+      "loss": 0.9836,
+      "step": 9460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006215600000000001,
+      "loss": 1.0739,
+      "step": 9461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062152,
+      "loss": 1.1499,
+      "step": 9462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062148,
+      "loss": 1.0992,
+      "step": 9463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062144,
+      "loss": 1.0564,
+      "step": 9464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006213999999999999,
+      "loss": 1.0815,
+      "step": 9465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006213600000000001,
+      "loss": 1.1283,
+      "step": 9466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062132,
+      "loss": 1.0781,
+      "step": 9467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006212800000000001,
+      "loss": 1.115,
+      "step": 9468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062124,
+      "loss": 1.1808,
+      "step": 9469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006212,
+      "loss": 1.0604,
+      "step": 9470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062116,
+      "loss": 1.0809,
+      "step": 9471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006211200000000001,
+      "loss": 1.0913,
+      "step": 9472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062108,
+      "loss": 1.1236,
+      "step": 9473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062104,
+      "loss": 1.0584,
+      "step": 9474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000621,
+      "loss": 1.0605,
+      "step": 9475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006209599999999999,
+      "loss": 1.0719,
+      "step": 9476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006209200000000001,
+      "loss": 1.0846,
+      "step": 9477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062088,
+      "loss": 1.0775,
+      "step": 9478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062084,
+      "loss": 1.1922,
+      "step": 9479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006208,
+      "loss": 1.1322,
+      "step": 9480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062076,
+      "loss": 0.9813,
+      "step": 9481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006207200000000001,
+      "loss": 1.1237,
+      "step": 9482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062068,
+      "loss": 0.9853,
+      "step": 9483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062064,
+      "loss": 1.0696,
+      "step": 9484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006206,
+      "loss": 1.1062,
+      "step": 9485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062056,
+      "loss": 1.0588,
+      "step": 9486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062052,
+      "loss": 1.1935,
+      "step": 9487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006204800000000001,
+      "loss": 1.1176,
+      "step": 9488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062044,
+      "loss": 1.0939,
+      "step": 9489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006204,
+      "loss": 1.116,
+      "step": 9490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062036,
+      "loss": 0.9966,
+      "step": 9491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006203199999999999,
+      "loss": 1.1286,
+      "step": 9492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006202800000000001,
+      "loss": 1.091,
+      "step": 9493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062024,
+      "loss": 1.0395,
+      "step": 9494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006202,
+      "loss": 1.0831,
+      "step": 9495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062016,
+      "loss": 1.1753,
+      "step": 9496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062012,
+      "loss": 1.1198,
+      "step": 9497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062008,
+      "loss": 1.1267,
+      "step": 9498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0006200400000000001,
+      "loss": 1.0778,
+      "step": 9499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00062,
+      "loss": 1.1139,
+      "step": 9500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00061996,
+      "loss": 1.1944,
+      "step": 9501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00061992,
+      "loss": 1.0986,
+      "step": 9502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061988,
+      "loss": 1.0399,
+      "step": 9503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061984,
+      "loss": 1.0737,
+      "step": 9504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006198,
+      "loss": 1.0549,
+      "step": 9505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061976,
+      "loss": 1.0587,
+      "step": 9506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061972,
+      "loss": 1.0797,
+      "step": 9507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006196800000000001,
+      "loss": 1.0882,
+      "step": 9508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061964,
+      "loss": 1.0931,
+      "step": 9509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006196000000000001,
+      "loss": 1.0525,
+      "step": 9510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061956,
+      "loss": 1.1107,
+      "step": 9511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006195199999999999,
+      "loss": 1.0026,
+      "step": 9512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061948,
+      "loss": 1.123,
+      "step": 9513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061944,
+      "loss": 1.0807,
+      "step": 9514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006194,
+      "loss": 1.107,
+      "step": 9515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061936,
+      "loss": 1.0831,
+      "step": 9516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061932,
+      "loss": 1.119,
+      "step": 9517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061928,
+      "loss": 1.1019,
+      "step": 9518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006192400000000001,
+      "loss": 1.0676,
+      "step": 9519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006192,
+      "loss": 1.0405,
+      "step": 9520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006191600000000001,
+      "loss": 1.1402,
+      "step": 9521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061912,
+      "loss": 1.122,
+      "step": 9522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006190799999999999,
+      "loss": 1.0401,
+      "step": 9523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006190400000000001,
+      "loss": 1.0489,
+      "step": 9524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000619,
+      "loss": 1.0985,
+      "step": 9525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061896,
+      "loss": 1.0339,
+      "step": 9526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061892,
+      "loss": 1.0783,
+      "step": 9527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061888,
+      "loss": 1.0193,
+      "step": 9528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061884,
+      "loss": 1.073,
+      "step": 9529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006188000000000001,
+      "loss": 1.1333,
+      "step": 9530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061876,
+      "loss": 1.0296,
+      "step": 9531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061872,
+      "loss": 1.0621,
+      "step": 9532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061868,
+      "loss": 1.0609,
+      "step": 9533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006186399999999999,
+      "loss": 1.0966,
+      "step": 9534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006186000000000001,
+      "loss": 1.0825,
+      "step": 9535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061856,
+      "loss": 1.1133,
+      "step": 9536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061852,
+      "loss": 1.0369,
+      "step": 9537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061848,
+      "loss": 1.117,
+      "step": 9538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061844,
+      "loss": 1.1508,
+      "step": 9539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006184,
+      "loss": 1.0685,
+      "step": 9540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061836,
+      "loss": 1.0813,
+      "step": 9541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061832,
+      "loss": 1.1243,
+      "step": 9542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061828,
+      "loss": 1.1181,
+      "step": 9543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061824,
+      "loss": 1.0601,
+      "step": 9544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006182,
+      "loss": 1.1225,
+      "step": 9545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006181600000000001,
+      "loss": 1.0546,
+      "step": 9546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061812,
+      "loss": 0.9967,
+      "step": 9547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061808,
+      "loss": 1.0635,
+      "step": 9548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061804,
+      "loss": 1.0262,
+      "step": 9549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006180000000000001,
+      "loss": 1.0406,
+      "step": 9550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061796,
+      "loss": 1.1068,
+      "step": 9551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061792,
+      "loss": 1.0464,
+      "step": 9552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061788,
+      "loss": 1.0751,
+      "step": 9553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006178399999999999,
+      "loss": 1.0701,
+      "step": 9554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006178000000000001,
+      "loss": 1.1058,
+      "step": 9555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061776,
+      "loss": 1.1229,
+      "step": 9556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006177200000000001,
+      "loss": 1.1127,
+      "step": 9557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061768,
+      "loss": 1.0894,
+      "step": 9558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061764,
+      "loss": 1.108,
+      "step": 9559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006176,
+      "loss": 1.0842,
+      "step": 9560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061756,
+      "loss": 1.1163,
+      "step": 9561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061752,
+      "loss": 1.1,
+      "step": 9562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061748,
+      "loss": 1.0696,
+      "step": 9563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061744,
+      "loss": 1.1211,
+      "step": 9564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006173999999999999,
+      "loss": 1.0941,
+      "step": 9565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006173600000000001,
+      "loss": 1.0908,
+      "step": 9566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061732,
+      "loss": 1.0803,
+      "step": 9567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006172800000000001,
+      "loss": 1.033,
+      "step": 9568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061724,
+      "loss": 1.1124,
+      "step": 9569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006171999999999999,
+      "loss": 1.1344,
+      "step": 9570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006171600000000001,
+      "loss": 1.1591,
+      "step": 9571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061712,
+      "loss": 1.1742,
+      "step": 9572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061708,
+      "loss": 1.0633,
+      "step": 9573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061704,
+      "loss": 1.0893,
+      "step": 9574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000617,
+      "loss": 1.0627,
+      "step": 9575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061696,
+      "loss": 1.0984,
+      "step": 9576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006169200000000001,
+      "loss": 1.086,
+      "step": 9577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061688,
+      "loss": 1.1441,
+      "step": 9578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006168400000000001,
+      "loss": 1.1409,
+      "step": 9579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006168,
+      "loss": 1.0892,
+      "step": 9580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006167599999999999,
+      "loss": 1.101,
+      "step": 9581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006167200000000001,
+      "loss": 1.0147,
+      "step": 9582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061668,
+      "loss": 1.1163,
+      "step": 9583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061664,
+      "loss": 1.0209,
+      "step": 9584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006166,
+      "loss": 1.0026,
+      "step": 9585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061656,
+      "loss": 1.0998,
+      "step": 9586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061652,
+      "loss": 1.1054,
+      "step": 9587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006164800000000001,
+      "loss": 1.0647,
+      "step": 9588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061644,
+      "loss": 1.1098,
+      "step": 9589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006163999999999999,
+      "loss": 1.1511,
+      "step": 9590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061636,
+      "loss": 1.1187,
+      "step": 9591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061632,
+      "loss": 1.0853,
+      "step": 9592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006162800000000001,
+      "loss": 1.1389,
+      "step": 9593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061624,
+      "loss": 1.0662,
+      "step": 9594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006162,
+      "loss": 1.0785,
+      "step": 9595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061616,
+      "loss": 1.1165,
+      "step": 9596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006161200000000001,
+      "loss": 1.0951,
+      "step": 9597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061608,
+      "loss": 1.0829,
+      "step": 9598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006160400000000001,
+      "loss": 1.0422,
+      "step": 9599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000616,
+      "loss": 1.0663,
+      "step": 9600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006159599999999999,
+      "loss": 1.0751,
+      "step": 9601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061592,
+      "loss": 1.0413,
+      "step": 9602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061588,
+      "loss": 1.2308,
+      "step": 9603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006158400000000001,
+      "loss": 1.033,
+      "step": 9604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006158,
+      "loss": 1.0913,
+      "step": 9605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061576,
+      "loss": 1.0836,
+      "step": 9606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061572,
+      "loss": 1.101,
+      "step": 9607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006156800000000001,
+      "loss": 1.0982,
+      "step": 9608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061564,
+      "loss": 1.0931,
+      "step": 9609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006156,
+      "loss": 1.1534,
+      "step": 9610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061556,
+      "loss": 1.0174,
+      "step": 9611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006155199999999999,
+      "loss": 1.058,
+      "step": 9612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006154800000000001,
+      "loss": 1.1076,
+      "step": 9613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061544,
+      "loss": 1.0253,
+      "step": 9614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006154,
+      "loss": 1.1523,
+      "step": 9615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061536,
+      "loss": 1.0995,
+      "step": 9616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061532,
+      "loss": 1.1528,
+      "step": 9617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006152800000000001,
+      "loss": 1.0384,
+      "step": 9618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061524,
+      "loss": 1.0649,
+      "step": 9619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006152,
+      "loss": 1.066,
+      "step": 9620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061516,
+      "loss": 1.1281,
+      "step": 9621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061512,
+      "loss": 1.1709,
+      "step": 9622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006150799999999999,
+      "loss": 1.0945,
+      "step": 9623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006150400000000001,
+      "loss": 1.1358,
+      "step": 9624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000615,
+      "loss": 1.0815,
+      "step": 9625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061496,
+      "loss": 1.0426,
+      "step": 9626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061492,
+      "loss": 1.0117,
+      "step": 9627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061488,
+      "loss": 1.099,
+      "step": 9628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006148400000000001,
+      "loss": 1.1051,
+      "step": 9629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006148,
+      "loss": 1.058,
+      "step": 9630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061476,
+      "loss": 1.1193,
+      "step": 9631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061472,
+      "loss": 1.0771,
+      "step": 9632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061468,
+      "loss": 0.9452,
+      "step": 9633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061464,
+      "loss": 1.0783,
+      "step": 9634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006146000000000001,
+      "loss": 1.0864,
+      "step": 9635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061456,
+      "loss": 1.0887,
+      "step": 9636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061452,
+      "loss": 1.0837,
+      "step": 9637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061448,
+      "loss": 1.1003,
+      "step": 9638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061444,
+      "loss": 1.0419,
+      "step": 9639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006144,
+      "loss": 1.0046,
+      "step": 9640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061436,
+      "loss": 1.0491,
+      "step": 9641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061432,
+      "loss": 1.16,
+      "step": 9642
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061428,
+      "loss": 1.0654,
+      "step": 9643
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006142400000000001,
+      "loss": 1.0923,
+      "step": 9644
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006142,
+      "loss": 1.1211,
+      "step": 9645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006141600000000001,
+      "loss": 1.0331,
+      "step": 9646
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061412,
+      "loss": 1.1144,
+      "step": 9647
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006140799999999999,
+      "loss": 1.0696,
+      "step": 9648
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061404,
+      "loss": 1.1373,
+      "step": 9649
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000614,
+      "loss": 1.0832,
+      "step": 9650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061396,
+      "loss": 1.2051,
+      "step": 9651
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061392,
+      "loss": 1.1588,
+      "step": 9652
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061388,
+      "loss": 1.0783,
+      "step": 9653
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061384,
+      "loss": 1.0221,
+      "step": 9654
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006138000000000001,
+      "loss": 1.0736,
+      "step": 9655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061376,
+      "loss": 1.077,
+      "step": 9656
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006137200000000001,
+      "loss": 1.1004,
+      "step": 9657
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061368,
+      "loss": 1.0166,
+      "step": 9658
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006136399999999999,
+      "loss": 1.1288,
+      "step": 9659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006136000000000001,
+      "loss": 1.0576,
+      "step": 9660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061356,
+      "loss": 1.0267,
+      "step": 9661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061352,
+      "loss": 1.041,
+      "step": 9662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061348,
+      "loss": 1.0845,
+      "step": 9663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061344,
+      "loss": 1.0354,
+      "step": 9664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006134,
+      "loss": 1.1359,
+      "step": 9665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006133600000000001,
+      "loss": 1.0879,
+      "step": 9666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061332,
+      "loss": 1.0751,
+      "step": 9667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061328,
+      "loss": 1.1135,
+      "step": 9668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061324,
+      "loss": 1.0761,
+      "step": 9669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006131999999999999,
+      "loss": 1.0449,
+      "step": 9670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006131600000000001,
+      "loss": 1.096,
+      "step": 9671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061312,
+      "loss": 1.0308,
+      "step": 9672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061308,
+      "loss": 1.0825,
+      "step": 9673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061304,
+      "loss": 1.1154,
+      "step": 9674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000613,
+      "loss": 1.1243,
+      "step": 9675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061296,
+      "loss": 1.1329,
+      "step": 9676
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061292,
+      "loss": 1.1553,
+      "step": 9677
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061288,
+      "loss": 1.1071,
+      "step": 9678
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061284,
+      "loss": 1.0429,
+      "step": 9679
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006128,
+      "loss": 1.1144,
+      "step": 9680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061276,
+      "loss": 1.0958,
+      "step": 9681
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006127200000000001,
+      "loss": 1.0424,
+      "step": 9682
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061268,
+      "loss": 1.039,
+      "step": 9683
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061264,
+      "loss": 1.0577,
+      "step": 9684
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006126,
+      "loss": 1.107,
+      "step": 9685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006125600000000001,
+      "loss": 1.0791,
+      "step": 9686
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061252,
+      "loss": 1.1243,
+      "step": 9687
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061248,
+      "loss": 1.0286,
+      "step": 9688
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061244,
+      "loss": 1.0712,
+      "step": 9689
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006124,
+      "loss": 0.9749,
+      "step": 9690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061236,
+      "loss": 1.0917,
+      "step": 9691
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061232,
+      "loss": 1.1196,
+      "step": 9692
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006122800000000001,
+      "loss": 1.074,
+      "step": 9693
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061224,
+      "loss": 1.1497,
+      "step": 9694
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006122,
+      "loss": 1.1087,
+      "step": 9695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061216,
+      "loss": 1.0569,
+      "step": 9696
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061212,
+      "loss": 1.1137,
+      "step": 9697
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061208,
+      "loss": 1.1442,
+      "step": 9698
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061204,
+      "loss": 0.9817,
+      "step": 9699
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000612,
+      "loss": 1.026,
+      "step": 9700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006119599999999999,
+      "loss": 1.032,
+      "step": 9701
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006119200000000001,
+      "loss": 1.0944,
+      "step": 9702
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061188,
+      "loss": 1.0523,
+      "step": 9703
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006118400000000001,
+      "loss": 1.0517,
+      "step": 9704
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006118,
+      "loss": 1.1248,
+      "step": 9705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006117599999999999,
+      "loss": 1.071,
+      "step": 9706
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006117200000000001,
+      "loss": 1.1061,
+      "step": 9707
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061168,
+      "loss": 1.1006,
+      "step": 9708
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061164,
+      "loss": 1.1373,
+      "step": 9709
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006116,
+      "loss": 1.173,
+      "step": 9710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061156,
+      "loss": 1.1113,
+      "step": 9711
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006115199999999999,
+      "loss": 1.1142,
+      "step": 9712
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006114800000000001,
+      "loss": 1.1127,
+      "step": 9713
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061144,
+      "loss": 1.0993,
+      "step": 9714
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006114000000000001,
+      "loss": 1.0656,
+      "step": 9715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061136,
+      "loss": 1.0929,
+      "step": 9716
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006113199999999999,
+      "loss": 1.054,
+      "step": 9717
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006112800000000001,
+      "loss": 1.0199,
+      "step": 9718
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061124,
+      "loss": 1.1715,
+      "step": 9719
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006112,
+      "loss": 1.0469,
+      "step": 9720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061116,
+      "loss": 1.0907,
+      "step": 9721
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061112,
+      "loss": 1.1613,
+      "step": 9722
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061108,
+      "loss": 1.0218,
+      "step": 9723
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006110400000000001,
+      "loss": 1.026,
+      "step": 9724
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000611,
+      "loss": 1.129,
+      "step": 9725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006109599999999999,
+      "loss": 1.1016,
+      "step": 9726
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061092,
+      "loss": 1.0758,
+      "step": 9727
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061088,
+      "loss": 1.087,
+      "step": 9728
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006108400000000001,
+      "loss": 1.1365,
+      "step": 9729
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006108,
+      "loss": 1.0581,
+      "step": 9730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061076,
+      "loss": 1.1289,
+      "step": 9731
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061072,
+      "loss": 1.1049,
+      "step": 9732
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006106800000000001,
+      "loss": 1.0608,
+      "step": 9733
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061064,
+      "loss": 1.098,
+      "step": 9734
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006106000000000001,
+      "loss": 1.0844,
+      "step": 9735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061056,
+      "loss": 1.0687,
+      "step": 9736
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006105199999999999,
+      "loss": 1.0908,
+      "step": 9737
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061048,
+      "loss": 1.1112,
+      "step": 9738
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061044,
+      "loss": 1.0625,
+      "step": 9739
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006104000000000001,
+      "loss": 1.0075,
+      "step": 9740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061036,
+      "loss": 1.1059,
+      "step": 9741
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061032,
+      "loss": 0.9774,
+      "step": 9742
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061028,
+      "loss": 1.1085,
+      "step": 9743
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006102400000000001,
+      "loss": 1.1722,
+      "step": 9744
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006102,
+      "loss": 1.0091,
+      "step": 9745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061016,
+      "loss": 1.0728,
+      "step": 9746
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061012,
+      "loss": 1.0262,
+      "step": 9747
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006100799999999999,
+      "loss": 1.075,
+      "step": 9748
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006100400000000001,
+      "loss": 1.0701,
+      "step": 9749
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00061,
+      "loss": 1.1106,
+      "step": 9750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00060996,
+      "loss": 1.0764,
+      "step": 9751
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00060992,
+      "loss": 0.9847,
+      "step": 9752
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00060988,
+      "loss": 1.1035,
+      "step": 9753
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006098400000000001,
+      "loss": 1.0514,
+      "step": 9754
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0006098,
+      "loss": 1.0129,
+      "step": 9755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060976,
+      "loss": 1.0559,
+      "step": 9756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060972,
+      "loss": 1.0612,
+      "step": 9757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060968,
+      "loss": 1.0448,
+      "step": 9758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006096399999999999,
+      "loss": 1.0282,
+      "step": 9759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006096000000000001,
+      "loss": 1.1109,
+      "step": 9760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060956,
+      "loss": 1.123,
+      "step": 9761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060952,
+      "loss": 1.1024,
+      "step": 9762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060948,
+      "loss": 1.127,
+      "step": 9763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060944,
+      "loss": 1.0877,
+      "step": 9764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006094000000000001,
+      "loss": 1.104,
+      "step": 9765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060936,
+      "loss": 1.0973,
+      "step": 9766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060932,
+      "loss": 1.1058,
+      "step": 9767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060928,
+      "loss": 1.1038,
+      "step": 9768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060924,
+      "loss": 1.1881,
+      "step": 9769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006092,
+      "loss": 1.0848,
+      "step": 9770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006091600000000001,
+      "loss": 1.1536,
+      "step": 9771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060912,
+      "loss": 1.1194,
+      "step": 9772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060908,
+      "loss": 1.1074,
+      "step": 9773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060904,
+      "loss": 1.0596,
+      "step": 9774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000609,
+      "loss": 1.0496,
+      "step": 9775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060896,
+      "loss": 1.117,
+      "step": 9776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060892,
+      "loss": 1.1374,
+      "step": 9777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060888,
+      "loss": 1.025,
+      "step": 9778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060884,
+      "loss": 1.1799,
+      "step": 9779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006088,
+      "loss": 1.059,
+      "step": 9780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060876,
+      "loss": 1.0593,
+      "step": 9781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006087200000000001,
+      "loss": 1.0144,
+      "step": 9782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060868,
+      "loss": 1.0926,
+      "step": 9783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006086399999999999,
+      "loss": 1.1465,
+      "step": 9784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006086,
+      "loss": 1.0428,
+      "step": 9785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060856,
+      "loss": 1.0406,
+      "step": 9786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060852,
+      "loss": 1.1014,
+      "step": 9787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060848,
+      "loss": 1.0378,
+      "step": 9788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060844,
+      "loss": 0.9779,
+      "step": 9789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006084,
+      "loss": 1.0552,
+      "step": 9790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006083600000000001,
+      "loss": 1.0125,
+      "step": 9791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060832,
+      "loss": 1.0798,
+      "step": 9792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006082800000000001,
+      "loss": 1.0675,
+      "step": 9793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060824,
+      "loss": 1.0972,
+      "step": 9794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006081999999999999,
+      "loss": 1.0753,
+      "step": 9795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006081600000000001,
+      "loss": 1.1429,
+      "step": 9796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060812,
+      "loss": 1.0695,
+      "step": 9797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060808,
+      "loss": 1.1457,
+      "step": 9798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060804,
+      "loss": 1.1229,
+      "step": 9799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000608,
+      "loss": 1.0725,
+      "step": 9800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006079599999999999,
+      "loss": 1.0788,
+      "step": 9801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006079200000000001,
+      "loss": 1.0586,
+      "step": 9802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060788,
+      "loss": 1.1841,
+      "step": 9803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060784,
+      "loss": 1.073,
+      "step": 9804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006078,
+      "loss": 1.0849,
+      "step": 9805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006077599999999999,
+      "loss": 1.1233,
+      "step": 9806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006077200000000001,
+      "loss": 1.1034,
+      "step": 9807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060768,
+      "loss": 1.0589,
+      "step": 9808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060764,
+      "loss": 1.0981,
+      "step": 9809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006076,
+      "loss": 1.1172,
+      "step": 9810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060756,
+      "loss": 1.0656,
+      "step": 9811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060752,
+      "loss": 1.1061,
+      "step": 9812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006074800000000001,
+      "loss": 1.0942,
+      "step": 9813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060744,
+      "loss": 1.1595,
+      "step": 9814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006074,
+      "loss": 1.0532,
+      "step": 9815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060736,
+      "loss": 1.1046,
+      "step": 9816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060732,
+      "loss": 1.0369,
+      "step": 9817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006072800000000001,
+      "loss": 1.1267,
+      "step": 9818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060724,
+      "loss": 1.0367,
+      "step": 9819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006072,
+      "loss": 1.1111,
+      "step": 9820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060716,
+      "loss": 1.0757,
+      "step": 9821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006071200000000001,
+      "loss": 1.0717,
+      "step": 9822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060708,
+      "loss": 1.115,
+      "step": 9823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060704,
+      "loss": 1.1006,
+      "step": 9824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000607,
+      "loss": 1.1399,
+      "step": 9825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060696,
+      "loss": 1.1263,
+      "step": 9826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060692,
+      "loss": 1.0581,
+      "step": 9827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060688,
+      "loss": 1.1246,
+      "step": 9828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006068400000000001,
+      "loss": 1.102,
+      "step": 9829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006068,
+      "loss": 1.1127,
+      "step": 9830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060676,
+      "loss": 0.9917,
+      "step": 9831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060672,
+      "loss": 1.1222,
+      "step": 9832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060668,
+      "loss": 1.0967,
+      "step": 9833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060664,
+      "loss": 1.0763,
+      "step": 9834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006066,
+      "loss": 1.0269,
+      "step": 9835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060656,
+      "loss": 0.9952,
+      "step": 9836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006065199999999999,
+      "loss": 1.1014,
+      "step": 9837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006064800000000001,
+      "loss": 1.0615,
+      "step": 9838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060644,
+      "loss": 1.0992,
+      "step": 9839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006064000000000001,
+      "loss": 1.12,
+      "step": 9840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060636,
+      "loss": 1.0561,
+      "step": 9841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060632,
+      "loss": 1.1221,
+      "step": 9842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006062800000000001,
+      "loss": 1.1259,
+      "step": 9843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060624,
+      "loss": 1.0876,
+      "step": 9844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006062,
+      "loss": 1.13,
+      "step": 9845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060616,
+      "loss": 1.0833,
+      "step": 9846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060612,
+      "loss": 1.0593,
+      "step": 9847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006060799999999999,
+      "loss": 1.0745,
+      "step": 9848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006060400000000001,
+      "loss": 1.0621,
+      "step": 9849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000606,
+      "loss": 1.0956,
+      "step": 9850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006059600000000001,
+      "loss": 1.1514,
+      "step": 9851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060592,
+      "loss": 1.0156,
+      "step": 9852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006058799999999999,
+      "loss": 1.0155,
+      "step": 9853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006058400000000001,
+      "loss": 1.1683,
+      "step": 9854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006058,
+      "loss": 1.0818,
+      "step": 9855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060576,
+      "loss": 1.1428,
+      "step": 9856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060572,
+      "loss": 1.0188,
+      "step": 9857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060568,
+      "loss": 1.0592,
+      "step": 9858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060564,
+      "loss": 1.0342,
+      "step": 9859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006056000000000001,
+      "loss": 1.1097,
+      "step": 9860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060556,
+      "loss": 1.1423,
+      "step": 9861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006055199999999999,
+      "loss": 1.0854,
+      "step": 9862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060548,
+      "loss": 1.0503,
+      "step": 9863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060544,
+      "loss": 1.0315,
+      "step": 9864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006054000000000001,
+      "loss": 1.1104,
+      "step": 9865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060536,
+      "loss": 1.1084,
+      "step": 9866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060532,
+      "loss": 1.1711,
+      "step": 9867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060528,
+      "loss": 1.0143,
+      "step": 9868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060524,
+      "loss": 1.0359,
+      "step": 9869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006052,
+      "loss": 1.0363,
+      "step": 9870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006051600000000001,
+      "loss": 1.0315,
+      "step": 9871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060512,
+      "loss": 1.0507,
+      "step": 9872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006050799999999999,
+      "loss": 1.1175,
+      "step": 9873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060504,
+      "loss": 1.041,
+      "step": 9874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000605,
+      "loss": 1.1088,
+      "step": 9875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006049600000000001,
+      "loss": 1.1011,
+      "step": 9876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060492,
+      "loss": 1.0258,
+      "step": 9877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060488,
+      "loss": 1.0242,
+      "step": 9878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060484,
+      "loss": 0.9946,
+      "step": 9879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006048000000000001,
+      "loss": 1.1255,
+      "step": 9880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060476,
+      "loss": 1.093,
+      "step": 9881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060472,
+      "loss": 1.0898,
+      "step": 9882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060468,
+      "loss": 1.2109,
+      "step": 9883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006046399999999999,
+      "loss": 1.0752,
+      "step": 9884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006046000000000001,
+      "loss": 1.1205,
+      "step": 9885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060456,
+      "loss": 1.0817,
+      "step": 9886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060452,
+      "loss": 1.062,
+      "step": 9887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060448,
+      "loss": 1.1876,
+      "step": 9888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060444,
+      "loss": 1.0922,
+      "step": 9889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006044000000000001,
+      "loss": 0.9833,
+      "step": 9890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006043600000000001,
+      "loss": 1.1429,
+      "step": 9891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060432,
+      "loss": 1.0201,
+      "step": 9892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060428,
+      "loss": 1.003,
+      "step": 9893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060424,
+      "loss": 1.0697,
+      "step": 9894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006041999999999999,
+      "loss": 1.0554,
+      "step": 9895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006041600000000001,
+      "loss": 1.0485,
+      "step": 9896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060412,
+      "loss": 1.0847,
+      "step": 9897
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060408,
+      "loss": 1.0699,
+      "step": 9898
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060404,
+      "loss": 1.1031,
+      "step": 9899
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000604,
+      "loss": 1.074,
+      "step": 9900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006039600000000001,
+      "loss": 1.026,
+      "step": 9901
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060392,
+      "loss": 1.0542,
+      "step": 9902
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060388,
+      "loss": 1.0826,
+      "step": 9903
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060384,
+      "loss": 1.1182,
+      "step": 9904
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006038,
+      "loss": 1.0839,
+      "step": 9905
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060376,
+      "loss": 1.074,
+      "step": 9906
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006037200000000001,
+      "loss": 1.1306,
+      "step": 9907
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060368,
+      "loss": 1.1313,
+      "step": 9908
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060364,
+      "loss": 1.0689,
+      "step": 9909
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006036,
+      "loss": 1.055,
+      "step": 9910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060356,
+      "loss": 1.0913,
+      "step": 9911
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060352,
+      "loss": 1.0238,
+      "step": 9912
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060348,
+      "loss": 1.102,
+      "step": 9913
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060344,
+      "loss": 1.0423,
+      "step": 9914
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006034,
+      "loss": 1.0242,
+      "step": 9915
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060336,
+      "loss": 0.986,
+      "step": 9916
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060332,
+      "loss": 1.023,
+      "step": 9917
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006032800000000001,
+      "loss": 1.0279,
+      "step": 9918
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060324,
+      "loss": 1.0142,
+      "step": 9919
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006031999999999999,
+      "loss": 1.1145,
+      "step": 9920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060316,
+      "loss": 1.0522,
+      "step": 9921
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060312,
+      "loss": 1.0794,
+      "step": 9922
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060308,
+      "loss": 1.1113,
+      "step": 9923
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060304,
+      "loss": 1.0863,
+      "step": 9924
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000603,
+      "loss": 1.0752,
+      "step": 9925
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060296,
+      "loss": 1.1049,
+      "step": 9926
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006029200000000001,
+      "loss": 1.0524,
+      "step": 9927
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060288,
+      "loss": 1.0525,
+      "step": 9928
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006028400000000001,
+      "loss": 1.0964,
+      "step": 9929
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006028,
+      "loss": 1.0593,
+      "step": 9930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006027599999999999,
+      "loss": 1.1295,
+      "step": 9931
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006027200000000001,
+      "loss": 1.0971,
+      "step": 9932
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060268,
+      "loss": 1.1118,
+      "step": 9933
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060264,
+      "loss": 1.0451,
+      "step": 9934
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006026,
+      "loss": 1.0969,
+      "step": 9935
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060256,
+      "loss": 1.0492,
+      "step": 9936
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006025199999999999,
+      "loss": 1.1714,
+      "step": 9937
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006024800000000001,
+      "loss": 1.1618,
+      "step": 9938
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060244,
+      "loss": 1.1277,
+      "step": 9939
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006024,
+      "loss": 1.0144,
+      "step": 9940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060236,
+      "loss": 1.0679,
+      "step": 9941
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006023199999999999,
+      "loss": 1.1191,
+      "step": 9942
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006022800000000001,
+      "loss": 1.0759,
+      "step": 9943
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060224,
+      "loss": 1.1097,
+      "step": 9944
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006022,
+      "loss": 1.066,
+      "step": 9945
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060216,
+      "loss": 1.0953,
+      "step": 9946
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060212,
+      "loss": 1.0422,
+      "step": 9947
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060208,
+      "loss": 1.0896,
+      "step": 9948
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006020400000000001,
+      "loss": 1.0926,
+      "step": 9949
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000602,
+      "loss": 1.0765,
+      "step": 9950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060196,
+      "loss": 1.0912,
+      "step": 9951
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060192,
+      "loss": 1.0679,
+      "step": 9952
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060188,
+      "loss": 1.1386,
+      "step": 9953
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006018400000000001,
+      "loss": 1.0647,
+      "step": 9954
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006018,
+      "loss": 1.011,
+      "step": 9955
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060176,
+      "loss": 1.1257,
+      "step": 9956
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060172,
+      "loss": 1.0668,
+      "step": 9957
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060168,
+      "loss": 1.1058,
+      "step": 9958
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060164,
+      "loss": 1.0353,
+      "step": 9959
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006016,
+      "loss": 1.1758,
+      "step": 9960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060156,
+      "loss": 1.0817,
+      "step": 9961
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060152,
+      "loss": 1.031,
+      "step": 9962
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060148,
+      "loss": 1.1539,
+      "step": 9963
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060144,
+      "loss": 1.0711,
+      "step": 9964
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006014000000000001,
+      "loss": 1.1151,
+      "step": 9965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060136,
+      "loss": 1.0649,
+      "step": 9966
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060132,
+      "loss": 1.0388,
+      "step": 9967
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060128,
+      "loss": 0.9663,
+      "step": 9968
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060124,
+      "loss": 1.1502,
+      "step": 9969
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006012,
+      "loss": 1.1334,
+      "step": 9970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060116,
+      "loss": 1.0801,
+      "step": 9971
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060112,
+      "loss": 1.1645,
+      "step": 9972
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006010799999999999,
+      "loss": 1.1455,
+      "step": 9973
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006010400000000001,
+      "loss": 1.0236,
+      "step": 9974
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000601,
+      "loss": 1.0625,
+      "step": 9975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006009600000000001,
+      "loss": 1.1575,
+      "step": 9976
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060092,
+      "loss": 1.0226,
+      "step": 9977
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060088,
+      "loss": 1.0679,
+      "step": 9978
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006008400000000001,
+      "loss": 1.0401,
+      "step": 9979
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006008,
+      "loss": 1.0415,
+      "step": 9980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060076,
+      "loss": 1.0409,
+      "step": 9981
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060072,
+      "loss": 1.0866,
+      "step": 9982
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060068,
+      "loss": 1.0059,
+      "step": 9983
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006006399999999999,
+      "loss": 1.1399,
+      "step": 9984
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006006000000000001,
+      "loss": 1.0607,
+      "step": 9985
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060056,
+      "loss": 1.0468,
+      "step": 9986
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006005200000000001,
+      "loss": 1.0641,
+      "step": 9987
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060048,
+      "loss": 1.0804,
+      "step": 9988
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006004399999999999,
+      "loss": 1.1046,
+      "step": 9989
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006004000000000001,
+      "loss": 1.0216,
+      "step": 9990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060036,
+      "loss": 1.1453,
+      "step": 9991
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060032,
+      "loss": 1.0386,
+      "step": 9992
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060028,
+      "loss": 1.0135,
+      "step": 9993
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060024,
+      "loss": 1.0969,
+      "step": 9994
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006002,
+      "loss": 1.1463,
+      "step": 9995
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006001600000000001,
+      "loss": 1.0407,
+      "step": 9996
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060012,
+      "loss": 1.0475,
+      "step": 9997
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006000799999999999,
+      "loss": 1.0969,
+      "step": 9998
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00060004,
+      "loss": 1.1366,
+      "step": 9999
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0006,
+      "loss": 1.0307,
+      "step": 10000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0005999600000000001,
+      "loss": 1.0559,
+      "step": 10001
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00059992,
+      "loss": 1.0929,
+      "step": 10002
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00059988,
+      "loss": 1.1278,
+      "step": 10003
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00059984,
+      "loss": 1.1044,
+      "step": 10004
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0005998,
+      "loss": 0.9711,
+      "step": 10005
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00059976,
+      "loss": 1.1329,
+      "step": 10006
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0005997200000000001,
+      "loss": 1.0002,
+      "step": 10007
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00059968,
+      "loss": 1.0542,
+      "step": 10008
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0005996399999999999,
+      "loss": 1.0848,
+      "step": 10009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005996,
+      "loss": 1.0041,
+      "step": 10010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059956,
+      "loss": 1.0198,
+      "step": 10011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005995200000000001,
+      "loss": 1.0032,
+      "step": 10012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059948,
+      "loss": 1.042,
+      "step": 10013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059944,
+      "loss": 1.0207,
+      "step": 10014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005994,
+      "loss": 1.1194,
+      "step": 10015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005993600000000001,
+      "loss": 1.0756,
+      "step": 10016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059932,
+      "loss": 1.1259,
+      "step": 10017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059928,
+      "loss": 1.0378,
+      "step": 10018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059924,
+      "loss": 1.004,
+      "step": 10019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005991999999999999,
+      "loss": 1.1222,
+      "step": 10020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005991600000000001,
+      "loss": 1.0358,
+      "step": 10021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059912,
+      "loss": 1.0116,
+      "step": 10022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059908,
+      "loss": 1.1154,
+      "step": 10023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059904,
+      "loss": 1.0877,
+      "step": 10024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000599,
+      "loss": 1.1295,
+      "step": 10025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059896,
+      "loss": 1.0644,
+      "step": 10026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005989200000000001,
+      "loss": 1.1017,
+      "step": 10027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059888,
+      "loss": 1.1004,
+      "step": 10028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059884,
+      "loss": 1.038,
+      "step": 10029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005988,
+      "loss": 1.1154,
+      "step": 10030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005987599999999999,
+      "loss": 1.0696,
+      "step": 10031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005987200000000001,
+      "loss": 1.069,
+      "step": 10032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059868,
+      "loss": 1.101,
+      "step": 10033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059864,
+      "loss": 1.0616,
+      "step": 10034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005986,
+      "loss": 1.0512,
+      "step": 10035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059856,
+      "loss": 1.0881,
+      "step": 10036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005985200000000001,
+      "loss": 1.1887,
+      "step": 10037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059848,
+      "loss": 1.0213,
+      "step": 10038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059844,
+      "loss": 1.113,
+      "step": 10039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005984,
+      "loss": 1.0281,
+      "step": 10040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059836,
+      "loss": 1.1005,
+      "step": 10041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059832,
+      "loss": 1.1091,
+      "step": 10042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005982800000000001,
+      "loss": 1.1957,
+      "step": 10043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059824,
+      "loss": 1.0642,
+      "step": 10044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005982,
+      "loss": 1.1331,
+      "step": 10045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059816,
+      "loss": 1.0715,
+      "step": 10046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005981199999999999,
+      "loss": 1.106,
+      "step": 10047
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059808,
+      "loss": 1.0473,
+      "step": 10048
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059804,
+      "loss": 1.0032,
+      "step": 10049
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000598,
+      "loss": 1.05,
+      "step": 10050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059796,
+      "loss": 1.1499,
+      "step": 10051
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059792,
+      "loss": 1.1709,
+      "step": 10052
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059788,
+      "loss": 1.1423,
+      "step": 10053
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005978400000000001,
+      "loss": 1.0756,
+      "step": 10054
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005978,
+      "loss": 1.0328,
+      "step": 10055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059776,
+      "loss": 1.083,
+      "step": 10056
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059772,
+      "loss": 1.044,
+      "step": 10057
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059768,
+      "loss": 1.0518,
+      "step": 10058
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059764,
+      "loss": 1.1678,
+      "step": 10059
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005976,
+      "loss": 1.0463,
+      "step": 10060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059756,
+      "loss": 1.0999,
+      "step": 10061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059752,
+      "loss": 1.0957,
+      "step": 10062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005974800000000001,
+      "loss": 1.0877,
+      "step": 10063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059744,
+      "loss": 1.0942,
+      "step": 10064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005974000000000001,
+      "loss": 1.1206,
+      "step": 10065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059736,
+      "loss": 1.1666,
+      "step": 10066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005973199999999999,
+      "loss": 1.0864,
+      "step": 10067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005972800000000001,
+      "loss": 1.0017,
+      "step": 10068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059724,
+      "loss": 1.1294,
+      "step": 10069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005972,
+      "loss": 1.1256,
+      "step": 10070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059716,
+      "loss": 1.1205,
+      "step": 10071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059712,
+      "loss": 1.1412,
+      "step": 10072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059708,
+      "loss": 1.1084,
+      "step": 10073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005970400000000001,
+      "loss": 1.1199,
+      "step": 10074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000597,
+      "loss": 1.0932,
+      "step": 10075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005969600000000001,
+      "loss": 1.1392,
+      "step": 10076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059692,
+      "loss": 1.0553,
+      "step": 10077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005968799999999999,
+      "loss": 1.0894,
+      "step": 10078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005968400000000001,
+      "loss": 1.171,
+      "step": 10079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005968,
+      "loss": 1.0223,
+      "step": 10080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059676,
+      "loss": 1.1416,
+      "step": 10081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059672,
+      "loss": 1.0204,
+      "step": 10082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059668,
+      "loss": 1.0621,
+      "step": 10083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059664,
+      "loss": 1.0354,
+      "step": 10084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005966000000000001,
+      "loss": 1.0407,
+      "step": 10085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059656,
+      "loss": 1.1038,
+      "step": 10086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059652,
+      "loss": 1.1159,
+      "step": 10087
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059648,
+      "loss": 1.0488,
+      "step": 10088
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059644,
+      "loss": 1.0576,
+      "step": 10089
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005964000000000001,
+      "loss": 1.0672,
+      "step": 10090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059636,
+      "loss": 1.0222,
+      "step": 10091
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059632,
+      "loss": 1.0076,
+      "step": 10092
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059628,
+      "loss": 1.1675,
+      "step": 10093
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059624,
+      "loss": 1.0173,
+      "step": 10094
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005962,
+      "loss": 1.0707,
+      "step": 10095
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059616,
+      "loss": 1.0537,
+      "step": 10096
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059612,
+      "loss": 1.0842,
+      "step": 10097
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059608,
+      "loss": 1.0041,
+      "step": 10098
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059604,
+      "loss": 1.055,
+      "step": 10099
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000596,
+      "loss": 1.0925,
+      "step": 10100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005959600000000001,
+      "loss": 1.0098,
+      "step": 10101
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059592,
+      "loss": 1.0892,
+      "step": 10102
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059588,
+      "loss": 1.0957,
+      "step": 10103
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059584,
+      "loss": 1.051,
+      "step": 10104
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005958000000000001,
+      "loss": 1.0915,
+      "step": 10105
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059576,
+      "loss": 1.0784,
+      "step": 10106
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059572,
+      "loss": 1.0631,
+      "step": 10107
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059568,
+      "loss": 1.1093,
+      "step": 10108
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005956399999999999,
+      "loss": 1.0978,
+      "step": 10109
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005956000000000001,
+      "loss": 1.0515,
+      "step": 10110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059556,
+      "loss": 1.1306,
+      "step": 10111
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005955200000000001,
+      "loss": 1.0773,
+      "step": 10112
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059548,
+      "loss": 1.1077,
+      "step": 10113
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059544,
+      "loss": 1.0974,
+      "step": 10114
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005954,
+      "loss": 1.1798,
+      "step": 10115
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059536,
+      "loss": 1.1019,
+      "step": 10116
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059532,
+      "loss": 1.0702,
+      "step": 10117
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059528,
+      "loss": 1.0443,
+      "step": 10118
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059524,
+      "loss": 1.0022,
+      "step": 10119
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005951999999999999,
+      "loss": 1.0792,
+      "step": 10120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005951600000000001,
+      "loss": 1.0599,
+      "step": 10121
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059512,
+      "loss": 1.0448,
+      "step": 10122
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005950800000000001,
+      "loss": 1.0665,
+      "step": 10123
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059504,
+      "loss": 1.1204,
+      "step": 10124
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005949999999999999,
+      "loss": 1.1165,
+      "step": 10125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005949600000000001,
+      "loss": 1.0861,
+      "step": 10126
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059492,
+      "loss": 1.0603,
+      "step": 10127
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059488,
+      "loss": 1.0256,
+      "step": 10128
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059484,
+      "loss": 1.1376,
+      "step": 10129
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005948,
+      "loss": 1.0265,
+      "step": 10130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059476,
+      "loss": 1.0258,
+      "step": 10131
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005947200000000001,
+      "loss": 1.1236,
+      "step": 10132
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059468,
+      "loss": 1.167,
+      "step": 10133
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059464,
+      "loss": 1.0976,
+      "step": 10134
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005946,
+      "loss": 1.0702,
+      "step": 10135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005945599999999999,
+      "loss": 1.0582,
+      "step": 10136
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005945200000000001,
+      "loss": 1.1315,
+      "step": 10137
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059448,
+      "loss": 1.0922,
+      "step": 10138
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059444,
+      "loss": 1.0767,
+      "step": 10139
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005944,
+      "loss": 1.0109,
+      "step": 10140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059436,
+      "loss": 1.0073,
+      "step": 10141
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059432,
+      "loss": 1.1709,
+      "step": 10142
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005942800000000001,
+      "loss": 1.0171,
+      "step": 10143
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059424,
+      "loss": 1.0905,
+      "step": 10144
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005941999999999999,
+      "loss": 1.0848,
+      "step": 10145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059416,
+      "loss": 1.0375,
+      "step": 10146
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059412,
+      "loss": 1.1358,
+      "step": 10147
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005940800000000001,
+      "loss": 1.1281,
+      "step": 10148
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059404,
+      "loss": 1.0838,
+      "step": 10149
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000594,
+      "loss": 1.1527,
+      "step": 10150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059396,
+      "loss": 1.0237,
+      "step": 10151
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005939200000000001,
+      "loss": 1.0272,
+      "step": 10152
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059388,
+      "loss": 1.103,
+      "step": 10153
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059384,
+      "loss": 1.0848,
+      "step": 10154
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005938,
+      "loss": 1.0335,
+      "step": 10155
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005937599999999999,
+      "loss": 1.1025,
+      "step": 10156
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005937200000000001,
+      "loss": 1.0853,
+      "step": 10157
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059368,
+      "loss": 1.0443,
+      "step": 10158
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059364,
+      "loss": 1.0695,
+      "step": 10159
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005936,
+      "loss": 1.1453,
+      "step": 10160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059356,
+      "loss": 1.1919,
+      "step": 10161
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059352,
+      "loss": 1.1195,
+      "step": 10162
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005934800000000001,
+      "loss": 1.0409,
+      "step": 10163
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059344,
+      "loss": 1.1175,
+      "step": 10164
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005934,
+      "loss": 1.0586,
+      "step": 10165
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059336,
+      "loss": 1.1363,
+      "step": 10166
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005933199999999999,
+      "loss": 1.1093,
+      "step": 10167
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005932800000000001,
+      "loss": 1.0434,
+      "step": 10168
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059324,
+      "loss": 1.0811,
+      "step": 10169
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005932,
+      "loss": 1.0516,
+      "step": 10170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059316,
+      "loss": 0.9422,
+      "step": 10171
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059312,
+      "loss": 1.0838,
+      "step": 10172
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005930800000000001,
+      "loss": 1.0448,
+      "step": 10173
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059304,
+      "loss": 1.1235,
+      "step": 10174
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000593,
+      "loss": 1.104,
+      "step": 10175
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059296,
+      "loss": 1.079,
+      "step": 10176
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059292,
+      "loss": 1.0843,
+      "step": 10177
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059288,
+      "loss": 1.0414,
+      "step": 10178
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005928400000000001,
+      "loss": 1.0598,
+      "step": 10179
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005928,
+      "loss": 1.1159,
+      "step": 10180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059276,
+      "loss": 1.034,
+      "step": 10181
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059272,
+      "loss": 1.0633,
+      "step": 10182
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005926799999999999,
+      "loss": 1.0833,
+      "step": 10183
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059264,
+      "loss": 1.0861,
+      "step": 10184
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005926,
+      "loss": 1.1574,
+      "step": 10185
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059256,
+      "loss": 1.1194,
+      "step": 10186
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059252,
+      "loss": 1.1128,
+      "step": 10187
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059248,
+      "loss": 1.0301,
+      "step": 10188
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059244,
+      "loss": 1.0493,
+      "step": 10189
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005924000000000001,
+      "loss": 1.0451,
+      "step": 10190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059236,
+      "loss": 0.9777,
+      "step": 10191
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059232,
+      "loss": 1.044,
+      "step": 10192
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059228,
+      "loss": 1.0377,
+      "step": 10193
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059224,
+      "loss": 1.1316,
+      "step": 10194
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005922,
+      "loss": 1.1376,
+      "step": 10195
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059216,
+      "loss": 1.1136,
+      "step": 10196
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059212,
+      "loss": 1.1737,
+      "step": 10197
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059208,
+      "loss": 1.1593,
+      "step": 10198
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005920400000000001,
+      "loss": 1.1043,
+      "step": 10199
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000592,
+      "loss": 1.0614,
+      "step": 10200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005919600000000001,
+      "loss": 1.1428,
+      "step": 10201
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059192,
+      "loss": 1.133,
+      "step": 10202
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005918799999999999,
+      "loss": 1.0611,
+      "step": 10203
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059184,
+      "loss": 1.0654,
+      "step": 10204
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005918,
+      "loss": 0.9855,
+      "step": 10205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059176,
+      "loss": 1.0117,
+      "step": 10206
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059172,
+      "loss": 1.0685,
+      "step": 10207
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059168,
+      "loss": 0.98,
+      "step": 10208
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059164,
+      "loss": 1.1413,
+      "step": 10209
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005916000000000001,
+      "loss": 1.1334,
+      "step": 10210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059156,
+      "loss": 1.1131,
+      "step": 10211
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005915200000000001,
+      "loss": 1.1018,
+      "step": 10212
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059148,
+      "loss": 1.0261,
+      "step": 10213
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005914399999999999,
+      "loss": 1.1096,
+      "step": 10214
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005914000000000001,
+      "loss": 1.1413,
+      "step": 10215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059136,
+      "loss": 1.0784,
+      "step": 10216
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059132,
+      "loss": 1.1343,
+      "step": 10217
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059128,
+      "loss": 1.125,
+      "step": 10218
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059124,
+      "loss": 1.0304,
+      "step": 10219
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005912,
+      "loss": 1.1216,
+      "step": 10220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005911600000000001,
+      "loss": 1.0436,
+      "step": 10221
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059112,
+      "loss": 1.1156,
+      "step": 10222
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059108,
+      "loss": 1.0024,
+      "step": 10223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059104,
+      "loss": 1.0319,
+      "step": 10224
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005909999999999999,
+      "loss": 0.9904,
+      "step": 10225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005909600000000001,
+      "loss": 1.0584,
+      "step": 10226
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059092,
+      "loss": 1.0409,
+      "step": 10227
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059088,
+      "loss": 1.0448,
+      "step": 10228
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059084,
+      "loss": 1.0825,
+      "step": 10229
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005908,
+      "loss": 1.1304,
+      "step": 10230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059076,
+      "loss": 1.1551,
+      "step": 10231
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059072,
+      "loss": 1.0419,
+      "step": 10232
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059068,
+      "loss": 0.9883,
+      "step": 10233
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059064,
+      "loss": 1.0784,
+      "step": 10234
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005906,
+      "loss": 1.0498,
+      "step": 10235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059056,
+      "loss": 1.0987,
+      "step": 10236
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005905200000000001,
+      "loss": 1.1031,
+      "step": 10237
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059048,
+      "loss": 1.0564,
+      "step": 10238
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059044,
+      "loss": 1.0495,
+      "step": 10239
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005904,
+      "loss": 1.1008,
+      "step": 10240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005903600000000001,
+      "loss": 1.1167,
+      "step": 10241
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059032,
+      "loss": 1.0459,
+      "step": 10242
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059028,
+      "loss": 1.0524,
+      "step": 10243
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059024,
+      "loss": 1.0225,
+      "step": 10244
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005901999999999999,
+      "loss": 1.1236,
+      "step": 10245
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005901600000000001,
+      "loss": 1.101,
+      "step": 10246
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059012,
+      "loss": 1.0766,
+      "step": 10247
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005900800000000001,
+      "loss": 1.0315,
+      "step": 10248
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059004,
+      "loss": 1.0738,
+      "step": 10249
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00059,
+      "loss": 1.0585,
+      "step": 10250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00058996,
+      "loss": 1.0722,
+      "step": 10251
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00058992,
+      "loss": 1.0371,
+      "step": 10252
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00058988,
+      "loss": 1.0813,
+      "step": 10253
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00058984,
+      "loss": 1.0731,
+      "step": 10254
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005898,
+      "loss": 0.9752,
+      "step": 10255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005897599999999999,
+      "loss": 1.1304,
+      "step": 10256
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005897200000000001,
+      "loss": 1.1029,
+      "step": 10257
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00058968,
+      "loss": 1.1009,
+      "step": 10258
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005896400000000001,
+      "loss": 1.1026,
+      "step": 10259
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005896,
+      "loss": 1.062,
+      "step": 10260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005895599999999999,
+      "loss": 1.1091,
+      "step": 10261
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0005895200000000001,
+      "loss": 1.1246,
+      "step": 10262
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058948,
+      "loss": 1.1459,
+      "step": 10263
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058944,
+      "loss": 1.1143,
+      "step": 10264
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005894,
+      "loss": 1.1201,
+      "step": 10265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058936,
+      "loss": 0.9867,
+      "step": 10266
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058932,
+      "loss": 1.0755,
+      "step": 10267
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005892800000000001,
+      "loss": 1.1436,
+      "step": 10268
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058924,
+      "loss": 1.1164,
+      "step": 10269
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005892,
+      "loss": 1.1112,
+      "step": 10270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058916,
+      "loss": 1.0142,
+      "step": 10271
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005891199999999999,
+      "loss": 1.1158,
+      "step": 10272
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005890800000000001,
+      "loss": 1.0607,
+      "step": 10273
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058904,
+      "loss": 1.1024,
+      "step": 10274
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000589,
+      "loss": 1.0644,
+      "step": 10275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058896,
+      "loss": 1.1665,
+      "step": 10276
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058892,
+      "loss": 1.015,
+      "step": 10277
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058888,
+      "loss": 1.1522,
+      "step": 10278
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005888400000000001,
+      "loss": 1.1401,
+      "step": 10279
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005888,
+      "loss": 1.0325,
+      "step": 10280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005887599999999999,
+      "loss": 1.0117,
+      "step": 10281
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058872,
+      "loss": 1.0904,
+      "step": 10282
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058868,
+      "loss": 1.0524,
+      "step": 10283
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005886400000000001,
+      "loss": 1.0097,
+      "step": 10284
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005886,
+      "loss": 1.0658,
+      "step": 10285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058856,
+      "loss": 1.0988,
+      "step": 10286
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058852,
+      "loss": 1.08,
+      "step": 10287
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005884800000000001,
+      "loss": 1.0937,
+      "step": 10288
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058844,
+      "loss": 1.0517,
+      "step": 10289
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005884000000000001,
+      "loss": 1.0332,
+      "step": 10290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058836,
+      "loss": 1.1085,
+      "step": 10291
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005883199999999999,
+      "loss": 1.1415,
+      "step": 10292
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058828,
+      "loss": 1.1129,
+      "step": 10293
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058824,
+      "loss": 1.049,
+      "step": 10294
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005882,
+      "loss": 1.1503,
+      "step": 10295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058816,
+      "loss": 1.1072,
+      "step": 10296
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058812,
+      "loss": 1.0926,
+      "step": 10297
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058808,
+      "loss": 1.0917,
+      "step": 10298
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005880400000000001,
+      "loss": 1.1185,
+      "step": 10299
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000588,
+      "loss": 1.0651,
+      "step": 10300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058796,
+      "loss": 1.0812,
+      "step": 10301
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058792,
+      "loss": 1.1171,
+      "step": 10302
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005878799999999999,
+      "loss": 1.1915,
+      "step": 10303
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005878400000000001,
+      "loss": 1.093,
+      "step": 10304
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005878,
+      "loss": 1.0999,
+      "step": 10305
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058776,
+      "loss": 1.109,
+      "step": 10306
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058772,
+      "loss": 1.0531,
+      "step": 10307
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058768,
+      "loss": 1.1084,
+      "step": 10308
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005876400000000001,
+      "loss": 1.0869,
+      "step": 10309
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005876,
+      "loss": 1.0466,
+      "step": 10310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058756,
+      "loss": 1.0973,
+      "step": 10311
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058752,
+      "loss": 1.0311,
+      "step": 10312
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058748,
+      "loss": 0.9717,
+      "step": 10313
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005874399999999999,
+      "loss": 1.2021,
+      "step": 10314
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005874000000000001,
+      "loss": 1.1239,
+      "step": 10315
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058736,
+      "loss": 1.1808,
+      "step": 10316
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058732,
+      "loss": 1.0431,
+      "step": 10317
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058728,
+      "loss": 1.1004,
+      "step": 10318
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058724,
+      "loss": 1.0172,
+      "step": 10319
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005872000000000001,
+      "loss": 1.1098,
+      "step": 10320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058716,
+      "loss": 1.0877,
+      "step": 10321
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058712,
+      "loss": 1.1015,
+      "step": 10322
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058708,
+      "loss": 1.0799,
+      "step": 10323
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058704,
+      "loss": 1.0487,
+      "step": 10324
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000587,
+      "loss": 1.0982,
+      "step": 10325
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005869600000000001,
+      "loss": 1.052,
+      "step": 10326
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058692,
+      "loss": 1.0897,
+      "step": 10327
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058688,
+      "loss": 1.0909,
+      "step": 10328
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058684,
+      "loss": 1.0045,
+      "step": 10329
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005868,
+      "loss": 1.045,
+      "step": 10330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058676,
+      "loss": 1.0637,
+      "step": 10331
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058672,
+      "loss": 1.1096,
+      "step": 10332
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058668,
+      "loss": 1.0278,
+      "step": 10333
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058664,
+      "loss": 1.0711,
+      "step": 10334
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005866000000000001,
+      "loss": 1.0313,
+      "step": 10335
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058656,
+      "loss": 1.1545,
+      "step": 10336
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005865200000000001,
+      "loss": 1.1349,
+      "step": 10337
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058648,
+      "loss": 1.1364,
+      "step": 10338
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005864399999999999,
+      "loss": 1.1145,
+      "step": 10339
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005864,
+      "loss": 1.06,
+      "step": 10340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058636,
+      "loss": 1.1105,
+      "step": 10341
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058632,
+      "loss": 1.0682,
+      "step": 10342
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058628,
+      "loss": 1.0805,
+      "step": 10343
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058624,
+      "loss": 1.0405,
+      "step": 10344
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005862,
+      "loss": 1.0438,
+      "step": 10345
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005861600000000001,
+      "loss": 1.0018,
+      "step": 10346
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058612,
+      "loss": 1.1556,
+      "step": 10347
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005860800000000001,
+      "loss": 1.0665,
+      "step": 10348
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058604,
+      "loss": 1.1558,
+      "step": 10349
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005859999999999999,
+      "loss": 1.0603,
+      "step": 10350
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005859600000000001,
+      "loss": 1.0597,
+      "step": 10351
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058592,
+      "loss": 1.1483,
+      "step": 10352
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058588,
+      "loss": 1.0637,
+      "step": 10353
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058584,
+      "loss": 1.0424,
+      "step": 10354
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005858,
+      "loss": 1.1226,
+      "step": 10355
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058576,
+      "loss": 1.0545,
+      "step": 10356
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005857200000000001,
+      "loss": 0.9875,
+      "step": 10357
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058568,
+      "loss": 1.0059,
+      "step": 10358
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058564,
+      "loss": 1.1147,
+      "step": 10359
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005856,
+      "loss": 0.9865,
+      "step": 10360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005855599999999999,
+      "loss": 1.1457,
+      "step": 10361
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005855200000000001,
+      "loss": 1.0995,
+      "step": 10362
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058548,
+      "loss": 1.0914,
+      "step": 10363
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058544,
+      "loss": 1.0808,
+      "step": 10364
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005854,
+      "loss": 1.0411,
+      "step": 10365
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058536,
+      "loss": 1.0557,
+      "step": 10366
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058532,
+      "loss": 1.1651,
+      "step": 10367
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005852800000000001,
+      "loss": 1.0184,
+      "step": 10368
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058524,
+      "loss": 1.0782,
+      "step": 10369
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005852,
+      "loss": 1.1185,
+      "step": 10370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058516,
+      "loss": 0.971,
+      "step": 10371
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058512,
+      "loss": 1.0549,
+      "step": 10372
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005850800000000001,
+      "loss": 1.072,
+      "step": 10373
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058504,
+      "loss": 1.0066,
+      "step": 10374
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000585,
+      "loss": 1.1808,
+      "step": 10375
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058496,
+      "loss": 1.0016,
+      "step": 10376
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005849200000000001,
+      "loss": 1.0399,
+      "step": 10377
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058488,
+      "loss": 1.0773,
+      "step": 10378
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058484,
+      "loss": 1.0475,
+      "step": 10379
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005848,
+      "loss": 1.1424,
+      "step": 10380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005847599999999999,
+      "loss": 1.0412,
+      "step": 10381
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058472,
+      "loss": 1.1295,
+      "step": 10382
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058468,
+      "loss": 1.125,
+      "step": 10383
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005846400000000001,
+      "loss": 1.1385,
+      "step": 10384
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005846,
+      "loss": 1.0507,
+      "step": 10385
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058456,
+      "loss": 0.9632,
+      "step": 10386
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058452,
+      "loss": 1.078,
+      "step": 10387
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058448,
+      "loss": 1.0901,
+      "step": 10388
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058444,
+      "loss": 1.1343,
+      "step": 10389
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005844,
+      "loss": 1.0926,
+      "step": 10390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058436,
+      "loss": 1.0656,
+      "step": 10391
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005843199999999999,
+      "loss": 1.07,
+      "step": 10392
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005842800000000001,
+      "loss": 1.0763,
+      "step": 10393
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058424,
+      "loss": 1.0647,
+      "step": 10394
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005842000000000001,
+      "loss": 1.1282,
+      "step": 10395
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058416,
+      "loss": 1.1275,
+      "step": 10396
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058412,
+      "loss": 1.1431,
+      "step": 10397
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005840800000000001,
+      "loss": 1.0468,
+      "step": 10398
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058404,
+      "loss": 1.0336,
+      "step": 10399
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000584,
+      "loss": 1.077,
+      "step": 10400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058396,
+      "loss": 1.0419,
+      "step": 10401
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058392,
+      "loss": 1.0986,
+      "step": 10402
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058388,
+      "loss": 1.0377,
+      "step": 10403
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005838400000000001,
+      "loss": 1.0524,
+      "step": 10404
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005838,
+      "loss": 1.0634,
+      "step": 10405
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058376,
+      "loss": 1.0426,
+      "step": 10406
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058372,
+      "loss": 1.081,
+      "step": 10407
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005836799999999999,
+      "loss": 1.0475,
+      "step": 10408
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005836400000000001,
+      "loss": 1.024,
+      "step": 10409
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005836,
+      "loss": 1.1266,
+      "step": 10410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058356,
+      "loss": 1.0839,
+      "step": 10411
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058352,
+      "loss": 1.0331,
+      "step": 10412
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058348,
+      "loss": 1.1459,
+      "step": 10413
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058344,
+      "loss": 1.0619,
+      "step": 10414
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005834000000000001,
+      "loss": 1.0896,
+      "step": 10415
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058336,
+      "loss": 1.0449,
+      "step": 10416
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005833199999999999,
+      "loss": 1.0953,
+      "step": 10417
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058328,
+      "loss": 1.0433,
+      "step": 10418
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058324,
+      "loss": 1.0954,
+      "step": 10419
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005832000000000001,
+      "loss": 1.0254,
+      "step": 10420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058316,
+      "loss": 1.0548,
+      "step": 10421
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058312,
+      "loss": 1.0148,
+      "step": 10422
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058308,
+      "loss": 1.0879,
+      "step": 10423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005830400000000001,
+      "loss": 1.103,
+      "step": 10424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000583,
+      "loss": 1.0749,
+      "step": 10425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005829600000000001,
+      "loss": 1.0607,
+      "step": 10426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058292,
+      "loss": 1.087,
+      "step": 10427
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005828799999999999,
+      "loss": 1.0103,
+      "step": 10428
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058284,
+      "loss": 1.0691,
+      "step": 10429
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005828,
+      "loss": 1.2089,
+      "step": 10430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058276,
+      "loss": 1.0823,
+      "step": 10431
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058272,
+      "loss": 1.1462,
+      "step": 10432
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058268,
+      "loss": 1.1133,
+      "step": 10433
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058264,
+      "loss": 1.1169,
+      "step": 10434
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005826000000000001,
+      "loss": 1.137,
+      "step": 10435
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058256,
+      "loss": 1.0427,
+      "step": 10436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058252,
+      "loss": 1.1207,
+      "step": 10437
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058248,
+      "loss": 1.0944,
+      "step": 10438
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005824399999999999,
+      "loss": 1.0753,
+      "step": 10439
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005824000000000001,
+      "loss": 1.0895,
+      "step": 10440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058236,
+      "loss": 1.063,
+      "step": 10441
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058232,
+      "loss": 1.064,
+      "step": 10442
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058228,
+      "loss": 1.0176,
+      "step": 10443
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058224,
+      "loss": 1.0569,
+      "step": 10444
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005822000000000001,
+      "loss": 1.0342,
+      "step": 10445
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058216,
+      "loss": 1.0625,
+      "step": 10446
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058212,
+      "loss": 1.1029,
+      "step": 10447
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058208,
+      "loss": 1.1149,
+      "step": 10448
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058204,
+      "loss": 1.1675,
+      "step": 10449
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005819999999999999,
+      "loss": 1.1406,
+      "step": 10450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005819600000000001,
+      "loss": 1.0461,
+      "step": 10451
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058192,
+      "loss": 1.0183,
+      "step": 10452
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058188,
+      "loss": 1.0477,
+      "step": 10453
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058184,
+      "loss": 1.0696,
+      "step": 10454
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005818,
+      "loss": 1.0377,
+      "step": 10455
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005817600000000001,
+      "loss": 1.0974,
+      "step": 10456
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058172,
+      "loss": 1.0952,
+      "step": 10457
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058168,
+      "loss": 1.1283,
+      "step": 10458
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058164,
+      "loss": 1.0789,
+      "step": 10459
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005816,
+      "loss": 1.0655,
+      "step": 10460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058156,
+      "loss": 1.1274,
+      "step": 10461
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005815200000000001,
+      "loss": 1.0423,
+      "step": 10462
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058148,
+      "loss": 1.123,
+      "step": 10463
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058144,
+      "loss": 1.0744,
+      "step": 10464
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005814,
+      "loss": 1.1049,
+      "step": 10465
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058136,
+      "loss": 1.0548,
+      "step": 10466
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058132,
+      "loss": 1.1406,
+      "step": 10467
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058128,
+      "loss": 1.084,
+      "step": 10468
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058124,
+      "loss": 1.0887,
+      "step": 10469
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005812,
+      "loss": 1.0396,
+      "step": 10470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058116,
+      "loss": 1.0956,
+      "step": 10471
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058112,
+      "loss": 1.0891,
+      "step": 10472
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005810800000000001,
+      "loss": 0.9775,
+      "step": 10473
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058104,
+      "loss": 1.1152,
+      "step": 10474
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005809999999999999,
+      "loss": 1.0657,
+      "step": 10475
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058096,
+      "loss": 1.0921,
+      "step": 10476
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058092,
+      "loss": 1.0721,
+      "step": 10477
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058088,
+      "loss": 1.0617,
+      "step": 10478
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058084,
+      "loss": 1.0841,
+      "step": 10479
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005808,
+      "loss": 1.1009,
+      "step": 10480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058076,
+      "loss": 1.0549,
+      "step": 10481
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005807200000000001,
+      "loss": 1.0637,
+      "step": 10482
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058068,
+      "loss": 1.0969,
+      "step": 10483
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005806400000000001,
+      "loss": 1.0031,
+      "step": 10484
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005806,
+      "loss": 1.0379,
+      "step": 10485
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005805599999999999,
+      "loss": 1.1116,
+      "step": 10486
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005805200000000001,
+      "loss": 1.0861,
+      "step": 10487
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058048,
+      "loss": 1.0966,
+      "step": 10488
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058044,
+      "loss": 0.9868,
+      "step": 10489
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005804,
+      "loss": 1.0723,
+      "step": 10490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058036,
+      "loss": 1.0416,
+      "step": 10491
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058032,
+      "loss": 1.1296,
+      "step": 10492
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005802800000000001,
+      "loss": 1.1297,
+      "step": 10493
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058024,
+      "loss": 1.1407,
+      "step": 10494
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005802,
+      "loss": 1.0987,
+      "step": 10495
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058016,
+      "loss": 1.0591,
+      "step": 10496
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005801199999999999,
+      "loss": 1.0893,
+      "step": 10497
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005800800000000001,
+      "loss": 1.24,
+      "step": 10498
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058004,
+      "loss": 1.094,
+      "step": 10499
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00058,
+      "loss": 1.0651,
+      "step": 10500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057996,
+      "loss": 1.1161,
+      "step": 10501
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057992,
+      "loss": 1.0756,
+      "step": 10502
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057988,
+      "loss": 1.0685,
+      "step": 10503
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005798400000000001,
+      "loss": 1.0803,
+      "step": 10504
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005798,
+      "loss": 1.1271,
+      "step": 10505
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057976,
+      "loss": 1.1581,
+      "step": 10506
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057972,
+      "loss": 1.0079,
+      "step": 10507
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057968,
+      "loss": 1.0752,
+      "step": 10508
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005796400000000001,
+      "loss": 1.0171,
+      "step": 10509
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005796,
+      "loss": 1.1058,
+      "step": 10510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057956,
+      "loss": 1.108,
+      "step": 10511
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057952,
+      "loss": 1.101,
+      "step": 10512
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005794800000000001,
+      "loss": 1.0533,
+      "step": 10513
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057944,
+      "loss": 1.0572,
+      "step": 10514
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0005794,
+      "loss": 1.0947,
+      "step": 10515
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00057936,
+      "loss": 1.0786,
+      "step": 10516
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005793199999999999,
+      "loss": 1.1183,
+      "step": 10517
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057928,
+      "loss": 1.0997,
+      "step": 10518
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057924,
+      "loss": 1.0891,
+      "step": 10519
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005792000000000001,
+      "loss": 1.0093,
+      "step": 10520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057916,
+      "loss": 1.0736,
+      "step": 10521
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057912,
+      "loss": 1.0211,
+      "step": 10522
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057908,
+      "loss": 1.142,
+      "step": 10523
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057904,
+      "loss": 1.1133,
+      "step": 10524
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000579,
+      "loss": 1.0391,
+      "step": 10525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057896,
+      "loss": 1.0533,
+      "step": 10526
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057892,
+      "loss": 1.0653,
+      "step": 10527
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005788799999999999,
+      "loss": 1.0127,
+      "step": 10528
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005788400000000001,
+      "loss": 1.0905,
+      "step": 10529
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005788,
+      "loss": 1.0519,
+      "step": 10530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005787600000000001,
+      "loss": 0.9899,
+      "step": 10531
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057872,
+      "loss": 1.1222,
+      "step": 10532
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057868,
+      "loss": 1.1211,
+      "step": 10533
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005786400000000001,
+      "loss": 1.0271,
+      "step": 10534
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005786,
+      "loss": 1.1102,
+      "step": 10535
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057856,
+      "loss": 1.0375,
+      "step": 10536
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057852,
+      "loss": 1.0489,
+      "step": 10537
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057848,
+      "loss": 1.0696,
+      "step": 10538
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005784399999999999,
+      "loss": 1.0442,
+      "step": 10539
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005784000000000001,
+      "loss": 1.051,
+      "step": 10540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057836,
+      "loss": 1.1016,
+      "step": 10541
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057832,
+      "loss": 1.0969,
+      "step": 10542
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057828,
+      "loss": 1.0727,
+      "step": 10543
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005782399999999999,
+      "loss": 1.0443,
+      "step": 10544
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005782000000000001,
+      "loss": 1.1032,
+      "step": 10545
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057816,
+      "loss": 1.1867,
+      "step": 10546
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057812,
+      "loss": 1.0354,
+      "step": 10547
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057808,
+      "loss": 1.0959,
+      "step": 10548
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057804,
+      "loss": 1.1154,
+      "step": 10549
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000578,
+      "loss": 1.0639,
+      "step": 10550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005779600000000001,
+      "loss": 1.0859,
+      "step": 10551
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057792,
+      "loss": 1.1585,
+      "step": 10552
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005778799999999999,
+      "loss": 1.003,
+      "step": 10553
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057784,
+      "loss": 1.0983,
+      "step": 10554
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005778,
+      "loss": 1.0658,
+      "step": 10555
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005777600000000001,
+      "loss": 1.0993,
+      "step": 10556
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057772,
+      "loss": 1.0573,
+      "step": 10557
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057768,
+      "loss": 1.1123,
+      "step": 10558
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057764,
+      "loss": 1.0102,
+      "step": 10559
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005776,
+      "loss": 1.0727,
+      "step": 10560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057756,
+      "loss": 1.1715,
+      "step": 10561
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005775200000000001,
+      "loss": 1.1185,
+      "step": 10562
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057748,
+      "loss": 1.0741,
+      "step": 10563
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005774399999999999,
+      "loss": 1.0413,
+      "step": 10564
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005774,
+      "loss": 1.0285,
+      "step": 10565
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057736,
+      "loss": 1.1579,
+      "step": 10566
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005773200000000001,
+      "loss": 1.071,
+      "step": 10567
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057728,
+      "loss": 1.0836,
+      "step": 10568
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057724,
+      "loss": 1.0212,
+      "step": 10569
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005772,
+      "loss": 1.0783,
+      "step": 10570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005771600000000001,
+      "loss": 1.0495,
+      "step": 10571
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057712,
+      "loss": 1.0595,
+      "step": 10572
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057708,
+      "loss": 1.0896,
+      "step": 10573
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057704,
+      "loss": 1.032,
+      "step": 10574
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005769999999999999,
+      "loss": 1.1558,
+      "step": 10575
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005769600000000001,
+      "loss": 1.0792,
+      "step": 10576
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057692,
+      "loss": 1.0212,
+      "step": 10577
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057688,
+      "loss": 1.0858,
+      "step": 10578
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057684,
+      "loss": 1.0348,
+      "step": 10579
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005768,
+      "loss": 1.1188,
+      "step": 10580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005767600000000001,
+      "loss": 0.9746,
+      "step": 10581
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005767200000000001,
+      "loss": 1.1346,
+      "step": 10582
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057668,
+      "loss": 1.0544,
+      "step": 10583
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057664,
+      "loss": 1.1088,
+      "step": 10584
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005766,
+      "loss": 1.0045,
+      "step": 10585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005765599999999999,
+      "loss": 1.1019,
+      "step": 10586
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005765200000000001,
+      "loss": 1.1412,
+      "step": 10587
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057648,
+      "loss": 1.0588,
+      "step": 10588
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057644,
+      "loss": 1.1634,
+      "step": 10589
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005764,
+      "loss": 1.0676,
+      "step": 10590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057636,
+      "loss": 1.0195,
+      "step": 10591
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005763200000000001,
+      "loss": 0.9999,
+      "step": 10592
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057628,
+      "loss": 1.0558,
+      "step": 10593
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057624,
+      "loss": 1.1066,
+      "step": 10594
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005762,
+      "loss": 1.0098,
+      "step": 10595
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057616,
+      "loss": 1.0813,
+      "step": 10596
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057612,
+      "loss": 1.0456,
+      "step": 10597
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005760800000000001,
+      "loss": 1.077,
+      "step": 10598
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057604,
+      "loss": 0.9595,
+      "step": 10599
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000576,
+      "loss": 1.1191,
+      "step": 10600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057596,
+      "loss": 1.0303,
+      "step": 10601
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057592,
+      "loss": 1.1156,
+      "step": 10602
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057588,
+      "loss": 1.0376,
+      "step": 10603
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057584,
+      "loss": 1.1214,
+      "step": 10604
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005758,
+      "loss": 1.0859,
+      "step": 10605
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057576,
+      "loss": 1.0012,
+      "step": 10606
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057572,
+      "loss": 1.0879,
+      "step": 10607
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057568,
+      "loss": 1.1038,
+      "step": 10608
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005756400000000001,
+      "loss": 1.2085,
+      "step": 10609
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005756,
+      "loss": 0.9898,
+      "step": 10610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057556,
+      "loss": 1.138,
+      "step": 10611
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057552,
+      "loss": 1.0934,
+      "step": 10612
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057548,
+      "loss": 1.1004,
+      "step": 10613
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057544,
+      "loss": 1.0082,
+      "step": 10614
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005754,
+      "loss": 1.1746,
+      "step": 10615
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057536,
+      "loss": 1.0452,
+      "step": 10616
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057532,
+      "loss": 1.1222,
+      "step": 10617
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005752800000000001,
+      "loss": 1.0873,
+      "step": 10618
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057524,
+      "loss": 1.0947,
+      "step": 10619
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005752000000000001,
+      "loss": 1.1657,
+      "step": 10620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057516,
+      "loss": 1.0696,
+      "step": 10621
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005751199999999999,
+      "loss": 1.107,
+      "step": 10622
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005750800000000001,
+      "loss": 1.0072,
+      "step": 10623
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057504,
+      "loss": 1.0064,
+      "step": 10624
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000575,
+      "loss": 1.1165,
+      "step": 10625
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057496,
+      "loss": 1.0688,
+      "step": 10626
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057492,
+      "loss": 1.0688,
+      "step": 10627
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005748799999999999,
+      "loss": 1.0839,
+      "step": 10628
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005748400000000001,
+      "loss": 1.1366,
+      "step": 10629
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005748,
+      "loss": 1.0112,
+      "step": 10630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005747600000000001,
+      "loss": 1.0577,
+      "step": 10631
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057472,
+      "loss": 1.1106,
+      "step": 10632
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005746799999999999,
+      "loss": 1.0841,
+      "step": 10633
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005746400000000001,
+      "loss": 1.1175,
+      "step": 10634
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005746,
+      "loss": 1.1384,
+      "step": 10635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057456,
+      "loss": 1.0867,
+      "step": 10636
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057452,
+      "loss": 0.9953,
+      "step": 10637
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057448,
+      "loss": 1.0789,
+      "step": 10638
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057444,
+      "loss": 1.0302,
+      "step": 10639
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005744000000000001,
+      "loss": 1.0975,
+      "step": 10640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057436,
+      "loss": 1.1158,
+      "step": 10641
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057432,
+      "loss": 1.1083,
+      "step": 10642
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057428,
+      "loss": 1.099,
+      "step": 10643
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057424,
+      "loss": 1.1141,
+      "step": 10644
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005742000000000001,
+      "loss": 1.0615,
+      "step": 10645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057416,
+      "loss": 1.1572,
+      "step": 10646
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057412,
+      "loss": 1.0305,
+      "step": 10647
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057408,
+      "loss": 1.121,
+      "step": 10648
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057404,
+      "loss": 1.0347,
+      "step": 10649
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000574,
+      "loss": 1.0619,
+      "step": 10650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057396,
+      "loss": 1.1155,
+      "step": 10651
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057392,
+      "loss": 1.1008,
+      "step": 10652
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005738799999999999,
+      "loss": 0.9985,
+      "step": 10653
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057384,
+      "loss": 0.9743,
+      "step": 10654
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005738,
+      "loss": 1.0705,
+      "step": 10655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005737600000000001,
+      "loss": 1.0855,
+      "step": 10656
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057372,
+      "loss": 1.0113,
+      "step": 10657
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057368,
+      "loss": 1.0882,
+      "step": 10658
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057364,
+      "loss": 1.0188,
+      "step": 10659
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005736000000000001,
+      "loss": 1.1819,
+      "step": 10660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057356,
+      "loss": 1.0672,
+      "step": 10661
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057352,
+      "loss": 1.0096,
+      "step": 10662
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057348,
+      "loss": 1.1134,
+      "step": 10663
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005734399999999999,
+      "loss": 1.0674,
+      "step": 10664
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005734000000000001,
+      "loss": 0.9934,
+      "step": 10665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057336,
+      "loss": 1.0671,
+      "step": 10666
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005733200000000001,
+      "loss": 1.0405,
+      "step": 10667
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057328,
+      "loss": 1.1198,
+      "step": 10668
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057324,
+      "loss": 1.1117,
+      "step": 10669
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005732000000000001,
+      "loss": 1.1186,
+      "step": 10670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057316,
+      "loss": 1.1176,
+      "step": 10671
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057312,
+      "loss": 1.0507,
+      "step": 10672
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057308,
+      "loss": 0.9401,
+      "step": 10673
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057304,
+      "loss": 1.1098,
+      "step": 10674
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005729999999999999,
+      "loss": 0.9357,
+      "step": 10675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005729600000000001,
+      "loss": 1.0873,
+      "step": 10676
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057292,
+      "loss": 1.1483,
+      "step": 10677
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057288,
+      "loss": 1.1279,
+      "step": 10678
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057284,
+      "loss": 1.0978,
+      "step": 10679
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005727999999999999,
+      "loss": 1.0808,
+      "step": 10680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005727600000000001,
+      "loss": 1.0533,
+      "step": 10681
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057272,
+      "loss": 1.0079,
+      "step": 10682
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057268,
+      "loss": 1.1527,
+      "step": 10683
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057264,
+      "loss": 1.0818,
+      "step": 10684
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005726,
+      "loss": 1.1041,
+      "step": 10685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057256,
+      "loss": 1.0353,
+      "step": 10686
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005725200000000001,
+      "loss": 1.0974,
+      "step": 10687
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057248,
+      "loss": 1.0486,
+      "step": 10688
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057244,
+      "loss": 1.0157,
+      "step": 10689
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005724,
+      "loss": 1.0697,
+      "step": 10690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057236,
+      "loss": 1.0294,
+      "step": 10691
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005723200000000001,
+      "loss": 1.0476,
+      "step": 10692
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057228,
+      "loss": 1.0557,
+      "step": 10693
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057224,
+      "loss": 1.0605,
+      "step": 10694
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005722,
+      "loss": 1.0571,
+      "step": 10695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057216,
+      "loss": 1.1274,
+      "step": 10696
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057212,
+      "loss": 1.1602,
+      "step": 10697
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005720800000000001,
+      "loss": 1.0078,
+      "step": 10698
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057204,
+      "loss": 1.1116,
+      "step": 10699
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005719999999999999,
+      "loss": 0.9839,
+      "step": 10700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057196,
+      "loss": 1.1253,
+      "step": 10701
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057192,
+      "loss": 1.1279,
+      "step": 10702
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005718800000000001,
+      "loss": 1.068,
+      "step": 10703
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057184,
+      "loss": 1.057,
+      "step": 10704
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005718,
+      "loss": 1.1471,
+      "step": 10705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057176,
+      "loss": 1.1101,
+      "step": 10706
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005717200000000001,
+      "loss": 1.1228,
+      "step": 10707
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057168,
+      "loss": 1.0291,
+      "step": 10708
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057164,
+      "loss": 1.1559,
+      "step": 10709
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005716,
+      "loss": 1.0643,
+      "step": 10710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005715599999999999,
+      "loss": 1.0285,
+      "step": 10711
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005715200000000001,
+      "loss": 1.0256,
+      "step": 10712
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057148,
+      "loss": 1.0669,
+      "step": 10713
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057144,
+      "loss": 1.1344,
+      "step": 10714
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005714,
+      "loss": 1.0784,
+      "step": 10715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057136,
+      "loss": 1.046,
+      "step": 10716
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057132,
+      "loss": 1.0456,
+      "step": 10717
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005712800000000001,
+      "loss": 1.0943,
+      "step": 10718
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057124,
+      "loss": 1.0581,
+      "step": 10719
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005712,
+      "loss": 1.0986,
+      "step": 10720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057116,
+      "loss": 1.1208,
+      "step": 10721
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005711199999999999,
+      "loss": 1.0956,
+      "step": 10722
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005710800000000001,
+      "loss": 1.0549,
+      "step": 10723
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057104,
+      "loss": 0.9658,
+      "step": 10724
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000571,
+      "loss": 1.1223,
+      "step": 10725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057096,
+      "loss": 1.0584,
+      "step": 10726
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057092,
+      "loss": 1.0422,
+      "step": 10727
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005708800000000001,
+      "loss": 0.9988,
+      "step": 10728
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057084,
+      "loss": 1.0373,
+      "step": 10729
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005708,
+      "loss": 1.0692,
+      "step": 10730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057076,
+      "loss": 1.029,
+      "step": 10731
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057072,
+      "loss": 0.9559,
+      "step": 10732
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057068,
+      "loss": 1.033,
+      "step": 10733
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005706400000000001,
+      "loss": 1.1146,
+      "step": 10734
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005706,
+      "loss": 1.0728,
+      "step": 10735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057056,
+      "loss": 1.1079,
+      "step": 10736
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057052,
+      "loss": 1.0127,
+      "step": 10737
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005704799999999999,
+      "loss": 1.1458,
+      "step": 10738
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057044,
+      "loss": 1.0284,
+      "step": 10739
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005704,
+      "loss": 1.0564,
+      "step": 10740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057036,
+      "loss": 1.1076,
+      "step": 10741
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057032,
+      "loss": 1.1231,
+      "step": 10742
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057028,
+      "loss": 1.0902,
+      "step": 10743
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057024,
+      "loss": 1.0546,
+      "step": 10744
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005702000000000001,
+      "loss": 1.078,
+      "step": 10745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057016,
+      "loss": 1.0769,
+      "step": 10746
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057012,
+      "loss": 1.1228,
+      "step": 10747
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057008,
+      "loss": 1.0824,
+      "step": 10748
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057004,
+      "loss": 1.0108,
+      "step": 10749
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00057,
+      "loss": 1.0434,
+      "step": 10750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056996,
+      "loss": 1.0165,
+      "step": 10751
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056992,
+      "loss": 1.0441,
+      "step": 10752
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056988,
+      "loss": 1.0515,
+      "step": 10753
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005698400000000001,
+      "loss": 1.0477,
+      "step": 10754
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005698,
+      "loss": 1.0919,
+      "step": 10755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005697600000000001,
+      "loss": 1.0987,
+      "step": 10756
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056972,
+      "loss": 1.0885,
+      "step": 10757
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005696799999999999,
+      "loss": 1.0545,
+      "step": 10758
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005696400000000001,
+      "loss": 1.0624,
+      "step": 10759
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005696,
+      "loss": 0.9717,
+      "step": 10760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056956,
+      "loss": 1.09,
+      "step": 10761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056952,
+      "loss": 1.1965,
+      "step": 10762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056948,
+      "loss": 1.0772,
+      "step": 10763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005694399999999999,
+      "loss": 1.0012,
+      "step": 10764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005694000000000001,
+      "loss": 1.0957,
+      "step": 10765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056936,
+      "loss": 1.1265,
+      "step": 10766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005693200000000001,
+      "loss": 1.1071,
+      "step": 10767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00056928,
+      "loss": 1.1751,
+      "step": 10768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0005692399999999999,
+      "loss": 0.9871,
+      "step": 10769
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005692000000000001,
+      "loss": 1.01,
+      "step": 10770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056916,
+      "loss": 1.0604,
+      "step": 10771
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056912,
+      "loss": 1.0875,
+      "step": 10772
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056908,
+      "loss": 1.1057,
+      "step": 10773
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056904,
+      "loss": 1.0784,
+      "step": 10774
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000569,
+      "loss": 1.0357,
+      "step": 10775
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005689600000000001,
+      "loss": 1.0554,
+      "step": 10776
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056892,
+      "loss": 1.0864,
+      "step": 10777
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056888,
+      "loss": 1.0193,
+      "step": 10778
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056884,
+      "loss": 1.0784,
+      "step": 10779
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005688,
+      "loss": 1.1046,
+      "step": 10780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005687600000000001,
+      "loss": 1.0664,
+      "step": 10781
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056872,
+      "loss": 1.0518,
+      "step": 10782
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056868,
+      "loss": 1.0489,
+      "step": 10783
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056864,
+      "loss": 1.1161,
+      "step": 10784
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005686,
+      "loss": 1.0434,
+      "step": 10785
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056856,
+      "loss": 0.9999,
+      "step": 10786
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056852,
+      "loss": 1.0574,
+      "step": 10787
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056848,
+      "loss": 1.0808,
+      "step": 10788
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005684399999999999,
+      "loss": 1.0151,
+      "step": 10789
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005684,
+      "loss": 1.038,
+      "step": 10790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056836,
+      "loss": 1.0175,
+      "step": 10791
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005683200000000001,
+      "loss": 1.0926,
+      "step": 10792
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056828,
+      "loss": 0.973,
+      "step": 10793
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056824,
+      "loss": 0.9981,
+      "step": 10794
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005682,
+      "loss": 1.1244,
+      "step": 10795
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005681600000000001,
+      "loss": 0.9672,
+      "step": 10796
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056812,
+      "loss": 1.0896,
+      "step": 10797
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056808,
+      "loss": 1.0893,
+      "step": 10798
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056804,
+      "loss": 1.023,
+      "step": 10799
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005679999999999999,
+      "loss": 1.0292,
+      "step": 10800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005679600000000001,
+      "loss": 1.0905,
+      "step": 10801
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056792,
+      "loss": 1.0649,
+      "step": 10802
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005678800000000001,
+      "loss": 1.0258,
+      "step": 10803
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056784,
+      "loss": 1.074,
+      "step": 10804
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005678,
+      "loss": 1.0423,
+      "step": 10805
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056776,
+      "loss": 1.1125,
+      "step": 10806
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056772,
+      "loss": 1.0463,
+      "step": 10807
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056768,
+      "loss": 1.1095,
+      "step": 10808
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056764,
+      "loss": 1.1159,
+      "step": 10809
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005676,
+      "loss": 1.0735,
+      "step": 10810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005675599999999999,
+      "loss": 1.0475,
+      "step": 10811
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005675200000000001,
+      "loss": 1.0039,
+      "step": 10812
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056748,
+      "loss": 1.0501,
+      "step": 10813
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005674400000000001,
+      "loss": 1.023,
+      "step": 10814
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005674,
+      "loss": 0.9738,
+      "step": 10815
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005673599999999999,
+      "loss": 1.0409,
+      "step": 10816
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005673200000000001,
+      "loss": 1.054,
+      "step": 10817
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056728,
+      "loss": 1.0413,
+      "step": 10818
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056724,
+      "loss": 1.088,
+      "step": 10819
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005672,
+      "loss": 1.1232,
+      "step": 10820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056716,
+      "loss": 1.0364,
+      "step": 10821
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056712,
+      "loss": 1.0741,
+      "step": 10822
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005670800000000001,
+      "loss": 1.1316,
+      "step": 10823
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056704,
+      "loss": 1.0831,
+      "step": 10824
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000567,
+      "loss": 1.1094,
+      "step": 10825
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056696,
+      "loss": 1.1042,
+      "step": 10826
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005669199999999999,
+      "loss": 1.0557,
+      "step": 10827
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005668800000000001,
+      "loss": 1.0427,
+      "step": 10828
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056684,
+      "loss": 1.0517,
+      "step": 10829
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005668,
+      "loss": 1.0199,
+      "step": 10830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056676,
+      "loss": 1.0338,
+      "step": 10831
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056672,
+      "loss": 1.055,
+      "step": 10832
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056668,
+      "loss": 1.0194,
+      "step": 10833
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005666400000000001,
+      "loss": 1.0429,
+      "step": 10834
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005666,
+      "loss": 1.0568,
+      "step": 10835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005665599999999999,
+      "loss": 1.1398,
+      "step": 10836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056652,
+      "loss": 1.0425,
+      "step": 10837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056648,
+      "loss": 0.9953,
+      "step": 10838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005664400000000001,
+      "loss": 1.056,
+      "step": 10839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005664,
+      "loss": 1.0084,
+      "step": 10840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056636,
+      "loss": 1.0433,
+      "step": 10841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056632,
+      "loss": 1.0874,
+      "step": 10842
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005662800000000001,
+      "loss": 1.0425,
+      "step": 10843
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056624,
+      "loss": 1.0119,
+      "step": 10844
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005662000000000001,
+      "loss": 1.1204,
+      "step": 10845
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056616,
+      "loss": 1.1256,
+      "step": 10846
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005661199999999999,
+      "loss": 1.1315,
+      "step": 10847
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005660800000000001,
+      "loss": 1.0112,
+      "step": 10848
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056604,
+      "loss": 1.0074,
+      "step": 10849
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000566,
+      "loss": 1.09,
+      "step": 10850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056596,
+      "loss": 1.0826,
+      "step": 10851
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056592,
+      "loss": 1.0616,
+      "step": 10852
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056588,
+      "loss": 1.1908,
+      "step": 10853
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005658400000000001,
+      "loss": 1.0724,
+      "step": 10854
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005658,
+      "loss": 1.0902,
+      "step": 10855
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056576,
+      "loss": 1.108,
+      "step": 10856
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056572,
+      "loss": 1.0862,
+      "step": 10857
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005656799999999999,
+      "loss": 0.9918,
+      "step": 10858
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005656400000000001,
+      "loss": 1.1171,
+      "step": 10859
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005656,
+      "loss": 1.0667,
+      "step": 10860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056556,
+      "loss": 0.918,
+      "step": 10861
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056552,
+      "loss": 1.0331,
+      "step": 10862
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056548,
+      "loss": 1.0497,
+      "step": 10863
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005654400000000001,
+      "loss": 1.097,
+      "step": 10864
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005654,
+      "loss": 1.0924,
+      "step": 10865
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056536,
+      "loss": 1.076,
+      "step": 10866
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056532,
+      "loss": 1.0772,
+      "step": 10867
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056528,
+      "loss": 1.0078,
+      "step": 10868
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056524,
+      "loss": 1.0017,
+      "step": 10869
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005652000000000001,
+      "loss": 1.1232,
+      "step": 10870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056516,
+      "loss": 1.0052,
+      "step": 10871
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056512,
+      "loss": 1.0101,
+      "step": 10872
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056508,
+      "loss": 1.0047,
+      "step": 10873
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056504,
+      "loss": 1.0429,
+      "step": 10874
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000565,
+      "loss": 1.1189,
+      "step": 10875
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056496,
+      "loss": 1.0732,
+      "step": 10876
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056492,
+      "loss": 1.0763,
+      "step": 10877
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056488,
+      "loss": 0.9882,
+      "step": 10878
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056484,
+      "loss": 1.0407,
+      "step": 10879
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005648,
+      "loss": 1.0534,
+      "step": 10880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005647600000000001,
+      "loss": 1.1106,
+      "step": 10881
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056472,
+      "loss": 1.0769,
+      "step": 10882
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056468,
+      "loss": 1.1458,
+      "step": 10883
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056464,
+      "loss": 1.0887,
+      "step": 10884
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005646,
+      "loss": 1.021,
+      "step": 10885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056456,
+      "loss": 1.0194,
+      "step": 10886
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056452,
+      "loss": 1.0737,
+      "step": 10887
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056448,
+      "loss": 1.1484,
+      "step": 10888
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056444,
+      "loss": 1.049,
+      "step": 10889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005644000000000001,
+      "loss": 1.0566,
+      "step": 10890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056436,
+      "loss": 1.0447,
+      "step": 10891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005643200000000001,
+      "loss": 0.9931,
+      "step": 10892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056428,
+      "loss": 1.0915,
+      "step": 10893
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005642399999999999,
+      "loss": 1.0106,
+      "step": 10894
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005642,
+      "loss": 1.0125,
+      "step": 10895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056416,
+      "loss": 1.083,
+      "step": 10896
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056412,
+      "loss": 0.9931,
+      "step": 10897
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056408,
+      "loss": 0.9512,
+      "step": 10898
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056404,
+      "loss": 1.101,
+      "step": 10899
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005639999999999999,
+      "loss": 1.0744,
+      "step": 10900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005639600000000001,
+      "loss": 1.0712,
+      "step": 10901
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056392,
+      "loss": 1.0377,
+      "step": 10902
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005638800000000001,
+      "loss": 1.0954,
+      "step": 10903
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056384,
+      "loss": 1.1617,
+      "step": 10904
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005637999999999999,
+      "loss": 1.0711,
+      "step": 10905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005637600000000001,
+      "loss": 1.0717,
+      "step": 10906
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056372,
+      "loss": 0.9893,
+      "step": 10907
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056368,
+      "loss": 1.0625,
+      "step": 10908
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056364,
+      "loss": 1.0629,
+      "step": 10909
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005636,
+      "loss": 1.0366,
+      "step": 10910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056356,
+      "loss": 1.1072,
+      "step": 10911
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005635200000000001,
+      "loss": 1.0702,
+      "step": 10912
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056348,
+      "loss": 1.1191,
+      "step": 10913
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056344,
+      "loss": 0.9708,
+      "step": 10914
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005634,
+      "loss": 1.0022,
+      "step": 10915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056336,
+      "loss": 1.0197,
+      "step": 10916
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005633200000000001,
+      "loss": 1.0215,
+      "step": 10917
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056328,
+      "loss": 1.0517,
+      "step": 10918
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056324,
+      "loss": 1.022,
+      "step": 10919
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005632,
+      "loss": 1.1361,
+      "step": 10920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056316,
+      "loss": 1.089,
+      "step": 10921
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056312,
+      "loss": 1.0865,
+      "step": 10922
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005630800000000001,
+      "loss": 1.067,
+      "step": 10923
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056304,
+      "loss": 1.0743,
+      "step": 10924
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005629999999999999,
+      "loss": 1.0208,
+      "step": 10925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056296,
+      "loss": 1.1713,
+      "step": 10926
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056292,
+      "loss": 1.1421,
+      "step": 10927
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005628800000000001,
+      "loss": 1.0431,
+      "step": 10928
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056284,
+      "loss": 1.0302,
+      "step": 10929
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005628,
+      "loss": 1.0632,
+      "step": 10930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056276,
+      "loss": 1.0368,
+      "step": 10931
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005627200000000001,
+      "loss": 1.0121,
+      "step": 10932
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056268,
+      "loss": 1.0447,
+      "step": 10933
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056264,
+      "loss": 0.987,
+      "step": 10934
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005626,
+      "loss": 1.0635,
+      "step": 10935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005625599999999999,
+      "loss": 1.013,
+      "step": 10936
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005625200000000001,
+      "loss": 1.1242,
+      "step": 10937
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056248,
+      "loss": 0.9646,
+      "step": 10938
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005624400000000001,
+      "loss": 1.0037,
+      "step": 10939
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005624,
+      "loss": 1.0058,
+      "step": 10940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056236,
+      "loss": 1.1086,
+      "step": 10941
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056232,
+      "loss": 1.0801,
+      "step": 10942
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056228,
+      "loss": 1.0117,
+      "step": 10943
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056224,
+      "loss": 1.0787,
+      "step": 10944
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005622,
+      "loss": 1.0847,
+      "step": 10945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056216,
+      "loss": 1.0429,
+      "step": 10946
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005621199999999999,
+      "loss": 1.0109,
+      "step": 10947
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005620800000000001,
+      "loss": 1.0563,
+      "step": 10948
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056204,
+      "loss": 1.0885,
+      "step": 10949
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005620000000000001,
+      "loss": 1.0452,
+      "step": 10950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056196,
+      "loss": 1.0282,
+      "step": 10951
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005619199999999999,
+      "loss": 1.1429,
+      "step": 10952
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005618800000000001,
+      "loss": 0.9994,
+      "step": 10953
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056184,
+      "loss": 0.9692,
+      "step": 10954
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005618,
+      "loss": 0.9944,
+      "step": 10955
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056176,
+      "loss": 0.9712,
+      "step": 10956
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056172,
+      "loss": 1.0274,
+      "step": 10957
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056168,
+      "loss": 1.0136,
+      "step": 10958
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005616400000000001,
+      "loss": 1.0655,
+      "step": 10959
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005616,
+      "loss": 0.9207,
+      "step": 10960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056156,
+      "loss": 1.0969,
+      "step": 10961
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056152,
+      "loss": 1.1391,
+      "step": 10962
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005614799999999999,
+      "loss": 0.9752,
+      "step": 10963
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005614400000000001,
+      "loss": 1.0603,
+      "step": 10964
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005614,
+      "loss": 1.0592,
+      "step": 10965
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056136,
+      "loss": 1.1204,
+      "step": 10966
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056132,
+      "loss": 1.0487,
+      "step": 10967
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056128,
+      "loss": 0.9687,
+      "step": 10968
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056124,
+      "loss": 1.0391,
+      "step": 10969
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005612000000000001,
+      "loss": 1.1357,
+      "step": 10970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056116,
+      "loss": 1.0892,
+      "step": 10971
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005611199999999999,
+      "loss": 1.0881,
+      "step": 10972
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056108,
+      "loss": 1.0594,
+      "step": 10973
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056104,
+      "loss": 1.1168,
+      "step": 10974
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005610000000000001,
+      "loss": 1.0448,
+      "step": 10975
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056096,
+      "loss": 0.951,
+      "step": 10976
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056092,
+      "loss": 1.0226,
+      "step": 10977
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056088,
+      "loss": 1.1644,
+      "step": 10978
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005608400000000001,
+      "loss": 1.0073,
+      "step": 10979
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005608,
+      "loss": 1.0693,
+      "step": 10980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005607600000000001,
+      "loss": 1.098,
+      "step": 10981
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056072,
+      "loss": 1.0529,
+      "step": 10982
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005606799999999999,
+      "loss": 0.9613,
+      "step": 10983
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056064,
+      "loss": 0.9633,
+      "step": 10984
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005606,
+      "loss": 1.0588,
+      "step": 10985
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056056,
+      "loss": 1.0361,
+      "step": 10986
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056052,
+      "loss": 1.004,
+      "step": 10987
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056048,
+      "loss": 1.0388,
+      "step": 10988
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056044,
+      "loss": 1.0617,
+      "step": 10989
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005604000000000001,
+      "loss": 1.0098,
+      "step": 10990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056036,
+      "loss": 1.207,
+      "step": 10991
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056032,
+      "loss": 1.0923,
+      "step": 10992
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056028,
+      "loss": 1.0607,
+      "step": 10993
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005602399999999999,
+      "loss": 1.0339,
+      "step": 10994
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005602000000000001,
+      "loss": 1.1052,
+      "step": 10995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056016,
+      "loss": 1.1014,
+      "step": 10996
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056012,
+      "loss": 0.9429,
+      "step": 10997
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056008,
+      "loss": 1.0639,
+      "step": 10998
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00056004,
+      "loss": 1.0998,
+      "step": 10999
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005600000000000001,
+      "loss": 1.1525,
+      "step": 11000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055996,
+      "loss": 1.0852,
+      "step": 11001
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055992,
+      "loss": 1.1299,
+      "step": 11002
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055988,
+      "loss": 1.0677,
+      "step": 11003
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055984,
+      "loss": 1.0645,
+      "step": 11004
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005598,
+      "loss": 1.0655,
+      "step": 11005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005597600000000001,
+      "loss": 1.0588,
+      "step": 11006
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055972,
+      "loss": 1.0768,
+      "step": 11007
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055968,
+      "loss": 1.1487,
+      "step": 11008
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055964,
+      "loss": 1.108,
+      "step": 11009
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005596,
+      "loss": 1.0254,
+      "step": 11010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055956,
+      "loss": 1.021,
+      "step": 11011
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055952,
+      "loss": 1.0773,
+      "step": 11012
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055948,
+      "loss": 1.1788,
+      "step": 11013
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055944,
+      "loss": 1.0214,
+      "step": 11014
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005594,
+      "loss": 1.0406,
+      "step": 11015
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055936,
+      "loss": 1.1173,
+      "step": 11016
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005593200000000001,
+      "loss": 1.1293,
+      "step": 11017
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055928,
+      "loss": 1.1175,
+      "step": 11018
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055924,
+      "loss": 0.957,
+      "step": 11019
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0005592,
+      "loss": 1.0782,
+      "step": 11020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055916,
+      "loss": 1.0791,
+      "step": 11021
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055912,
+      "loss": 1.1438,
+      "step": 11022
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00055908,
+      "loss": 1.0122,
+      "step": 11023
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055904,
+      "loss": 1.0228,
+      "step": 11024
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000559,
+      "loss": 1.0388,
+      "step": 11025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005589600000000001,
+      "loss": 1.0524,
+      "step": 11026
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055892,
+      "loss": 1.0646,
+      "step": 11027
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005588800000000001,
+      "loss": 1.127,
+      "step": 11028
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055884,
+      "loss": 1.1258,
+      "step": 11029
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005587999999999999,
+      "loss": 1.0538,
+      "step": 11030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055876,
+      "loss": 1.0585,
+      "step": 11031
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055872,
+      "loss": 1.0813,
+      "step": 11032
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055868,
+      "loss": 0.9776,
+      "step": 11033
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055864,
+      "loss": 1.0799,
+      "step": 11034
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005586,
+      "loss": 1.0455,
+      "step": 11035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005585599999999999,
+      "loss": 1.0057,
+      "step": 11036
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005585200000000001,
+      "loss": 1.0253,
+      "step": 11037
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055848,
+      "loss": 1.054,
+      "step": 11038
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005584400000000001,
+      "loss": 1.0563,
+      "step": 11039
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005584,
+      "loss": 1.1707,
+      "step": 11040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005583599999999999,
+      "loss": 1.1358,
+      "step": 11041
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005583200000000001,
+      "loss": 1.0763,
+      "step": 11042
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055828,
+      "loss": 1.05,
+      "step": 11043
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055824,
+      "loss": 1.0506,
+      "step": 11044
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005582,
+      "loss": 1.0387,
+      "step": 11045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055816,
+      "loss": 1.0128,
+      "step": 11046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055812,
+      "loss": 1.0813,
+      "step": 11047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005580800000000001,
+      "loss": 1.1378,
+      "step": 11048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055804,
+      "loss": 1.0556,
+      "step": 11049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000558,
+      "loss": 1.0237,
+      "step": 11050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055796,
+      "loss": 1.1065,
+      "step": 11051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005579199999999999,
+      "loss": 1.1072,
+      "step": 11052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005578800000000001,
+      "loss": 1.0243,
+      "step": 11053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055784,
+      "loss": 0.983,
+      "step": 11054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005578,
+      "loss": 1.0512,
+      "step": 11055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055776,
+      "loss": 1.0362,
+      "step": 11056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055772,
+      "loss": 1.0602,
+      "step": 11057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055768,
+      "loss": 1.0126,
+      "step": 11058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005576400000000001,
+      "loss": 1.0769,
+      "step": 11059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005576,
+      "loss": 1.0916,
+      "step": 11060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005575599999999999,
+      "loss": 1.0218,
+      "step": 11061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055752,
+      "loss": 1.1008,
+      "step": 11062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055748,
+      "loss": 1.0866,
+      "step": 11063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005574400000000001,
+      "loss": 1.0355,
+      "step": 11064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005574,
+      "loss": 1.0692,
+      "step": 11065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055736,
+      "loss": 1.1361,
+      "step": 11066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055732,
+      "loss": 1.0172,
+      "step": 11067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005572800000000001,
+      "loss": 1.1839,
+      "step": 11068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055724,
+      "loss": 1.0558,
+      "step": 11069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005572,
+      "loss": 1.1025,
+      "step": 11070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055716,
+      "loss": 1.0739,
+      "step": 11071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005571199999999999,
+      "loss": 0.9955,
+      "step": 11072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055708,
+      "loss": 1.0706,
+      "step": 11073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055704,
+      "loss": 0.9936,
+      "step": 11074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005570000000000001,
+      "loss": 1.0684,
+      "step": 11075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055696,
+      "loss": 1.113,
+      "step": 11076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055692,
+      "loss": 1.0638,
+      "step": 11077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055688,
+      "loss": 1.0906,
+      "step": 11078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055684,
+      "loss": 1.0624,
+      "step": 11079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005568,
+      "loss": 1.081,
+      "step": 11080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055676,
+      "loss": 1.0221,
+      "step": 11081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055672,
+      "loss": 1.0704,
+      "step": 11082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005566799999999999,
+      "loss": 1.0101,
+      "step": 11083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005566400000000001,
+      "loss": 1.0494,
+      "step": 11084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005566,
+      "loss": 1.011,
+      "step": 11085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005565600000000001,
+      "loss": 1.1188,
+      "step": 11086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055652,
+      "loss": 1.0874,
+      "step": 11087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055648,
+      "loss": 1.0787,
+      "step": 11088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005564400000000001,
+      "loss": 1.0085,
+      "step": 11089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005564,
+      "loss": 1.1456,
+      "step": 11090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055636,
+      "loss": 1.0951,
+      "step": 11091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055632,
+      "loss": 1.0388,
+      "step": 11092
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055628,
+      "loss": 1.0379,
+      "step": 11093
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055624,
+      "loss": 1.0409,
+      "step": 11094
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005562000000000001,
+      "loss": 1.0589,
+      "step": 11095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055616,
+      "loss": 1.09,
+      "step": 11096
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055612,
+      "loss": 1.0915,
+      "step": 11097
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055608,
+      "loss": 1.0014,
+      "step": 11098
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005560399999999999,
+      "loss": 1.0453,
+      "step": 11099
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005560000000000001,
+      "loss": 0.9963,
+      "step": 11100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055596,
+      "loss": 1.0932,
+      "step": 11101
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055592,
+      "loss": 1.136,
+      "step": 11102
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055588,
+      "loss": 1.0995,
+      "step": 11103
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055584,
+      "loss": 0.9886,
+      "step": 11104
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005558,
+      "loss": 1.044,
+      "step": 11105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005557600000000001,
+      "loss": 1.0652,
+      "step": 11106
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055572,
+      "loss": 0.9698,
+      "step": 11107
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005556799999999999,
+      "loss": 0.9897,
+      "step": 11108
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055564,
+      "loss": 1.0138,
+      "step": 11109
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005556,
+      "loss": 0.973,
+      "step": 11110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005555600000000001,
+      "loss": 1.0622,
+      "step": 11111
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055552,
+      "loss": 1.0321,
+      "step": 11112
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055548,
+      "loss": 0.9989,
+      "step": 11113
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055544,
+      "loss": 1.0211,
+      "step": 11114
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005554000000000001,
+      "loss": 1.0427,
+      "step": 11115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055536,
+      "loss": 1.1447,
+      "step": 11116
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005553200000000001,
+      "loss": 1.0151,
+      "step": 11117
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055528,
+      "loss": 1.0169,
+      "step": 11118
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005552399999999999,
+      "loss": 1.0432,
+      "step": 11119
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005552,
+      "loss": 1.0509,
+      "step": 11120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055516,
+      "loss": 1.0796,
+      "step": 11121
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055512,
+      "loss": 1.0701,
+      "step": 11122
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055508,
+      "loss": 1.0525,
+      "step": 11123
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055504,
+      "loss": 1.1183,
+      "step": 11124
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000555,
+      "loss": 1.0375,
+      "step": 11125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005549600000000001,
+      "loss": 1.1015,
+      "step": 11126
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055492,
+      "loss": 1.0818,
+      "step": 11127
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055488,
+      "loss": 1.0573,
+      "step": 11128
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055484,
+      "loss": 1.1299,
+      "step": 11129
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005547999999999999,
+      "loss": 1.0191,
+      "step": 11130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005547600000000001,
+      "loss": 1.0542,
+      "step": 11131
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055472,
+      "loss": 1.0965,
+      "step": 11132
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055468,
+      "loss": 1.0512,
+      "step": 11133
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055464,
+      "loss": 1.0687,
+      "step": 11134
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005546,
+      "loss": 1.0669,
+      "step": 11135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005545600000000001,
+      "loss": 0.9944,
+      "step": 11136
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005545200000000001,
+      "loss": 1.1124,
+      "step": 11137
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055448,
+      "loss": 1.0534,
+      "step": 11138
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055444,
+      "loss": 0.9595,
+      "step": 11139
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005544,
+      "loss": 1.0739,
+      "step": 11140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005543599999999999,
+      "loss": 1.054,
+      "step": 11141
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005543200000000001,
+      "loss": 1.0073,
+      "step": 11142
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055428,
+      "loss": 1.0525,
+      "step": 11143
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055424,
+      "loss": 1.0524,
+      "step": 11144
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005542,
+      "loss": 0.9873,
+      "step": 11145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055416,
+      "loss": 1.0663,
+      "step": 11146
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055412,
+      "loss": 1.0824,
+      "step": 11147
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055408,
+      "loss": 1.0795,
+      "step": 11148
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055404,
+      "loss": 1.0811,
+      "step": 11149
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000554,
+      "loss": 1.0161,
+      "step": 11150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055396,
+      "loss": 0.9799,
+      "step": 11151
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055392,
+      "loss": 1.0035,
+      "step": 11152
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005538800000000001,
+      "loss": 1.0619,
+      "step": 11153
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055384,
+      "loss": 1.1109,
+      "step": 11154
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005538,
+      "loss": 1.0953,
+      "step": 11155
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055376,
+      "loss": 1.0974,
+      "step": 11156
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055372,
+      "loss": 1.0209,
+      "step": 11157
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055368,
+      "loss": 1.1423,
+      "step": 11158
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055364,
+      "loss": 1.1045,
+      "step": 11159
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005536,
+      "loss": 1.0307,
+      "step": 11160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055356,
+      "loss": 1.106,
+      "step": 11161
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055352,
+      "loss": 0.9948,
+      "step": 11162
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055348,
+      "loss": 1.1387,
+      "step": 11163
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005534400000000001,
+      "loss": 1.0158,
+      "step": 11164
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005534,
+      "loss": 1.0715,
+      "step": 11165
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055336,
+      "loss": 1.1336,
+      "step": 11166
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055332,
+      "loss": 1.1228,
+      "step": 11167
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055328,
+      "loss": 0.9965,
+      "step": 11168
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055324,
+      "loss": 1.0203,
+      "step": 11169
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005532,
+      "loss": 0.9676,
+      "step": 11170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055316,
+      "loss": 1.0282,
+      "step": 11171
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005531199999999999,
+      "loss": 1.0037,
+      "step": 11172
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005530800000000001,
+      "loss": 1.0059,
+      "step": 11173
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055304,
+      "loss": 1.0366,
+      "step": 11174
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005530000000000001,
+      "loss": 1.0193,
+      "step": 11175
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055296,
+      "loss": 1.0458,
+      "step": 11176
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005529199999999999,
+      "loss": 1.0862,
+      "step": 11177
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005528800000000001,
+      "loss": 1.0655,
+      "step": 11178
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055284,
+      "loss": 0.9918,
+      "step": 11179
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005528,
+      "loss": 1.0083,
+      "step": 11180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055276,
+      "loss": 1.0824,
+      "step": 11181
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055272,
+      "loss": 1.0117,
+      "step": 11182
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055268,
+      "loss": 1.0015,
+      "step": 11183
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005526400000000001,
+      "loss": 1.1299,
+      "step": 11184
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005526,
+      "loss": 0.9684,
+      "step": 11185
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055256,
+      "loss": 1.0091,
+      "step": 11186
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055252,
+      "loss": 0.9971,
+      "step": 11187
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005524799999999999,
+      "loss": 0.9868,
+      "step": 11188
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005524400000000001,
+      "loss": 1.0918,
+      "step": 11189
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005524,
+      "loss": 1.0473,
+      "step": 11190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055236,
+      "loss": 1.0916,
+      "step": 11191
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055232,
+      "loss": 1.0785,
+      "step": 11192
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055228,
+      "loss": 1.0623,
+      "step": 11193
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055224,
+      "loss": 1.136,
+      "step": 11194
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005522000000000001,
+      "loss": 1.0724,
+      "step": 11195
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055216,
+      "loss": 1.1243,
+      "step": 11196
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055212,
+      "loss": 1.0658,
+      "step": 11197
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055208,
+      "loss": 1.1158,
+      "step": 11198
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055204,
+      "loss": 1.01,
+      "step": 11199
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005520000000000001,
+      "loss": 1.0769,
+      "step": 11200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055196,
+      "loss": 1.0072,
+      "step": 11201
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055192,
+      "loss": 1.1382,
+      "step": 11202
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055188,
+      "loss": 1.0467,
+      "step": 11203
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005518400000000001,
+      "loss": 1.0848,
+      "step": 11204
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005518,
+      "loss": 1.0773,
+      "step": 11205
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055176,
+      "loss": 1.0763,
+      "step": 11206
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055172,
+      "loss": 0.9732,
+      "step": 11207
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005516799999999999,
+      "loss": 1.0369,
+      "step": 11208
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055164,
+      "loss": 1.1246,
+      "step": 11209
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005516,
+      "loss": 1.0413,
+      "step": 11210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005515600000000001,
+      "loss": 1.0941,
+      "step": 11211
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055152,
+      "loss": 0.9765,
+      "step": 11212
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055148,
+      "loss": 1.0585,
+      "step": 11213
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055144,
+      "loss": 1.1441,
+      "step": 11214
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005514,
+      "loss": 1.0849,
+      "step": 11215
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055136,
+      "loss": 1.1257,
+      "step": 11216
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055132,
+      "loss": 0.9756,
+      "step": 11217
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055128,
+      "loss": 1.0082,
+      "step": 11218
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005512399999999999,
+      "loss": 1.0251,
+      "step": 11219
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005512000000000001,
+      "loss": 1.0303,
+      "step": 11220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055116,
+      "loss": 1.072,
+      "step": 11221
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005511200000000001,
+      "loss": 1.1214,
+      "step": 11222
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055108,
+      "loss": 1.0035,
+      "step": 11223
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055104,
+      "loss": 1.0356,
+      "step": 11224
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005510000000000001,
+      "loss": 0.9727,
+      "step": 11225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055096,
+      "loss": 1.0742,
+      "step": 11226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055092,
+      "loss": 0.99,
+      "step": 11227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055088,
+      "loss": 0.9507,
+      "step": 11228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055084,
+      "loss": 1.0169,
+      "step": 11229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005507999999999999,
+      "loss": 1.0263,
+      "step": 11230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005507600000000001,
+      "loss": 0.9585,
+      "step": 11231
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055072,
+      "loss": 1.0177,
+      "step": 11232
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055068,
+      "loss": 1.0216,
+      "step": 11233
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055064,
+      "loss": 1.0865,
+      "step": 11234
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005505999999999999,
+      "loss": 1.0106,
+      "step": 11235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005505600000000001,
+      "loss": 1.0217,
+      "step": 11236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055052,
+      "loss": 1.0137,
+      "step": 11237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055048,
+      "loss": 1.0962,
+      "step": 11238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055044,
+      "loss": 1.0142,
+      "step": 11239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005504,
+      "loss": 1.0592,
+      "step": 11240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055036,
+      "loss": 1.0721,
+      "step": 11241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005503200000000001,
+      "loss": 1.0283,
+      "step": 11242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055028,
+      "loss": 1.0572,
+      "step": 11243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005502399999999999,
+      "loss": 1.0863,
+      "step": 11244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005502,
+      "loss": 1.123,
+      "step": 11245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055016,
+      "loss": 1.0366,
+      "step": 11246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005501200000000001,
+      "loss": 1.0307,
+      "step": 11247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055008,
+      "loss": 1.0121,
+      "step": 11248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055004,
+      "loss": 1.0093,
+      "step": 11249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00055,
+      "loss": 1.0013,
+      "step": 11250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054996,
+      "loss": 1.066,
+      "step": 11251
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054992,
+      "loss": 0.9969,
+      "step": 11252
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005498800000000001,
+      "loss": 1.1192,
+      "step": 11253
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054984,
+      "loss": 0.9991,
+      "step": 11254
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005497999999999999,
+      "loss": 1.0357,
+      "step": 11255
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054976,
+      "loss": 1.0964,
+      "step": 11256
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054972,
+      "loss": 1.1654,
+      "step": 11257
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054968,
+      "loss": 1.1071,
+      "step": 11258
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054964,
+      "loss": 1.1394,
+      "step": 11259
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005496,
+      "loss": 1.0403,
+      "step": 11260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054956,
+      "loss": 0.9978,
+      "step": 11261
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005495200000000001,
+      "loss": 1.0798,
+      "step": 11262
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054948,
+      "loss": 0.9923,
+      "step": 11263
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054944,
+      "loss": 1.0832,
+      "step": 11264
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005494,
+      "loss": 1.0939,
+      "step": 11265
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005493599999999999,
+      "loss": 1.0624,
+      "step": 11266
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005493200000000001,
+      "loss": 1.0901,
+      "step": 11267
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054928,
+      "loss": 1.1353,
+      "step": 11268
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054924,
+      "loss": 1.0826,
+      "step": 11269
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005492,
+      "loss": 1.0377,
+      "step": 11270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054916,
+      "loss": 1.1202,
+      "step": 11271
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005491200000000001,
+      "loss": 1.1173,
+      "step": 11272
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0005490800000000001,
+      "loss": 1.1189,
+      "step": 11273
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054904,
+      "loss": 1.0224,
+      "step": 11274
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000549,
+      "loss": 0.9605,
+      "step": 11275
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00054896,
+      "loss": 1.0459,
+      "step": 11276
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005489199999999999,
+      "loss": 1.041,
+      "step": 11277
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005488800000000001,
+      "loss": 1.1346,
+      "step": 11278
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054884,
+      "loss": 1.0931,
+      "step": 11279
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005488,
+      "loss": 0.9536,
+      "step": 11280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054876,
+      "loss": 0.9709,
+      "step": 11281
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054872,
+      "loss": 1.0481,
+      "step": 11282
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054868,
+      "loss": 1.0636,
+      "step": 11283
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054864,
+      "loss": 1.0658,
+      "step": 11284
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005486,
+      "loss": 1.0949,
+      "step": 11285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054856,
+      "loss": 1.0081,
+      "step": 11286
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054852,
+      "loss": 1.0655,
+      "step": 11287
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054848,
+      "loss": 1.0482,
+      "step": 11288
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005484400000000001,
+      "loss": 1.0803,
+      "step": 11289
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005484,
+      "loss": 0.9995,
+      "step": 11290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054836,
+      "loss": 1.0336,
+      "step": 11291
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054832,
+      "loss": 1.0387,
+      "step": 11292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054828,
+      "loss": 1.0883,
+      "step": 11293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054824,
+      "loss": 1.0763,
+      "step": 11294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005482,
+      "loss": 1.0388,
+      "step": 11295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054816,
+      "loss": 1.04,
+      "step": 11296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054812,
+      "loss": 0.9909,
+      "step": 11297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054808,
+      "loss": 1.0657,
+      "step": 11298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054804,
+      "loss": 1.0799,
+      "step": 11299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005480000000000001,
+      "loss": 0.9963,
+      "step": 11300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054796,
+      "loss": 1.0214,
+      "step": 11301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054792,
+      "loss": 1.0152,
+      "step": 11302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054788,
+      "loss": 1.0942,
+      "step": 11303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054784,
+      "loss": 1.0104,
+      "step": 11304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005478,
+      "loss": 1.0248,
+      "step": 11305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054776,
+      "loss": 0.9941,
+      "step": 11306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054772,
+      "loss": 1.0618,
+      "step": 11307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005476799999999999,
+      "loss": 1.0438,
+      "step": 11308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005476400000000001,
+      "loss": 1.0496,
+      "step": 11309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005476,
+      "loss": 1.0813,
+      "step": 11310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005475600000000001,
+      "loss": 1.0417,
+      "step": 11311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054752,
+      "loss": 1.1279,
+      "step": 11312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005474799999999999,
+      "loss": 1.1882,
+      "step": 11313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005474400000000001,
+      "loss": 1.0713,
+      "step": 11314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005474,
+      "loss": 1.0726,
+      "step": 11315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054736,
+      "loss": 1.0116,
+      "step": 11316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054732,
+      "loss": 1.1452,
+      "step": 11317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054728,
+      "loss": 1.0973,
+      "step": 11318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005472399999999999,
+      "loss": 0.9917,
+      "step": 11319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005472000000000001,
+      "loss": 1.0656,
+      "step": 11320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054716,
+      "loss": 1.0267,
+      "step": 11321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005471200000000001,
+      "loss": 1.0683,
+      "step": 11322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054708,
+      "loss": 1.0348,
+      "step": 11323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005470399999999999,
+      "loss": 1.0852,
+      "step": 11324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005470000000000001,
+      "loss": 1.0206,
+      "step": 11325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054696,
+      "loss": 1.0984,
+      "step": 11326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054692,
+      "loss": 1.0874,
+      "step": 11327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054688,
+      "loss": 1.0491,
+      "step": 11328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054684,
+      "loss": 1.0751,
+      "step": 11329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005468,
+      "loss": 1.078,
+      "step": 11330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005467600000000001,
+      "loss": 1.0653,
+      "step": 11331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054672,
+      "loss": 1.0989,
+      "step": 11332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054668,
+      "loss": 1.0239,
+      "step": 11333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054664,
+      "loss": 1.0698,
+      "step": 11334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005466,
+      "loss": 1.0237,
+      "step": 11335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005465600000000001,
+      "loss": 1.029,
+      "step": 11336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054652,
+      "loss": 1.0853,
+      "step": 11337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054648,
+      "loss": 1.0739,
+      "step": 11338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054644,
+      "loss": 1.0068,
+      "step": 11339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005464,
+      "loss": 0.9765,
+      "step": 11340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054636,
+      "loss": 1.0096,
+      "step": 11341
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054632,
+      "loss": 0.9991,
+      "step": 11342
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054628,
+      "loss": 1.0509,
+      "step": 11343
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005462399999999999,
+      "loss": 1.0824,
+      "step": 11344
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005462,
+      "loss": 1.0398,
+      "step": 11345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054616,
+      "loss": 0.9864,
+      "step": 11346
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005461200000000001,
+      "loss": 1.0455,
+      "step": 11347
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054608,
+      "loss": 1.0647,
+      "step": 11348
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054604,
+      "loss": 1.0493,
+      "step": 11349
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000546,
+      "loss": 1.0566,
+      "step": 11350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005459600000000001,
+      "loss": 1.1452,
+      "step": 11351
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054592,
+      "loss": 1.0817,
+      "step": 11352
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054588,
+      "loss": 1.035,
+      "step": 11353
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054584,
+      "loss": 1.056,
+      "step": 11354
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005457999999999999,
+      "loss": 1.0369,
+      "step": 11355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005457600000000001,
+      "loss": 1.0258,
+      "step": 11356
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054572,
+      "loss": 1.0333,
+      "step": 11357
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005456800000000001,
+      "loss": 1.0322,
+      "step": 11358
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054564,
+      "loss": 0.9996,
+      "step": 11359
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005456,
+      "loss": 1.0697,
+      "step": 11360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005455600000000001,
+      "loss": 0.9929,
+      "step": 11361
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054552,
+      "loss": 0.9397,
+      "step": 11362
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054548,
+      "loss": 1.0197,
+      "step": 11363
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054544,
+      "loss": 1.0036,
+      "step": 11364
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005454,
+      "loss": 1.018,
+      "step": 11365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005453599999999999,
+      "loss": 1.0299,
+      "step": 11366
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005453200000000001,
+      "loss": 1.0405,
+      "step": 11367
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054528,
+      "loss": 1.0382,
+      "step": 11368
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054524,
+      "loss": 1.0894,
+      "step": 11369
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005452,
+      "loss": 1.0746,
+      "step": 11370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005451599999999999,
+      "loss": 1.1129,
+      "step": 11371
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005451200000000001,
+      "loss": 1.0303,
+      "step": 11372
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054508,
+      "loss": 1.0564,
+      "step": 11373
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054504,
+      "loss": 1.0032,
+      "step": 11374
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000545,
+      "loss": 1.0638,
+      "step": 11375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054496,
+      "loss": 0.9711,
+      "step": 11376
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054492,
+      "loss": 1.0092,
+      "step": 11377
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005448800000000001,
+      "loss": 1.073,
+      "step": 11378
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054484,
+      "loss": 1.0087,
+      "step": 11379
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005448,
+      "loss": 1.0471,
+      "step": 11380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054476,
+      "loss": 1.0908,
+      "step": 11381
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054472,
+      "loss": 1.1086,
+      "step": 11382
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005446800000000001,
+      "loss": 1.0914,
+      "step": 11383
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054464,
+      "loss": 1.002,
+      "step": 11384
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005446,
+      "loss": 1.0737,
+      "step": 11385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054456,
+      "loss": 1.1226,
+      "step": 11386
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054452,
+      "loss": 1.1036,
+      "step": 11387
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054448,
+      "loss": 1.0396,
+      "step": 11388
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005444400000000001,
+      "loss": 1.0889,
+      "step": 11389
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005444,
+      "loss": 1.0546,
+      "step": 11390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005443599999999999,
+      "loss": 1.0741,
+      "step": 11391
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054432,
+      "loss": 1.0144,
+      "step": 11392
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054428,
+      "loss": 1.1186,
+      "step": 11393
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054424,
+      "loss": 1.0947,
+      "step": 11394
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005442,
+      "loss": 1.0513,
+      "step": 11395
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054416,
+      "loss": 1.0544,
+      "step": 11396
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054412,
+      "loss": 1.0464,
+      "step": 11397
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005440800000000001,
+      "loss": 1.1685,
+      "step": 11398
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054404,
+      "loss": 1.003,
+      "step": 11399
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005440000000000001,
+      "loss": 1.0477,
+      "step": 11400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054396,
+      "loss": 1.0671,
+      "step": 11401
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005439199999999999,
+      "loss": 0.9829,
+      "step": 11402
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005438800000000001,
+      "loss": 1.0565,
+      "step": 11403
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054384,
+      "loss": 1.0359,
+      "step": 11404
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005438,
+      "loss": 1.0617,
+      "step": 11405
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054376,
+      "loss": 1.0918,
+      "step": 11406
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054372,
+      "loss": 1.063,
+      "step": 11407
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054368,
+      "loss": 1.0269,
+      "step": 11408
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005436400000000001,
+      "loss": 1.062,
+      "step": 11409
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005436,
+      "loss": 1.0042,
+      "step": 11410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054356,
+      "loss": 1.0191,
+      "step": 11411
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054352,
+      "loss": 1.0279,
+      "step": 11412
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005434799999999999,
+      "loss": 0.9877,
+      "step": 11413
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005434400000000001,
+      "loss": 1.042,
+      "step": 11414
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005434,
+      "loss": 1.0156,
+      "step": 11415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054336,
+      "loss": 1.0755,
+      "step": 11416
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054332,
+      "loss": 1.0715,
+      "step": 11417
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054328,
+      "loss": 1.0507,
+      "step": 11418
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054324,
+      "loss": 1.129,
+      "step": 11419
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005432,
+      "loss": 1.0525,
+      "step": 11420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054316,
+      "loss": 0.951,
+      "step": 11421
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054312,
+      "loss": 1.0974,
+      "step": 11422
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054308,
+      "loss": 1.0315,
+      "step": 11423
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054304,
+      "loss": 1.041,
+      "step": 11424
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005430000000000001,
+      "loss": 0.9975,
+      "step": 11425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054296,
+      "loss": 1.0047,
+      "step": 11426
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054292,
+      "loss": 1.0691,
+      "step": 11427
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054288,
+      "loss": 0.9549,
+      "step": 11428
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005428400000000001,
+      "loss": 1.0759,
+      "step": 11429
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005428,
+      "loss": 1.1167,
+      "step": 11430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054276,
+      "loss": 1.0516,
+      "step": 11431
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054272,
+      "loss": 1.0297,
+      "step": 11432
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054268,
+      "loss": 1.048,
+      "step": 11433
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054264,
+      "loss": 1.0303,
+      "step": 11434
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005426,
+      "loss": 1.0424,
+      "step": 11435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005425600000000001,
+      "loss": 1.1036,
+      "step": 11436
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054252,
+      "loss": 1.0265,
+      "step": 11437
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054248,
+      "loss": 1.0188,
+      "step": 11438
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054244,
+      "loss": 0.9981,
+      "step": 11439
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005424,
+      "loss": 1.0442,
+      "step": 11440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054236,
+      "loss": 1.0253,
+      "step": 11441
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054232,
+      "loss": 1.0001,
+      "step": 11442
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054228,
+      "loss": 1.1091,
+      "step": 11443
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054224,
+      "loss": 1.0536,
+      "step": 11444
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005422000000000001,
+      "loss": 1.0219,
+      "step": 11445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054216,
+      "loss": 1.1033,
+      "step": 11446
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005421200000000001,
+      "loss": 1.0544,
+      "step": 11447
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054208,
+      "loss": 1.0151,
+      "step": 11448
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005420399999999999,
+      "loss": 1.0658,
+      "step": 11449
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005420000000000001,
+      "loss": 1.0309,
+      "step": 11450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054196,
+      "loss": 1.0282,
+      "step": 11451
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054192,
+      "loss": 1.0467,
+      "step": 11452
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054188,
+      "loss": 1.0683,
+      "step": 11453
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054184,
+      "loss": 1.0338,
+      "step": 11454
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005417999999999999,
+      "loss": 1.0089,
+      "step": 11455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005417600000000001,
+      "loss": 1.0573,
+      "step": 11456
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054172,
+      "loss": 1.0591,
+      "step": 11457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005416800000000001,
+      "loss": 1.0257,
+      "step": 11458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054164,
+      "loss": 0.9826,
+      "step": 11459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005415999999999999,
+      "loss": 1.1167,
+      "step": 11460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005415600000000001,
+      "loss": 1.0812,
+      "step": 11461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054152,
+      "loss": 1.0411,
+      "step": 11462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054148,
+      "loss": 1.0677,
+      "step": 11463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054144,
+      "loss": 1.0604,
+      "step": 11464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005414,
+      "loss": 0.9888,
+      "step": 11465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054136,
+      "loss": 1.0863,
+      "step": 11466
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005413200000000001,
+      "loss": 0.9886,
+      "step": 11467
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054128,
+      "loss": 1.069,
+      "step": 11468
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054124,
+      "loss": 1.0215,
+      "step": 11469
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005412,
+      "loss": 1.0247,
+      "step": 11470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054116,
+      "loss": 1.026,
+      "step": 11471
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005411200000000001,
+      "loss": 1.1472,
+      "step": 11472
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054108,
+      "loss": 1.0895,
+      "step": 11473
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054104,
+      "loss": 1.1236,
+      "step": 11474
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000541,
+      "loss": 1.0134,
+      "step": 11475
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054096,
+      "loss": 1.0573,
+      "step": 11476
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054092,
+      "loss": 1.107,
+      "step": 11477
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054088,
+      "loss": 1.0304,
+      "step": 11478
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054084,
+      "loss": 1.0574,
+      "step": 11479
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005407999999999999,
+      "loss": 1.0207,
+      "step": 11480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054076,
+      "loss": 1.024,
+      "step": 11481
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054072,
+      "loss": 1.0271,
+      "step": 11482
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005406800000000001,
+      "loss": 1.0197,
+      "step": 11483
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054064,
+      "loss": 1.1432,
+      "step": 11484
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005406,
+      "loss": 1.0114,
+      "step": 11485
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054056,
+      "loss": 1.0472,
+      "step": 11486
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005405200000000001,
+      "loss": 1.1057,
+      "step": 11487
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054048,
+      "loss": 1.0682,
+      "step": 11488
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054044,
+      "loss": 1.0197,
+      "step": 11489
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005404,
+      "loss": 1.0383,
+      "step": 11490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005403599999999999,
+      "loss": 0.9773,
+      "step": 11491
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005403200000000001,
+      "loss": 1.0292,
+      "step": 11492
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054028,
+      "loss": 1.0903,
+      "step": 11493
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005402400000000001,
+      "loss": 1.0659,
+      "step": 11494
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005402,
+      "loss": 1.0824,
+      "step": 11495
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054016,
+      "loss": 1.1301,
+      "step": 11496
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054012,
+      "loss": 1.0282,
+      "step": 11497
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054008,
+      "loss": 1.0643,
+      "step": 11498
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054004,
+      "loss": 1.0912,
+      "step": 11499
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00054,
+      "loss": 0.9547,
+      "step": 11500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053996,
+      "loss": 1.07,
+      "step": 11501
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005399199999999999,
+      "loss": 1.1149,
+      "step": 11502
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005398800000000001,
+      "loss": 1.0644,
+      "step": 11503
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053984,
+      "loss": 1.0665,
+      "step": 11504
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005398,
+      "loss": 1.1041,
+      "step": 11505
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053976,
+      "loss": 1.0271,
+      "step": 11506
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005397199999999999,
+      "loss": 1.1361,
+      "step": 11507
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005396800000000001,
+      "loss": 0.994,
+      "step": 11508
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053964,
+      "loss": 1.0571,
+      "step": 11509
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005396,
+      "loss": 1.028,
+      "step": 11510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053956,
+      "loss": 0.9935,
+      "step": 11511
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053952,
+      "loss": 1.0801,
+      "step": 11512
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053948,
+      "loss": 1.084,
+      "step": 11513
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005394400000000001,
+      "loss": 1.0502,
+      "step": 11514
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005394,
+      "loss": 1.0938,
+      "step": 11515
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053936,
+      "loss": 1.0735,
+      "step": 11516
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053932,
+      "loss": 1.0406,
+      "step": 11517
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053928,
+      "loss": 1.0079,
+      "step": 11518
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005392400000000001,
+      "loss": 1.0993,
+      "step": 11519
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005392,
+      "loss": 1.079,
+      "step": 11520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053916,
+      "loss": 1.0701,
+      "step": 11521
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053912,
+      "loss": 1.0752,
+      "step": 11522
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053908,
+      "loss": 1.09,
+      "step": 11523
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053904,
+      "loss": 1.0832,
+      "step": 11524
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005390000000000001,
+      "loss": 1.0476,
+      "step": 11525
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053896,
+      "loss": 1.0774,
+      "step": 11526
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0005389199999999999,
+      "loss": 1.038,
+      "step": 11527
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053888,
+      "loss": 1.1029,
+      "step": 11528
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00053884,
+      "loss": 0.9748,
+      "step": 11529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005388,
+      "loss": 1.0404,
+      "step": 11530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053876,
+      "loss": 1.0596,
+      "step": 11531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053872,
+      "loss": 0.9757,
+      "step": 11532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053868,
+      "loss": 1.0261,
+      "step": 11533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005386400000000001,
+      "loss": 1.0547,
+      "step": 11534
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005386,
+      "loss": 1.1136,
+      "step": 11535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005385600000000001,
+      "loss": 1.0397,
+      "step": 11536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053852,
+      "loss": 0.9791,
+      "step": 11537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005384799999999999,
+      "loss": 1.0747,
+      "step": 11538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005384400000000001,
+      "loss": 1.0937,
+      "step": 11539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005384,
+      "loss": 1.0671,
+      "step": 11540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053836,
+      "loss": 1.0084,
+      "step": 11541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053832,
+      "loss": 1.0435,
+      "step": 11542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053828,
+      "loss": 1.0301,
+      "step": 11543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053824,
+      "loss": 0.9278,
+      "step": 11544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005382000000000001,
+      "loss": 1.0242,
+      "step": 11545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053816,
+      "loss": 1.0333,
+      "step": 11546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053812,
+      "loss": 1.0039,
+      "step": 11547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053808,
+      "loss": 1.0615,
+      "step": 11548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005380399999999999,
+      "loss": 1.0727,
+      "step": 11549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005380000000000001,
+      "loss": 1.0301,
+      "step": 11550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053796,
+      "loss": 1.0647,
+      "step": 11551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053792,
+      "loss": 1.0425,
+      "step": 11552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053788,
+      "loss": 1.0679,
+      "step": 11553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053784,
+      "loss": 0.9851,
+      "step": 11554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005378,
+      "loss": 1.035,
+      "step": 11555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053776,
+      "loss": 1.0389,
+      "step": 11556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053772,
+      "loss": 1.0407,
+      "step": 11557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053768,
+      "loss": 0.9742,
+      "step": 11558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053764,
+      "loss": 1.1367,
+      "step": 11559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005376,
+      "loss": 1.0779,
+      "step": 11560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005375600000000001,
+      "loss": 1.0415,
+      "step": 11561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053752,
+      "loss": 0.9964,
+      "step": 11562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053748,
+      "loss": 1.008,
+      "step": 11563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053744,
+      "loss": 0.989,
+      "step": 11564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005374,
+      "loss": 0.9549,
+      "step": 11565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053736,
+      "loss": 1.0465,
+      "step": 11566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053732,
+      "loss": 1.0476,
+      "step": 11567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053728,
+      "loss": 1.1403,
+      "step": 11568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053724,
+      "loss": 1.0999,
+      "step": 11569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005372,
+      "loss": 1.0777,
+      "step": 11570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053716,
+      "loss": 1.0879,
+      "step": 11571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005371200000000001,
+      "loss": 1.021,
+      "step": 11572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053708,
+      "loss": 1.0057,
+      "step": 11573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053704,
+      "loss": 0.9207,
+      "step": 11574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000537,
+      "loss": 1.0783,
+      "step": 11575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053696,
+      "loss": 1.0191,
+      "step": 11576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053692,
+      "loss": 1.0464,
+      "step": 11577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053688,
+      "loss": 0.9531,
+      "step": 11578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053684,
+      "loss": 1.1072,
+      "step": 11579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005368,
+      "loss": 1.0674,
+      "step": 11580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005367600000000001,
+      "loss": 0.995,
+      "step": 11581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053672,
+      "loss": 1.0634,
+      "step": 11582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005366800000000001,
+      "loss": 1.0214,
+      "step": 11583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053664,
+      "loss": 1.0661,
+      "step": 11584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005365999999999999,
+      "loss": 0.8846,
+      "step": 11585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053656,
+      "loss": 1.0525,
+      "step": 11586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053652,
+      "loss": 1.1049,
+      "step": 11587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053648,
+      "loss": 1.0781,
+      "step": 11588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053644,
+      "loss": 1.0622,
+      "step": 11589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005364,
+      "loss": 1.093,
+      "step": 11590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005363599999999999,
+      "loss": 1.0044,
+      "step": 11591
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005363200000000001,
+      "loss": 1.0141,
+      "step": 11592
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053628,
+      "loss": 1.1421,
+      "step": 11593
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005362400000000001,
+      "loss": 1.0723,
+      "step": 11594
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005362,
+      "loss": 1.0702,
+      "step": 11595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005361599999999999,
+      "loss": 0.9996,
+      "step": 11596
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005361200000000001,
+      "loss": 1.0544,
+      "step": 11597
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053608,
+      "loss": 1.0368,
+      "step": 11598
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053604,
+      "loss": 1.0928,
+      "step": 11599
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000536,
+      "loss": 1.0505,
+      "step": 11600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053596,
+      "loss": 1.1096,
+      "step": 11601
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053592,
+      "loss": 1.0241,
+      "step": 11602
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005358800000000001,
+      "loss": 0.9231,
+      "step": 11603
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053584,
+      "loss": 1.103,
+      "step": 11604
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005358,
+      "loss": 1.0534,
+      "step": 11605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053576,
+      "loss": 1.0227,
+      "step": 11606
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053572,
+      "loss": 1.0269,
+      "step": 11607
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005356800000000001,
+      "loss": 0.9573,
+      "step": 11608
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053564,
+      "loss": 1.0871,
+      "step": 11609
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005356,
+      "loss": 1.0036,
+      "step": 11610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053556,
+      "loss": 1.0827,
+      "step": 11611
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053552,
+      "loss": 1.0514,
+      "step": 11612
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053548,
+      "loss": 1.0642,
+      "step": 11613
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005354400000000001,
+      "loss": 1.0588,
+      "step": 11614
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005354,
+      "loss": 1.06,
+      "step": 11615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005353599999999999,
+      "loss": 1.0807,
+      "step": 11616
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053532,
+      "loss": 1.0173,
+      "step": 11617
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053528,
+      "loss": 1.0349,
+      "step": 11618
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005352400000000001,
+      "loss": 0.9661,
+      "step": 11619
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005352,
+      "loss": 1.0414,
+      "step": 11620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053516,
+      "loss": 1.0389,
+      "step": 11621
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053512,
+      "loss": 1.0383,
+      "step": 11622
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005350800000000001,
+      "loss": 1.1503,
+      "step": 11623
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053504,
+      "loss": 1.0247,
+      "step": 11624
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000535,
+      "loss": 1.0034,
+      "step": 11625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053496,
+      "loss": 0.9834,
+      "step": 11626
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005349199999999999,
+      "loss": 1.0862,
+      "step": 11627
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005348800000000001,
+      "loss": 1.0851,
+      "step": 11628
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053484,
+      "loss": 1.0071,
+      "step": 11629
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005348000000000001,
+      "loss": 1.0974,
+      "step": 11630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053476,
+      "loss": 1.0313,
+      "step": 11631
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053472,
+      "loss": 1.0773,
+      "step": 11632
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053468,
+      "loss": 0.9864,
+      "step": 11633
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053464,
+      "loss": 1.0572,
+      "step": 11634
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005346,
+      "loss": 0.9684,
+      "step": 11635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053456,
+      "loss": 1.0356,
+      "step": 11636
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053452,
+      "loss": 1.0938,
+      "step": 11637
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005344799999999999,
+      "loss": 1.0007,
+      "step": 11638
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005344400000000001,
+      "loss": 1.0409,
+      "step": 11639
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005344,
+      "loss": 1.107,
+      "step": 11640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053436,
+      "loss": 1.0347,
+      "step": 11641
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053432,
+      "loss": 0.9683,
+      "step": 11642
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053428,
+      "loss": 1.0001,
+      "step": 11643
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005342400000000001,
+      "loss": 1.0588,
+      "step": 11644
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005342,
+      "loss": 1.1295,
+      "step": 11645
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053416,
+      "loss": 1.0619,
+      "step": 11646
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053412,
+      "loss": 1.0738,
+      "step": 11647
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053408,
+      "loss": 1.0743,
+      "step": 11648
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053404,
+      "loss": 1.0296,
+      "step": 11649
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005340000000000001,
+      "loss": 1.033,
+      "step": 11650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053396,
+      "loss": 1.1602,
+      "step": 11651
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053392,
+      "loss": 1.0433,
+      "step": 11652
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053388,
+      "loss": 1.0449,
+      "step": 11653
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005338399999999999,
+      "loss": 1.0884,
+      "step": 11654
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005338000000000001,
+      "loss": 1.1098,
+      "step": 11655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053376,
+      "loss": 1.0775,
+      "step": 11656
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053372,
+      "loss": 0.9195,
+      "step": 11657
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053368,
+      "loss": 1.1392,
+      "step": 11658
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053364,
+      "loss": 1.0642,
+      "step": 11659
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005336,
+      "loss": 1.098,
+      "step": 11660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005335600000000001,
+      "loss": 1.0621,
+      "step": 11661
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053352,
+      "loss": 1.1175,
+      "step": 11662
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005334799999999999,
+      "loss": 1.0261,
+      "step": 11663
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053344,
+      "loss": 1.0361,
+      "step": 11664
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005334,
+      "loss": 0.9372,
+      "step": 11665
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053336,
+      "loss": 0.9866,
+      "step": 11666
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053332,
+      "loss": 0.9991,
+      "step": 11667
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053328,
+      "loss": 1.0384,
+      "step": 11668
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053324,
+      "loss": 1.0383,
+      "step": 11669
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005332000000000001,
+      "loss": 0.9661,
+      "step": 11670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053316,
+      "loss": 1.016,
+      "step": 11671
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005331200000000001,
+      "loss": 1.0738,
+      "step": 11672
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053308,
+      "loss": 1.1089,
+      "step": 11673
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005330399999999999,
+      "loss": 0.9445,
+      "step": 11674
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000533,
+      "loss": 1.0061,
+      "step": 11675
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053296,
+      "loss": 1.0234,
+      "step": 11676
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053292,
+      "loss": 1.0829,
+      "step": 11677
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053288,
+      "loss": 1.0187,
+      "step": 11678
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053284,
+      "loss": 0.9513,
+      "step": 11679
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005328,
+      "loss": 1.0467,
+      "step": 11680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005327600000000001,
+      "loss": 1.0228,
+      "step": 11681
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053272,
+      "loss": 1.0559,
+      "step": 11682
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053268,
+      "loss": 1.0813,
+      "step": 11683
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053264,
+      "loss": 1.099,
+      "step": 11684
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005325999999999999,
+      "loss": 0.9815,
+      "step": 11685
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005325600000000001,
+      "loss": 1.0843,
+      "step": 11686
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053252,
+      "loss": 0.9949,
+      "step": 11687
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053248,
+      "loss": 1.0269,
+      "step": 11688
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053244,
+      "loss": 1.0569,
+      "step": 11689
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005324,
+      "loss": 1.0809,
+      "step": 11690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005323600000000001,
+      "loss": 1.025,
+      "step": 11691
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005323200000000001,
+      "loss": 1.0901,
+      "step": 11692
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053228,
+      "loss": 1.0722,
+      "step": 11693
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053224,
+      "loss": 1.0963,
+      "step": 11694
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005322,
+      "loss": 1.0035,
+      "step": 11695
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053216,
+      "loss": 1.1517,
+      "step": 11696
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005321200000000001,
+      "loss": 1.1228,
+      "step": 11697
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053208,
+      "loss": 1.1145,
+      "step": 11698
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053204,
+      "loss": 0.9782,
+      "step": 11699
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000532,
+      "loss": 1.0229,
+      "step": 11700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053196,
+      "loss": 1.014,
+      "step": 11701
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053192,
+      "loss": 1.0293,
+      "step": 11702
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053188,
+      "loss": 1.0626,
+      "step": 11703
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053184,
+      "loss": 1.0348,
+      "step": 11704
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005318,
+      "loss": 1.1327,
+      "step": 11705
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053176,
+      "loss": 1.1046,
+      "step": 11706
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053172,
+      "loss": 1.0662,
+      "step": 11707
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005316800000000001,
+      "loss": 1.0238,
+      "step": 11708
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053164,
+      "loss": 1.0879,
+      "step": 11709
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005316,
+      "loss": 0.9906,
+      "step": 11710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053156,
+      "loss": 1.0727,
+      "step": 11711
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053152,
+      "loss": 1.0498,
+      "step": 11712
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053148,
+      "loss": 1.0462,
+      "step": 11713
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053144,
+      "loss": 1.0409,
+      "step": 11714
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005314,
+      "loss": 1.1144,
+      "step": 11715
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053136,
+      "loss": 1.1217,
+      "step": 11716
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005313200000000001,
+      "loss": 1.0403,
+      "step": 11717
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053128,
+      "loss": 0.9964,
+      "step": 11718
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005312400000000001,
+      "loss": 1.0589,
+      "step": 11719
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005312,
+      "loss": 0.9824,
+      "step": 11720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053116,
+      "loss": 1.0206,
+      "step": 11721
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053112,
+      "loss": 1.0222,
+      "step": 11722
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053108,
+      "loss": 1.0461,
+      "step": 11723
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053104,
+      "loss": 1.0583,
+      "step": 11724
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000531,
+      "loss": 1.0459,
+      "step": 11725
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053096,
+      "loss": 0.9393,
+      "step": 11726
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005309199999999999,
+      "loss": 1.0343,
+      "step": 11727
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005308800000000001,
+      "loss": 1.074,
+      "step": 11728
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053084,
+      "loss": 1.0763,
+      "step": 11729
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005308000000000001,
+      "loss": 1.0888,
+      "step": 11730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053076,
+      "loss": 1.0098,
+      "step": 11731
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005307199999999999,
+      "loss": 0.9963,
+      "step": 11732
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005306800000000001,
+      "loss": 1.0433,
+      "step": 11733
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053064,
+      "loss": 1.02,
+      "step": 11734
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005306,
+      "loss": 1.0076,
+      "step": 11735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053056,
+      "loss": 1.117,
+      "step": 11736
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053052,
+      "loss": 1.0403,
+      "step": 11737
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053048,
+      "loss": 0.9893,
+      "step": 11738
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005304400000000001,
+      "loss": 0.9134,
+      "step": 11739
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005304,
+      "loss": 1.0405,
+      "step": 11740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053036,
+      "loss": 0.9721,
+      "step": 11741
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053032,
+      "loss": 0.9376,
+      "step": 11742
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005302799999999999,
+      "loss": 1.0492,
+      "step": 11743
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005302400000000001,
+      "loss": 1.053,
+      "step": 11744
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005302,
+      "loss": 1.0124,
+      "step": 11745
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053016,
+      "loss": 1.0717,
+      "step": 11746
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053012,
+      "loss": 1.014,
+      "step": 11747
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053008,
+      "loss": 1.0491,
+      "step": 11748
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00053004,
+      "loss": 1.1013,
+      "step": 11749
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005300000000000001,
+      "loss": 1.1101,
+      "step": 11750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052996,
+      "loss": 0.9881,
+      "step": 11751
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005299199999999999,
+      "loss": 1.0341,
+      "step": 11752
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052988,
+      "loss": 1.0417,
+      "step": 11753
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052984,
+      "loss": 1.0969,
+      "step": 11754
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005298000000000001,
+      "loss": 1.0518,
+      "step": 11755
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052976,
+      "loss": 1.0732,
+      "step": 11756
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052972,
+      "loss": 1.0872,
+      "step": 11757
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052968,
+      "loss": 1.0158,
+      "step": 11758
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005296400000000001,
+      "loss": 1.087,
+      "step": 11759
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005296,
+      "loss": 1.1217,
+      "step": 11760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052956,
+      "loss": 1.0659,
+      "step": 11761
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052952,
+      "loss": 1.0629,
+      "step": 11762
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005294799999999999,
+      "loss": 1.0392,
+      "step": 11763
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052944,
+      "loss": 1.071,
+      "step": 11764
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005294,
+      "loss": 1.0402,
+      "step": 11765
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005293600000000001,
+      "loss": 1.0084,
+      "step": 11766
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052932,
+      "loss": 1.0274,
+      "step": 11767
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052928,
+      "loss": 1.0964,
+      "step": 11768
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052924,
+      "loss": 0.9663,
+      "step": 11769
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005292,
+      "loss": 0.9867,
+      "step": 11770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052916,
+      "loss": 1.0315,
+      "step": 11771
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052912,
+      "loss": 1.0627,
+      "step": 11772
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052908,
+      "loss": 1.0582,
+      "step": 11773
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005290399999999999,
+      "loss": 0.9894,
+      "step": 11774
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005290000000000001,
+      "loss": 0.989,
+      "step": 11775
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052896,
+      "loss": 1.0965,
+      "step": 11776
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052892,
+      "loss": 1.0401,
+      "step": 11777
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052888,
+      "loss": 0.9994,
+      "step": 11778
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052884,
+      "loss": 1.0732,
+      "step": 11779
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0005288000000000001,
+      "loss": 1.0024,
+      "step": 11780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052876,
+      "loss": 1.0011,
+      "step": 11781
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052872,
+      "loss": 0.9653,
+      "step": 11782
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00052868,
+      "loss": 0.962,
+      "step": 11783
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052864,
+      "loss": 1.0564,
+      "step": 11784
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005286,
+      "loss": 0.9843,
+      "step": 11785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005285600000000001,
+      "loss": 1.0629,
+      "step": 11786
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052852,
+      "loss": 1.0337,
+      "step": 11787
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052848,
+      "loss": 1.0288,
+      "step": 11788
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052844,
+      "loss": 1.0275,
+      "step": 11789
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005283999999999999,
+      "loss": 1.0029,
+      "step": 11790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005283600000000001,
+      "loss": 1.0515,
+      "step": 11791
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052832,
+      "loss": 1.0869,
+      "step": 11792
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052828,
+      "loss": 1.0025,
+      "step": 11793
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052824,
+      "loss": 1.0639,
+      "step": 11794
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005282,
+      "loss": 1.021,
+      "step": 11795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052816,
+      "loss": 0.9996,
+      "step": 11796
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005281200000000001,
+      "loss": 1.0758,
+      "step": 11797
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052808,
+      "loss": 0.9986,
+      "step": 11798
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005280399999999999,
+      "loss": 1.0166,
+      "step": 11799
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000528,
+      "loss": 1.1404,
+      "step": 11800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052796,
+      "loss": 1.0091,
+      "step": 11801
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052792,
+      "loss": 1.0026,
+      "step": 11802
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052788,
+      "loss": 1.072,
+      "step": 11803
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052784,
+      "loss": 1.0027,
+      "step": 11804
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005278,
+      "loss": 1.0198,
+      "step": 11805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005277600000000001,
+      "loss": 1.0324,
+      "step": 11806
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052772,
+      "loss": 1.1201,
+      "step": 11807
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005276800000000001,
+      "loss": 1.0449,
+      "step": 11808
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052764,
+      "loss": 1.0412,
+      "step": 11809
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005275999999999999,
+      "loss": 1.0821,
+      "step": 11810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052756,
+      "loss": 1.0417,
+      "step": 11811
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052752,
+      "loss": 1.0605,
+      "step": 11812
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052748,
+      "loss": 1.0943,
+      "step": 11813
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052744,
+      "loss": 1.0174,
+      "step": 11814
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005274,
+      "loss": 0.9851,
+      "step": 11815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052736,
+      "loss": 1.057,
+      "step": 11816
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005273200000000001,
+      "loss": 1.0987,
+      "step": 11817
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052728,
+      "loss": 1.0758,
+      "step": 11818
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052724,
+      "loss": 1.0231,
+      "step": 11819
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005272,
+      "loss": 1.1033,
+      "step": 11820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005271599999999999,
+      "loss": 1.0459,
+      "step": 11821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005271200000000001,
+      "loss": 0.9656,
+      "step": 11822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052708,
+      "loss": 1.0605,
+      "step": 11823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052704,
+      "loss": 1.0714,
+      "step": 11824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000527,
+      "loss": 1.0458,
+      "step": 11825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052696,
+      "loss": 0.9863,
+      "step": 11826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005269200000000001,
+      "loss": 0.9847,
+      "step": 11827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005268800000000001,
+      "loss": 1.0412,
+      "step": 11828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052684,
+      "loss": 1.0205,
+      "step": 11829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005268,
+      "loss": 1.0295,
+      "step": 11830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052676,
+      "loss": 1.0681,
+      "step": 11831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005267199999999999,
+      "loss": 1.1121,
+      "step": 11832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005266800000000001,
+      "loss": 1.0344,
+      "step": 11833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052664,
+      "loss": 1.0527,
+      "step": 11834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005266,
+      "loss": 1.0648,
+      "step": 11835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052656,
+      "loss": 1.0225,
+      "step": 11836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052652,
+      "loss": 0.9427,
+      "step": 11837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052648,
+      "loss": 1.1293,
+      "step": 11838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052644,
+      "loss": 1.0283,
+      "step": 11839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005264,
+      "loss": 1.0244,
+      "step": 11840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052636,
+      "loss": 1.126,
+      "step": 11841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052632,
+      "loss": 1.0034,
+      "step": 11842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052628,
+      "loss": 1.0667,
+      "step": 11843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005262400000000001,
+      "loss": 0.9808,
+      "step": 11844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005262,
+      "loss": 1.0628,
+      "step": 11845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052616,
+      "loss": 1.0145,
+      "step": 11846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052612,
+      "loss": 1.0656,
+      "step": 11847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052608,
+      "loss": 1.0962,
+      "step": 11848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052604,
+      "loss": 0.9839,
+      "step": 11849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000526,
+      "loss": 1.0871,
+      "step": 11850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052596,
+      "loss": 1.0465,
+      "step": 11851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052592,
+      "loss": 1.0397,
+      "step": 11852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052588,
+      "loss": 1.038,
+      "step": 11853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052584,
+      "loss": 1.0455,
+      "step": 11854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005258000000000001,
+      "loss": 1.0303,
+      "step": 11855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052576,
+      "loss": 1.0942,
+      "step": 11856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052572,
+      "loss": 1.0941,
+      "step": 11857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052568,
+      "loss": 1.0255,
+      "step": 11858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052564,
+      "loss": 1.0582,
+      "step": 11859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005256,
+      "loss": 1.0782,
+      "step": 11860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052556,
+      "loss": 1.0537,
+      "step": 11861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052552,
+      "loss": 1.0795,
+      "step": 11862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005254799999999999,
+      "loss": 1.0984,
+      "step": 11863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005254400000000001,
+      "loss": 1.0258,
+      "step": 11864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005254,
+      "loss": 0.9735,
+      "step": 11865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005253600000000001,
+      "loss": 1.0148,
+      "step": 11866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052532,
+      "loss": 0.9787,
+      "step": 11867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005252799999999999,
+      "loss": 0.9653,
+      "step": 11868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005252400000000001,
+      "loss": 1.0625,
+      "step": 11869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005252,
+      "loss": 0.9819,
+      "step": 11870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052516,
+      "loss": 1.0039,
+      "step": 11871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052512,
+      "loss": 1.0266,
+      "step": 11872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052508,
+      "loss": 1.0859,
+      "step": 11873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052504,
+      "loss": 0.9464,
+      "step": 11874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005250000000000001,
+      "loss": 0.9742,
+      "step": 11875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052496,
+      "loss": 0.9973,
+      "step": 11876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005249200000000001,
+      "loss": 1.0769,
+      "step": 11877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052488,
+      "loss": 1.0315,
+      "step": 11878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005248399999999999,
+      "loss": 1.0069,
+      "step": 11879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005248000000000001,
+      "loss": 1.0308,
+      "step": 11880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052476,
+      "loss": 0.9147,
+      "step": 11881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052472,
+      "loss": 1.0249,
+      "step": 11882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052468,
+      "loss": 1.0298,
+      "step": 11883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052464,
+      "loss": 1.0488,
+      "step": 11884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005246,
+      "loss": 1.1019,
+      "step": 11885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005245600000000001,
+      "loss": 1.1177,
+      "step": 11886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052452,
+      "loss": 0.9509,
+      "step": 11887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005244799999999999,
+      "loss": 1.009,
+      "step": 11888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052444,
+      "loss": 1.0948,
+      "step": 11889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005244,
+      "loss": 1.1049,
+      "step": 11890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005243600000000001,
+      "loss": 0.9785,
+      "step": 11891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052432,
+      "loss": 0.9736,
+      "step": 11892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052428,
+      "loss": 1.0333,
+      "step": 11893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052424,
+      "loss": 0.9795,
+      "step": 11894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005242000000000001,
+      "loss": 1.0168,
+      "step": 11895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052416,
+      "loss": 1.1004,
+      "step": 11896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052412,
+      "loss": 1.1388,
+      "step": 11897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052408,
+      "loss": 1.0574,
+      "step": 11898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005240399999999999,
+      "loss": 1.0666,
+      "step": 11899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000524,
+      "loss": 1.0564,
+      "step": 11900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052396,
+      "loss": 1.0416,
+      "step": 11901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005239200000000001,
+      "loss": 1.0893,
+      "step": 11902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052388,
+      "loss": 1.0237,
+      "step": 11903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052384,
+      "loss": 1.0768,
+      "step": 11904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005238,
+      "loss": 1.1047,
+      "step": 11905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005237600000000001,
+      "loss": 1.0693,
+      "step": 11906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052372,
+      "loss": 1.057,
+      "step": 11907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052368,
+      "loss": 1.0354,
+      "step": 11908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052364,
+      "loss": 1.0927,
+      "step": 11909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005235999999999999,
+      "loss": 1.0226,
+      "step": 11910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005235600000000001,
+      "loss": 1.0878,
+      "step": 11911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052352,
+      "loss": 0.9946,
+      "step": 11912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052348,
+      "loss": 0.9306,
+      "step": 11913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052344,
+      "loss": 1.0529,
+      "step": 11914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005234,
+      "loss": 1.0536,
+      "step": 11915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005233600000000001,
+      "loss": 1.0966,
+      "step": 11916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052332,
+      "loss": 1.0096,
+      "step": 11917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052328,
+      "loss": 1.0463,
+      "step": 11918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052324,
+      "loss": 1.0454,
+      "step": 11919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005232,
+      "loss": 0.9415,
+      "step": 11920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005231599999999999,
+      "loss": 1.0092,
+      "step": 11921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005231200000000001,
+      "loss": 0.9768,
+      "step": 11922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052308,
+      "loss": 1.0518,
+      "step": 11923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052304,
+      "loss": 1.0,
+      "step": 11924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000523,
+      "loss": 1.0063,
+      "step": 11925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005229599999999999,
+      "loss": 1.0707,
+      "step": 11926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005229200000000001,
+      "loss": 1.0381,
+      "step": 11927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052288,
+      "loss": 1.0371,
+      "step": 11928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052284,
+      "loss": 1.1098,
+      "step": 11929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005228,
+      "loss": 1.034,
+      "step": 11930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052276,
+      "loss": 1.0024,
+      "step": 11931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052272,
+      "loss": 1.0499,
+      "step": 11932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005226800000000001,
+      "loss": 1.0614,
+      "step": 11933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052264,
+      "loss": 0.9727,
+      "step": 11934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005226,
+      "loss": 1.0231,
+      "step": 11935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052256,
+      "loss": 1.1021,
+      "step": 11936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052252,
+      "loss": 1.1036,
+      "step": 11937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005224800000000001,
+      "loss": 1.0742,
+      "step": 11938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052244,
+      "loss": 1.049,
+      "step": 11939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005224,
+      "loss": 1.0488,
+      "step": 11940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052236,
+      "loss": 0.9691,
+      "step": 11941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052232,
+      "loss": 1.0558,
+      "step": 11942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052228,
+      "loss": 1.0671,
+      "step": 11943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005222400000000001,
+      "loss": 1.0067,
+      "step": 11944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005222,
+      "loss": 1.0552,
+      "step": 11945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005221599999999999,
+      "loss": 1.0084,
+      "step": 11946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052212,
+      "loss": 1.1504,
+      "step": 11947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052208,
+      "loss": 0.9439,
+      "step": 11948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052204,
+      "loss": 1.1261,
+      "step": 11949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000522,
+      "loss": 1.1219,
+      "step": 11950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052196,
+      "loss": 1.037,
+      "step": 11951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052192,
+      "loss": 1.0046,
+      "step": 11952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005218800000000001,
+      "loss": 1.036,
+      "step": 11953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052184,
+      "loss": 0.9843,
+      "step": 11954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005218000000000001,
+      "loss": 1.0751,
+      "step": 11955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052176,
+      "loss": 1.0405,
+      "step": 11956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005217199999999999,
+      "loss": 1.0613,
+      "step": 11957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005216800000000001,
+      "loss": 0.9972,
+      "step": 11958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052164,
+      "loss": 1.0389,
+      "step": 11959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005216,
+      "loss": 0.9888,
+      "step": 11960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052156,
+      "loss": 1.1108,
+      "step": 11961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052152,
+      "loss": 1.1635,
+      "step": 11962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005214800000000001,
+      "loss": 1.1414,
+      "step": 11963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005214400000000001,
+      "loss": 1.0865,
+      "step": 11964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005214,
+      "loss": 1.1325,
+      "step": 11965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052136,
+      "loss": 1.0174,
+      "step": 11966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052132,
+      "loss": 1.0631,
+      "step": 11967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005212799999999999,
+      "loss": 1.0169,
+      "step": 11968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005212400000000001,
+      "loss": 1.0893,
+      "step": 11969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005212,
+      "loss": 1.1401,
+      "step": 11970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052116,
+      "loss": 0.9361,
+      "step": 11971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052112,
+      "loss": 0.9927,
+      "step": 11972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052108,
+      "loss": 1.1185,
+      "step": 11973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052104,
+      "loss": 1.1123,
+      "step": 11974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000521,
+      "loss": 1.0708,
+      "step": 11975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052096,
+      "loss": 1.083,
+      "step": 11976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052092,
+      "loss": 0.9743,
+      "step": 11977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052088,
+      "loss": 1.0194,
+      "step": 11978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052084,
+      "loss": 1.0356,
+      "step": 11979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005208000000000001,
+      "loss": 1.0032,
+      "step": 11980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052076,
+      "loss": 1.0479,
+      "step": 11981
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052072,
+      "loss": 1.1075,
+      "step": 11982
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052068,
+      "loss": 1.0369,
+      "step": 11983
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052064,
+      "loss": 1.1408,
+      "step": 11984
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005206,
+      "loss": 1.0985,
+      "step": 11985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052056,
+      "loss": 0.9981,
+      "step": 11986
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052052,
+      "loss": 0.967,
+      "step": 11987
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052048,
+      "loss": 0.9843,
+      "step": 11988
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052044,
+      "loss": 0.9716,
+      "step": 11989
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005204,
+      "loss": 1.008,
+      "step": 11990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005203600000000001,
+      "loss": 1.0,
+      "step": 11991
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052032,
+      "loss": 0.8948,
+      "step": 11992
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052028,
+      "loss": 1.0208,
+      "step": 11993
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052024,
+      "loss": 0.9866,
+      "step": 11994
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005202,
+      "loss": 1.0319,
+      "step": 11995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052016,
+      "loss": 0.9414,
+      "step": 11996
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052012,
+      "loss": 1.0143,
+      "step": 11997
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00052008,
+      "loss": 1.0383,
+      "step": 11998
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005200399999999999,
+      "loss": 1.027,
+      "step": 11999
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005200000000000001,
+      "loss": 1.0987,
+      "step": 12000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051996,
+      "loss": 1.0121,
+      "step": 12001
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005199200000000001,
+      "loss": 1.1073,
+      "step": 12002
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051988,
+      "loss": 0.9983,
+      "step": 12003
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005198399999999999,
+      "loss": 1.0017,
+      "step": 12004
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005198000000000001,
+      "loss": 1.0827,
+      "step": 12005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051976,
+      "loss": 1.0384,
+      "step": 12006
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051972,
+      "loss": 0.9593,
+      "step": 12007
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051968,
+      "loss": 1.0361,
+      "step": 12008
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051964,
+      "loss": 1.0692,
+      "step": 12009
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005195999999999999,
+      "loss": 1.0407,
+      "step": 12010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005195600000000001,
+      "loss": 0.9928,
+      "step": 12011
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051952,
+      "loss": 1.1084,
+      "step": 12012
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005194800000000001,
+      "loss": 1.1272,
+      "step": 12013
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051944,
+      "loss": 1.0886,
+      "step": 12014
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005193999999999999,
+      "loss": 1.0334,
+      "step": 12015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005193600000000001,
+      "loss": 1.0372,
+      "step": 12016
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051932,
+      "loss": 1.0741,
+      "step": 12017
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051928,
+      "loss": 1.0579,
+      "step": 12018
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051924,
+      "loss": 1.1013,
+      "step": 12019
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005192,
+      "loss": 1.0038,
+      "step": 12020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051916,
+      "loss": 0.9826,
+      "step": 12021
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005191200000000001,
+      "loss": 0.9574,
+      "step": 12022
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051908,
+      "loss": 1.0041,
+      "step": 12023
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005190399999999999,
+      "loss": 1.0594,
+      "step": 12024
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000519,
+      "loss": 1.0561,
+      "step": 12025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051896,
+      "loss": 1.0888,
+      "step": 12026
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005189200000000001,
+      "loss": 0.9262,
+      "step": 12027
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051888,
+      "loss": 0.9537,
+      "step": 12028
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051884,
+      "loss": 1.026,
+      "step": 12029
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005188,
+      "loss": 1.0228,
+      "step": 12030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005187600000000001,
+      "loss": 1.0223,
+      "step": 12031
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051872,
+      "loss": 1.0458,
+      "step": 12032
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051868,
+      "loss": 1.0224,
+      "step": 12033
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051864,
+      "loss": 1.0679,
+      "step": 12034
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0005185999999999999,
+      "loss": 1.0948,
+      "step": 12035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00051856,
+      "loss": 0.9643,
+      "step": 12036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051852,
+      "loss": 0.9922,
+      "step": 12037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005184800000000001,
+      "loss": 0.9723,
+      "step": 12038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051844,
+      "loss": 1.0764,
+      "step": 12039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005184,
+      "loss": 1.0294,
+      "step": 12040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051836,
+      "loss": 0.9802,
+      "step": 12041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005183200000000001,
+      "loss": 1.0966,
+      "step": 12042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051828,
+      "loss": 1.0685,
+      "step": 12043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051824,
+      "loss": 1.0467,
+      "step": 12044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005182,
+      "loss": 1.0996,
+      "step": 12045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005181599999999999,
+      "loss": 1.0509,
+      "step": 12046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005181200000000001,
+      "loss": 1.0803,
+      "step": 12047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051808,
+      "loss": 1.0172,
+      "step": 12048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051804,
+      "loss": 1.1143,
+      "step": 12049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000518,
+      "loss": 0.9513,
+      "step": 12050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051796,
+      "loss": 1.0002,
+      "step": 12051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005179200000000001,
+      "loss": 1.1383,
+      "step": 12052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051788,
+      "loss": 1.0609,
+      "step": 12053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051784,
+      "loss": 1.0087,
+      "step": 12054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005178,
+      "loss": 1.051,
+      "step": 12055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051776,
+      "loss": 1.0744,
+      "step": 12056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005177199999999999,
+      "loss": 1.017,
+      "step": 12057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005176800000000001,
+      "loss": 1.0239,
+      "step": 12058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051764,
+      "loss": 1.0337,
+      "step": 12059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005176,
+      "loss": 1.0919,
+      "step": 12060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051756,
+      "loss": 1.0231,
+      "step": 12061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005175199999999999,
+      "loss": 1.0006,
+      "step": 12062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005174800000000001,
+      "loss": 1.0011,
+      "step": 12063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051744,
+      "loss": 1.0116,
+      "step": 12064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005174,
+      "loss": 1.135,
+      "step": 12065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051736,
+      "loss": 1.0461,
+      "step": 12066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051732,
+      "loss": 1.0709,
+      "step": 12067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051728,
+      "loss": 1.0477,
+      "step": 12068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005172400000000001,
+      "loss": 0.9593,
+      "step": 12069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005172,
+      "loss": 1.0826,
+      "step": 12070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051716,
+      "loss": 1.0848,
+      "step": 12071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051712,
+      "loss": 1.003,
+      "step": 12072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051708,
+      "loss": 1.1254,
+      "step": 12073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005170400000000001,
+      "loss": 1.0693,
+      "step": 12074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000517,
+      "loss": 1.0314,
+      "step": 12075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051696,
+      "loss": 0.9692,
+      "step": 12076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051692,
+      "loss": 0.9819,
+      "step": 12077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051688,
+      "loss": 1.0076,
+      "step": 12078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051684,
+      "loss": 1.0275,
+      "step": 12079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005168000000000001,
+      "loss": 1.0033,
+      "step": 12080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051676,
+      "loss": 1.0196,
+      "step": 12081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005167199999999999,
+      "loss": 1.0616,
+      "step": 12082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051668,
+      "loss": 1.054,
+      "step": 12083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051664,
+      "loss": 0.965,
+      "step": 12084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005166,
+      "loss": 1.051,
+      "step": 12085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051656,
+      "loss": 1.0275,
+      "step": 12086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051652,
+      "loss": 0.939,
+      "step": 12087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051648,
+      "loss": 0.9879,
+      "step": 12088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005164400000000001,
+      "loss": 1.0292,
+      "step": 12089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005164,
+      "loss": 1.0972,
+      "step": 12090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005163600000000001,
+      "loss": 1.049,
+      "step": 12091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051632,
+      "loss": 1.081,
+      "step": 12092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005162799999999999,
+      "loss": 1.0829,
+      "step": 12093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005162400000000001,
+      "loss": 1.1184,
+      "step": 12094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005162,
+      "loss": 1.1015,
+      "step": 12095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051616,
+      "loss": 1.0606,
+      "step": 12096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051612,
+      "loss": 1.0587,
+      "step": 12097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051608,
+      "loss": 1.1265,
+      "step": 12098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051604,
+      "loss": 1.041,
+      "step": 12099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005160000000000001,
+      "loss": 0.9874,
+      "step": 12100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051596,
+      "loss": 1.0666,
+      "step": 12101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051592,
+      "loss": 1.0564,
+      "step": 12102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051588,
+      "loss": 0.9668,
+      "step": 12103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005158399999999999,
+      "loss": 1.0775,
+      "step": 12104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005158000000000001,
+      "loss": 1.1056,
+      "step": 12105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051576,
+      "loss": 0.9823,
+      "step": 12106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051572,
+      "loss": 0.9856,
+      "step": 12107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051568,
+      "loss": 1.0303,
+      "step": 12108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051564,
+      "loss": 1.0401,
+      "step": 12109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005156,
+      "loss": 1.0993,
+      "step": 12110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051556,
+      "loss": 0.9978,
+      "step": 12111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051552,
+      "loss": 1.0138,
+      "step": 12112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051548,
+      "loss": 1.0735,
+      "step": 12113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051544,
+      "loss": 1.0241,
+      "step": 12114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005154,
+      "loss": 1.0372,
+      "step": 12115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005153600000000001,
+      "loss": 0.989,
+      "step": 12116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051532,
+      "loss": 0.9999,
+      "step": 12117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051528,
+      "loss": 0.9882,
+      "step": 12118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051524,
+      "loss": 0.939,
+      "step": 12119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005152000000000001,
+      "loss": 1.0369,
+      "step": 12120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051516,
+      "loss": 1.0446,
+      "step": 12121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051512,
+      "loss": 1.0307,
+      "step": 12122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051508,
+      "loss": 1.0265,
+      "step": 12123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051504,
+      "loss": 0.9981,
+      "step": 12124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000515,
+      "loss": 1.0897,
+      "step": 12125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051496,
+      "loss": 1.0399,
+      "step": 12126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005149200000000001,
+      "loss": 1.0433,
+      "step": 12127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051488,
+      "loss": 0.9985,
+      "step": 12128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051484,
+      "loss": 0.9668,
+      "step": 12129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005148,
+      "loss": 0.9593,
+      "step": 12130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051476,
+      "loss": 1.0332,
+      "step": 12131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051472,
+      "loss": 0.9908,
+      "step": 12132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051468,
+      "loss": 1.0792,
+      "step": 12133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051464,
+      "loss": 1.0831,
+      "step": 12134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005145999999999999,
+      "loss": 0.9411,
+      "step": 12135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005145600000000001,
+      "loss": 1.0088,
+      "step": 12136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051452,
+      "loss": 0.9914,
+      "step": 12137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005144800000000001,
+      "loss": 1.0656,
+      "step": 12138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051444,
+      "loss": 1.0474,
+      "step": 12139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005143999999999999,
+      "loss": 1.0393,
+      "step": 12140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005143600000000001,
+      "loss": 1.0875,
+      "step": 12141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051432,
+      "loss": 1.0829,
+      "step": 12142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051428,
+      "loss": 1.0678,
+      "step": 12143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051424,
+      "loss": 1.0323,
+      "step": 12144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005142,
+      "loss": 1.0438,
+      "step": 12145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005141599999999999,
+      "loss": 0.974,
+      "step": 12146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005141200000000001,
+      "loss": 1.0205,
+      "step": 12147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051408,
+      "loss": 1.0246,
+      "step": 12148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005140400000000001,
+      "loss": 1.0238,
+      "step": 12149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000514,
+      "loss": 0.9981,
+      "step": 12150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005139599999999999,
+      "loss": 1.0198,
+      "step": 12151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005139200000000001,
+      "loss": 0.9427,
+      "step": 12152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051388,
+      "loss": 1.0054,
+      "step": 12153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051384,
+      "loss": 1.0442,
+      "step": 12154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005138,
+      "loss": 1.0778,
+      "step": 12155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051376,
+      "loss": 1.0132,
+      "step": 12156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051372,
+      "loss": 1.0172,
+      "step": 12157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005136800000000001,
+      "loss": 1.059,
+      "step": 12158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051364,
+      "loss": 1.011,
+      "step": 12159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005135999999999999,
+      "loss": 1.0448,
+      "step": 12160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051356,
+      "loss": 1.0232,
+      "step": 12161
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051352,
+      "loss": 1.017,
+      "step": 12162
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005134800000000001,
+      "loss": 1.0192,
+      "step": 12163
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051344,
+      "loss": 1.0547,
+      "step": 12164
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005134,
+      "loss": 1.095,
+      "step": 12165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051336,
+      "loss": 1.0531,
+      "step": 12166
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051332,
+      "loss": 1.0591,
+      "step": 12167
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051328,
+      "loss": 1.0429,
+      "step": 12168
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005132400000000001,
+      "loss": 1.1065,
+      "step": 12169
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005132,
+      "loss": 1.0707,
+      "step": 12170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005131599999999999,
+      "loss": 1.0051,
+      "step": 12171
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051312,
+      "loss": 1.0537,
+      "step": 12172
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051308,
+      "loss": 0.9992,
+      "step": 12173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005130400000000001,
+      "loss": 0.9276,
+      "step": 12174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000513,
+      "loss": 1.0621,
+      "step": 12175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051296,
+      "loss": 0.9797,
+      "step": 12176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051292,
+      "loss": 1.0728,
+      "step": 12177
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005128800000000001,
+      "loss": 1.0135,
+      "step": 12178
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051284,
+      "loss": 1.0721,
+      "step": 12179
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005128,
+      "loss": 1.0099,
+      "step": 12180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051276,
+      "loss": 0.9775,
+      "step": 12181
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005127199999999999,
+      "loss": 1.0157,
+      "step": 12182
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005126800000000001,
+      "loss": 1.0304,
+      "step": 12183
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051264,
+      "loss": 1.039,
+      "step": 12184
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005126,
+      "loss": 1.122,
+      "step": 12185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051256,
+      "loss": 1.074,
+      "step": 12186
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051252,
+      "loss": 1.0201,
+      "step": 12187
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051248,
+      "loss": 1.0299,
+      "step": 12188
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051244,
+      "loss": 1.0576,
+      "step": 12189
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005124,
+      "loss": 1.0228,
+      "step": 12190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051236,
+      "loss": 1.0387,
+      "step": 12191
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051232,
+      "loss": 1.0337,
+      "step": 12192
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005122799999999999,
+      "loss": 0.9994,
+      "step": 12193
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005122400000000001,
+      "loss": 1.0716,
+      "step": 12194
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005122,
+      "loss": 1.0402,
+      "step": 12195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051216,
+      "loss": 1.0384,
+      "step": 12196
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051212,
+      "loss": 1.0565,
+      "step": 12197
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051208,
+      "loss": 1.0616,
+      "step": 12198
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005120400000000001,
+      "loss": 1.0771,
+      "step": 12199
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000512,
+      "loss": 1.032,
+      "step": 12200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051196,
+      "loss": 0.983,
+      "step": 12201
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051192,
+      "loss": 1.032,
+      "step": 12202
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051188,
+      "loss": 0.997,
+      "step": 12203
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051184,
+      "loss": 1.1086,
+      "step": 12204
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005118000000000001,
+      "loss": 1.0642,
+      "step": 12205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051176,
+      "loss": 1.1078,
+      "step": 12206
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051172,
+      "loss": 1.1051,
+      "step": 12207
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051168,
+      "loss": 1.1026,
+      "step": 12208
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051164,
+      "loss": 1.0115,
+      "step": 12209
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005116000000000001,
+      "loss": 1.0986,
+      "step": 12210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051156,
+      "loss": 0.965,
+      "step": 12211
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051152,
+      "loss": 1.1151,
+      "step": 12212
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051148,
+      "loss": 1.0044,
+      "step": 12213
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051144,
+      "loss": 0.9863,
+      "step": 12214
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005114,
+      "loss": 1.0106,
+      "step": 12215
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005113600000000001,
+      "loss": 1.1774,
+      "step": 12216
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051132,
+      "loss": 1.0242,
+      "step": 12217
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005112799999999999,
+      "loss": 1.0637,
+      "step": 12218
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051124,
+      "loss": 1.134,
+      "step": 12219
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005112,
+      "loss": 1.015,
+      "step": 12220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051116,
+      "loss": 1.0542,
+      "step": 12221
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051112,
+      "loss": 0.9926,
+      "step": 12222
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051108,
+      "loss": 1.1099,
+      "step": 12223
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051104,
+      "loss": 1.0757,
+      "step": 12224
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005110000000000001,
+      "loss": 1.0563,
+      "step": 12225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051096,
+      "loss": 1.0995,
+      "step": 12226
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005109200000000001,
+      "loss": 1.0034,
+      "step": 12227
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051088,
+      "loss": 1.164,
+      "step": 12228
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005108399999999999,
+      "loss": 1.0095,
+      "step": 12229
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005108000000000001,
+      "loss": 0.953,
+      "step": 12230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051076,
+      "loss": 1.0627,
+      "step": 12231
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051072,
+      "loss": 1.046,
+      "step": 12232
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051068,
+      "loss": 1.0526,
+      "step": 12233
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051064,
+      "loss": 1.0705,
+      "step": 12234
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005106,
+      "loss": 1.016,
+      "step": 12235
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005105600000000001,
+      "loss": 1.0069,
+      "step": 12236
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051052,
+      "loss": 1.0361,
+      "step": 12237
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051048,
+      "loss": 1.0196,
+      "step": 12238
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051044,
+      "loss": 1.0087,
+      "step": 12239
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005103999999999999,
+      "loss": 1.0231,
+      "step": 12240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005103600000000001,
+      "loss": 1.0476,
+      "step": 12241
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051032,
+      "loss": 1.1236,
+      "step": 12242
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051028,
+      "loss": 0.9901,
+      "step": 12243
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051024,
+      "loss": 0.9769,
+      "step": 12244
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005102,
+      "loss": 1.0217,
+      "step": 12245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051016,
+      "loss": 0.9819,
+      "step": 12246
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051012,
+      "loss": 0.9862,
+      "step": 12247
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051008,
+      "loss": 1.0248,
+      "step": 12248
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051004,
+      "loss": 1.0058,
+      "step": 12249
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00051,
+      "loss": 0.9683,
+      "step": 12250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050996,
+      "loss": 1.162,
+      "step": 12251
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005099200000000001,
+      "loss": 0.9969,
+      "step": 12252
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050988,
+      "loss": 1.0438,
+      "step": 12253
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050984,
+      "loss": 1.0963,
+      "step": 12254
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005098,
+      "loss": 0.9766,
+      "step": 12255
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050976,
+      "loss": 1.078,
+      "step": 12256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050972,
+      "loss": 0.9877,
+      "step": 12257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050968,
+      "loss": 0.9916,
+      "step": 12258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050964,
+      "loss": 1.0218,
+      "step": 12259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005096,
+      "loss": 1.0883,
+      "step": 12260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050956,
+      "loss": 0.9739,
+      "step": 12261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050952,
+      "loss": 1.0567,
+      "step": 12262
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005094800000000001,
+      "loss": 1.0204,
+      "step": 12263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050944,
+      "loss": 0.9879,
+      "step": 12264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005094,
+      "loss": 1.1813,
+      "step": 12265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050936,
+      "loss": 1.0643,
+      "step": 12266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050932,
+      "loss": 1.0441,
+      "step": 12267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050928,
+      "loss": 0.9872,
+      "step": 12268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050924,
+      "loss": 0.9528,
+      "step": 12269
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005092,
+      "loss": 0.9939,
+      "step": 12270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005091599999999999,
+      "loss": 1.0701,
+      "step": 12271
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005091200000000001,
+      "loss": 1.0397,
+      "step": 12272
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050908,
+      "loss": 1.1313,
+      "step": 12273
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005090400000000001,
+      "loss": 0.9736,
+      "step": 12274
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000509,
+      "loss": 1.0152,
+      "step": 12275
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005089599999999999,
+      "loss": 1.0131,
+      "step": 12276
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050892,
+      "loss": 1.0397,
+      "step": 12277
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050888,
+      "loss": 1.0271,
+      "step": 12278
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050884,
+      "loss": 1.004,
+      "step": 12279
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005088,
+      "loss": 1.0112,
+      "step": 12280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050876,
+      "loss": 1.0557,
+      "step": 12281
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005087199999999999,
+      "loss": 1.0314,
+      "step": 12282
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005086800000000001,
+      "loss": 1.0339,
+      "step": 12283
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050864,
+      "loss": 1.0116,
+      "step": 12284
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005086000000000001,
+      "loss": 1.0566,
+      "step": 12285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050856,
+      "loss": 0.9856,
+      "step": 12286
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005085199999999999,
+      "loss": 1.0228,
+      "step": 12287
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005084800000000001,
+      "loss": 1.0295,
+      "step": 12288
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00050844,
+      "loss": 1.0822,
+      "step": 12289
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0005084,
+      "loss": 1.0708,
+      "step": 12290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050836,
+      "loss": 0.9601,
+      "step": 12291
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050832,
+      "loss": 1.0312,
+      "step": 12292
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050828,
+      "loss": 0.9646,
+      "step": 12293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005082400000000001,
+      "loss": 0.9956,
+      "step": 12294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005082,
+      "loss": 0.9679,
+      "step": 12295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005081599999999999,
+      "loss": 1.0535,
+      "step": 12296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050812,
+      "loss": 1.0711,
+      "step": 12297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050808,
+      "loss": 1.0515,
+      "step": 12298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005080400000000001,
+      "loss": 1.0249,
+      "step": 12299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000508,
+      "loss": 0.9797,
+      "step": 12300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050796,
+      "loss": 1.0689,
+      "step": 12301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050792,
+      "loss": 1.0023,
+      "step": 12302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050788,
+      "loss": 1.0436,
+      "step": 12303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050784,
+      "loss": 0.9854,
+      "step": 12304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005078000000000001,
+      "loss": 0.9239,
+      "step": 12305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050776,
+      "loss": 0.9952,
+      "step": 12306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005077199999999999,
+      "loss": 1.0539,
+      "step": 12307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050768,
+      "loss": 1.0681,
+      "step": 12308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050764,
+      "loss": 1.0464,
+      "step": 12309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005076000000000001,
+      "loss": 1.0294,
+      "step": 12310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050756,
+      "loss": 1.0121,
+      "step": 12311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050752,
+      "loss": 1.0389,
+      "step": 12312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050748,
+      "loss": 1.0401,
+      "step": 12313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005074400000000001,
+      "loss": 1.0615,
+      "step": 12314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005074,
+      "loss": 0.9881,
+      "step": 12315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050736,
+      "loss": 0.9618,
+      "step": 12316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050732,
+      "loss": 1.0537,
+      "step": 12317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005072799999999999,
+      "loss": 0.9803,
+      "step": 12318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005072400000000001,
+      "loss": 1.0336,
+      "step": 12319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005072,
+      "loss": 1.0153,
+      "step": 12320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005071600000000001,
+      "loss": 1.0794,
+      "step": 12321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050712,
+      "loss": 0.9866,
+      "step": 12322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050708,
+      "loss": 1.0565,
+      "step": 12323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050704,
+      "loss": 1.0041,
+      "step": 12324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000507,
+      "loss": 1.0174,
+      "step": 12325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050696,
+      "loss": 1.0032,
+      "step": 12326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050692,
+      "loss": 1.0463,
+      "step": 12327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050688,
+      "loss": 1.103,
+      "step": 12328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005068399999999999,
+      "loss": 1.0484,
+      "step": 12329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005068000000000001,
+      "loss": 1.0707,
+      "step": 12330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050676,
+      "loss": 1.0013,
+      "step": 12331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050672,
+      "loss": 1.0552,
+      "step": 12332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050668,
+      "loss": 1.0508,
+      "step": 12333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050664,
+      "loss": 1.0861,
+      "step": 12334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005066000000000001,
+      "loss": 0.9955,
+      "step": 12335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050656,
+      "loss": 0.9948,
+      "step": 12336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050652,
+      "loss": 1.0369,
+      "step": 12337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050648,
+      "loss": 1.0556,
+      "step": 12338
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050644,
+      "loss": 1.014,
+      "step": 12339
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005064,
+      "loss": 1.0795,
+      "step": 12340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005063600000000001,
+      "loss": 1.0303,
+      "step": 12341
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050632,
+      "loss": 1.0409,
+      "step": 12342
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050628,
+      "loss": 1.0765,
+      "step": 12343
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050624,
+      "loss": 1.1204,
+      "step": 12344
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005061999999999999,
+      "loss": 1.032,
+      "step": 12345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005061600000000001,
+      "loss": 0.9959,
+      "step": 12346
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050612,
+      "loss": 0.9994,
+      "step": 12347
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050608,
+      "loss": 1.0472,
+      "step": 12348
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050604,
+      "loss": 1.0601,
+      "step": 12349
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000506,
+      "loss": 1.0414,
+      "step": 12350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050596,
+      "loss": 1.067,
+      "step": 12351
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005059200000000001,
+      "loss": 1.0409,
+      "step": 12352
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050588,
+      "loss": 1.0351,
+      "step": 12353
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005058399999999999,
+      "loss": 1.0163,
+      "step": 12354
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005058,
+      "loss": 1.1003,
+      "step": 12355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050576,
+      "loss": 1.1508,
+      "step": 12356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050572,
+      "loss": 1.0011,
+      "step": 12357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050568,
+      "loss": 0.9887,
+      "step": 12358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050564,
+      "loss": 0.9787,
+      "step": 12359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005056,
+      "loss": 1.0301,
+      "step": 12360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005055600000000001,
+      "loss": 1.0751,
+      "step": 12361
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050552,
+      "loss": 1.0299,
+      "step": 12362
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005054800000000001,
+      "loss": 1.0363,
+      "step": 12363
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050544,
+      "loss": 1.0186,
+      "step": 12364
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005053999999999999,
+      "loss": 1.0269,
+      "step": 12365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050536,
+      "loss": 1.0351,
+      "step": 12366
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050532,
+      "loss": 0.9956,
+      "step": 12367
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050528,
+      "loss": 1.0008,
+      "step": 12368
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050524,
+      "loss": 1.0272,
+      "step": 12369
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005052,
+      "loss": 1.0249,
+      "step": 12370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050516,
+      "loss": 1.0718,
+      "step": 12371
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005051200000000001,
+      "loss": 0.9593,
+      "step": 12372
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050508,
+      "loss": 0.9366,
+      "step": 12373
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050504,
+      "loss": 1.0285,
+      "step": 12374
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000505,
+      "loss": 0.9964,
+      "step": 12375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005049599999999999,
+      "loss": 0.9952,
+      "step": 12376
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005049200000000001,
+      "loss": 0.9602,
+      "step": 12377
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050488,
+      "loss": 1.0542,
+      "step": 12378
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050484,
+      "loss": 1.0603,
+      "step": 12379
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005048,
+      "loss": 0.9606,
+      "step": 12380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050476,
+      "loss": 0.9711,
+      "step": 12381
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050472,
+      "loss": 1.0568,
+      "step": 12382
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005046800000000001,
+      "loss": 1.0464,
+      "step": 12383
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050464,
+      "loss": 1.0082,
+      "step": 12384
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005046,
+      "loss": 1.0139,
+      "step": 12385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050456,
+      "loss": 1.0705,
+      "step": 12386
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050452,
+      "loss": 1.006,
+      "step": 12387
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005044800000000001,
+      "loss": 0.9594,
+      "step": 12388
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050444,
+      "loss": 1.0826,
+      "step": 12389
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005044,
+      "loss": 1.0002,
+      "step": 12390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050436,
+      "loss": 0.963,
+      "step": 12391
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050432,
+      "loss": 1.0318,
+      "step": 12392
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050428,
+      "loss": 0.9238,
+      "step": 12393
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050424,
+      "loss": 0.986,
+      "step": 12394
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005042,
+      "loss": 0.9985,
+      "step": 12395
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050416,
+      "loss": 1.0743,
+      "step": 12396
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050412,
+      "loss": 0.9999,
+      "step": 12397
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050408,
+      "loss": 1.0341,
+      "step": 12398
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005040400000000001,
+      "loss": 1.075,
+      "step": 12399
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000504,
+      "loss": 0.9771,
+      "step": 12400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050396,
+      "loss": 1.0053,
+      "step": 12401
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050392,
+      "loss": 0.9094,
+      "step": 12402
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050388,
+      "loss": 1.1125,
+      "step": 12403
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050384,
+      "loss": 0.9395,
+      "step": 12404
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005038,
+      "loss": 1.0519,
+      "step": 12405
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050376,
+      "loss": 1.0319,
+      "step": 12406
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005037199999999999,
+      "loss": 0.9832,
+      "step": 12407
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005036800000000001,
+      "loss": 1.0722,
+      "step": 12408
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050364,
+      "loss": 1.108,
+      "step": 12409
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005036000000000001,
+      "loss": 1.0077,
+      "step": 12410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050356,
+      "loss": 1.0571,
+      "step": 12411
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050352,
+      "loss": 1.0629,
+      "step": 12412
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050348,
+      "loss": 1.0119,
+      "step": 12413
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050344,
+      "loss": 0.9905,
+      "step": 12414
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005034,
+      "loss": 0.9953,
+      "step": 12415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050336,
+      "loss": 1.0407,
+      "step": 12416
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050332,
+      "loss": 1.0415,
+      "step": 12417
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005032799999999999,
+      "loss": 1.1241,
+      "step": 12418
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005032400000000001,
+      "loss": 1.0977,
+      "step": 12419
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005032,
+      "loss": 1.0686,
+      "step": 12420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005031600000000001,
+      "loss": 1.0494,
+      "step": 12421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050312,
+      "loss": 1.0286,
+      "step": 12422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005030799999999999,
+      "loss": 1.1173,
+      "step": 12423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005030400000000001,
+      "loss": 1.0539,
+      "step": 12424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000503,
+      "loss": 1.0032,
+      "step": 12425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050296,
+      "loss": 0.9886,
+      "step": 12426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050292,
+      "loss": 1.0558,
+      "step": 12427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050288,
+      "loss": 0.998,
+      "step": 12428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050284,
+      "loss": 0.961,
+      "step": 12429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005028000000000001,
+      "loss": 1.0134,
+      "step": 12430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050276,
+      "loss": 0.9632,
+      "step": 12431
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005027199999999999,
+      "loss": 1.0705,
+      "step": 12432
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050268,
+      "loss": 1.0523,
+      "step": 12433
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005026399999999999,
+      "loss": 1.0781,
+      "step": 12434
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005026000000000001,
+      "loss": 1.0273,
+      "step": 12435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050256,
+      "loss": 1.1415,
+      "step": 12436
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050252,
+      "loss": 1.0267,
+      "step": 12437
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050248,
+      "loss": 0.9957,
+      "step": 12438
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050244,
+      "loss": 0.9857,
+      "step": 12439
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005024,
+      "loss": 0.9838,
+      "step": 12440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005023600000000001,
+      "loss": 0.9978,
+      "step": 12441
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050232,
+      "loss": 1.01,
+      "step": 12442
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005022799999999999,
+      "loss": 1.0604,
+      "step": 12443
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050224,
+      "loss": 1.0454,
+      "step": 12444
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005022,
+      "loss": 0.9706,
+      "step": 12445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005021600000000001,
+      "loss": 1.0235,
+      "step": 12446
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050212,
+      "loss": 1.0299,
+      "step": 12447
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050208,
+      "loss": 0.98,
+      "step": 12448
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050204,
+      "loss": 0.9764,
+      "step": 12449
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005020000000000001,
+      "loss": 1.1136,
+      "step": 12450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050196,
+      "loss": 0.9431,
+      "step": 12451
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050192,
+      "loss": 1.0137,
+      "step": 12452
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050188,
+      "loss": 1.0162,
+      "step": 12453
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005018399999999999,
+      "loss": 1.0566,
+      "step": 12454
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005018,
+      "loss": 1.1004,
+      "step": 12455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050176,
+      "loss": 1.0594,
+      "step": 12456
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005017200000000001,
+      "loss": 1.0318,
+      "step": 12457
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050168,
+      "loss": 1.1562,
+      "step": 12458
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050164,
+      "loss": 1.1064,
+      "step": 12459
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005016,
+      "loss": 1.0339,
+      "step": 12460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005015600000000001,
+      "loss": 1.0757,
+      "step": 12461
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050152,
+      "loss": 1.0145,
+      "step": 12462
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050148,
+      "loss": 0.9114,
+      "step": 12463
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050144,
+      "loss": 1.1124,
+      "step": 12464
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005013999999999999,
+      "loss": 1.0942,
+      "step": 12465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005013600000000001,
+      "loss": 0.949,
+      "step": 12466
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050132,
+      "loss": 1.062,
+      "step": 12467
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050128,
+      "loss": 1.0543,
+      "step": 12468
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050124,
+      "loss": 1.0212,
+      "step": 12469
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005012,
+      "loss": 0.9839,
+      "step": 12470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005011600000000001,
+      "loss": 1.0321,
+      "step": 12471
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050112,
+      "loss": 1.0092,
+      "step": 12472
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050108,
+      "loss": 1.0555,
+      "step": 12473
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050104,
+      "loss": 1.0138,
+      "step": 12474
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000501,
+      "loss": 1.008,
+      "step": 12475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050096,
+      "loss": 0.9999,
+      "step": 12476
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005009200000000001,
+      "loss": 1.0344,
+      "step": 12477
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050088,
+      "loss": 1.0428,
+      "step": 12478
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050084,
+      "loss": 0.9838,
+      "step": 12479
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005008,
+      "loss": 1.0104,
+      "step": 12480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005007599999999999,
+      "loss": 0.9938,
+      "step": 12481
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005007200000000001,
+      "loss": 1.0845,
+      "step": 12482
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050068,
+      "loss": 1.013,
+      "step": 12483
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050064,
+      "loss": 1.0364,
+      "step": 12484
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005006,
+      "loss": 1.089,
+      "step": 12485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050056,
+      "loss": 0.981,
+      "step": 12486
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050052,
+      "loss": 1.0333,
+      "step": 12487
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005004800000000001,
+      "loss": 0.9945,
+      "step": 12488
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050044,
+      "loss": 0.9753,
+      "step": 12489
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005004,
+      "loss": 0.9425,
+      "step": 12490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050036,
+      "loss": 0.9163,
+      "step": 12491
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050032,
+      "loss": 1.0224,
+      "step": 12492
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050028,
+      "loss": 1.0355,
+      "step": 12493
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050024,
+      "loss": 1.0389,
+      "step": 12494
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005002,
+      "loss": 1.0845,
+      "step": 12495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050016,
+      "loss": 1.0573,
+      "step": 12496
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005001200000000001,
+      "loss": 1.0619,
+      "step": 12497
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00050008,
+      "loss": 1.0483,
+      "step": 12498
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005000400000000001,
+      "loss": 1.0875,
+      "step": 12499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0005,
+      "loss": 1.0938,
+      "step": 12500
+    }
+  ],
+  "max_steps": 25000,
+  "num_train_epochs": 1,
+  "total_flos": 7.65554835062784e+19,
+  "trial_name": null,
+  "trial_params": null
+}