diff --git "a/checkpoint-20580/trainer_state.json" "b/checkpoint-20580/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-20580/trainer_state.json"
@@ -0,0 +1,123679 @@
+{
+  "best_metric": 0.013907751068472862,
+  "best_model_checkpoint": "depressed-tweet-detection-distilbert/checkpoint-20580",
+  "epoch": 20.0,
+  "eval_steps": 1,
+  "global_step": 20580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-08,
+      "loss": 0.7473,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-08,
+      "loss": 0.6623,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-08,
+      "loss": 0.6845,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-08,
+      "loss": 0.7351,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-08,
+      "loss": 0.5657,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6e-08,
+      "loss": 0.7478,
+      "step": 6
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7e-08,
+      "loss": 0.7155,
+      "step": 7
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8e-08,
+      "loss": 0.6262,
+      "step": 8
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9e-08,
+      "loss": 0.5688,
+      "step": 9
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1e-07,
+      "loss": 0.6751,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.0999999999999999e-07,
+      "loss": 0.6881,
+      "step": 11
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.2e-07,
+      "loss": 0.6566,
+      "step": 12
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.3e-07,
+      "loss": 0.682,
+      "step": 13
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4e-07,
+      "loss": 0.7304,
+      "step": 14
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.5e-07,
+      "loss": 0.7295,
+      "step": 15
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6e-07,
+      "loss": 0.722,
+      "step": 16
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.7000000000000001e-07,
+      "loss": 0.7703,
+      "step": 17
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8e-07,
+      "loss": 0.6071,
+      "step": 18
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.8999999999999998e-07,
+      "loss": 0.7038,
+      "step": 19
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2e-07,
+      "loss": 0.7082,
+      "step": 20
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.0999999999999997e-07,
+      "loss": 0.7486,
+      "step": 21
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.1999999999999998e-07,
+      "loss": 0.551,
+      "step": 22
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.3e-07,
+      "loss": 0.6535,
+      "step": 23
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.4e-07,
+      "loss": 0.6942,
+      "step": 24
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.5e-07,
+      "loss": 0.7208,
+      "step": 25
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.6e-07,
+      "loss": 0.5808,
+      "step": 26
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.7e-07,
+      "loss": 0.6948,
+      "step": 27
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.8e-07,
+      "loss": 0.738,
+      "step": 28
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2.9e-07,
+      "loss": 0.6841,
+      "step": 29
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3e-07,
+      "loss": 0.59,
+      "step": 30
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.1e-07,
+      "loss": 0.6606,
+      "step": 31
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.2e-07,
+      "loss": 0.6781,
+      "step": 32
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.3e-07,
+      "loss": 0.7466,
+      "step": 33
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.4000000000000003e-07,
+      "loss": 0.6626,
+      "step": 34
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.5e-07,
+      "loss": 0.6124,
+      "step": 35
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.6e-07,
+      "loss": 0.5859,
+      "step": 36
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.7e-07,
+      "loss": 0.6986,
+      "step": 37
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.7999999999999996e-07,
+      "loss": 0.6778,
+      "step": 38
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.8999999999999997e-07,
+      "loss": 0.6768,
+      "step": 39
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4e-07,
+      "loss": 0.5591,
+      "step": 40
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.0999999999999994e-07,
+      "loss": 0.6306,
+      "step": 41
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.1999999999999995e-07,
+      "loss": 0.6656,
+      "step": 42
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.2999999999999996e-07,
+      "loss": 0.6742,
+      "step": 43
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.3999999999999997e-07,
+      "loss": 0.6433,
+      "step": 44
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.5e-07,
+      "loss": 0.7217,
+      "step": 45
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.6e-07,
+      "loss": 0.6026,
+      "step": 46
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.6999999999999995e-07,
+      "loss": 0.5579,
+      "step": 47
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8e-07,
+      "loss": 0.6879,
+      "step": 48
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9e-07,
+      "loss": 0.6622,
+      "step": 49
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5e-07,
+      "loss": 0.7168,
+      "step": 50
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9997564539698e-07,
+      "loss": 0.6806,
+      "step": 51
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.999512907939601e-07,
+      "loss": 0.736,
+      "step": 52
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.9992693619094e-07,
+      "loss": 0.6887,
+      "step": 53
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.999025815879201e-07,
+      "loss": 0.6434,
+      "step": 54
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.998782269849001e-07,
+      "loss": 0.6691,
+      "step": 55
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.998538723818802e-07,
+      "loss": 0.5997,
+      "step": 56
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.998295177788601e-07,
+      "loss": 0.6797,
+      "step": 57
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.998051631758402e-07,
+      "loss": 0.7538,
+      "step": 58
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.997808085728202e-07,
+      "loss": 0.6424,
+      "step": 59
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.997564539698003e-07,
+      "loss": 0.6477,
+      "step": 60
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.997320993667803e-07,
+      "loss": 0.7178,
+      "step": 61
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.997077447637603e-07,
+      "loss": 0.682,
+      "step": 62
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.996833901607403e-07,
+      "loss": 0.6487,
+      "step": 63
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.996590355577204e-07,
+      "loss": 0.6997,
+      "step": 64
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.996346809547004e-07,
+      "loss": 0.6997,
+      "step": 65
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.996103263516805e-07,
+      "loss": 0.6258,
+      "step": 66
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.995859717486604e-07,
+      "loss": 0.6527,
+      "step": 67
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.995616171456405e-07,
+      "loss": 0.666,
+      "step": 68
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.995372625426205e-07,
+      "loss": 0.7311,
+      "step": 69
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.995129079396006e-07,
+      "loss": 0.5756,
+      "step": 70
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.994885533365805e-07,
+      "loss": 0.6865,
+      "step": 71
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.994641987335606e-07,
+      "loss": 0.6446,
+      "step": 72
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.994398441305406e-07,
+      "loss": 0.6821,
+      "step": 73
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.994154895275207e-07,
+      "loss": 0.6136,
+      "step": 74
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.993911349245007e-07,
+      "loss": 0.6134,
+      "step": 75
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.993667803214807e-07,
+      "loss": 0.6359,
+      "step": 76
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.993424257184607e-07,
+      "loss": 0.5695,
+      "step": 77
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.993180711154408e-07,
+      "loss": 0.7662,
+      "step": 78
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.992937165124208e-07,
+      "loss": 0.7065,
+      "step": 79
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.992693619094008e-07,
+      "loss": 0.7022,
+      "step": 80
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.992450073063808e-07,
+      "loss": 0.6463,
+      "step": 81
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.992206527033609e-07,
+      "loss": 0.5627,
+      "step": 82
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.991962981003409e-07,
+      "loss": 0.5653,
+      "step": 83
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.99171943497321e-07,
+      "loss": 0.7271,
+      "step": 84
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.991475888943009e-07,
+      "loss": 0.7716,
+      "step": 85
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.99123234291281e-07,
+      "loss": 0.6238,
+      "step": 86
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.99098879688261e-07,
+      "loss": 0.7056,
+      "step": 87
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.990745250852411e-07,
+      "loss": 0.7375,
+      "step": 88
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.990501704822211e-07,
+      "loss": 0.6329,
+      "step": 89
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.990258158792012e-07,
+      "loss": 0.7285,
+      "step": 90
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.990014612761812e-07,
+      "loss": 0.6914,
+      "step": 91
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.989771066731612e-07,
+      "loss": 0.7397,
+      "step": 92
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.989527520701412e-07,
+      "loss": 0.5605,
+      "step": 93
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.989283974671213e-07,
+      "loss": 0.675,
+      "step": 94
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.989040428641013e-07,
+      "loss": 0.7212,
+      "step": 95
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.988796882610813e-07,
+      "loss": 0.6365,
+      "step": 96
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.988553336580613e-07,
+      "loss": 0.6264,
+      "step": 97
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.988309790550414e-07,
+      "loss": 0.6528,
+      "step": 98
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.988066244520214e-07,
+      "loss": 0.7082,
+      "step": 99
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.987822698490015e-07,
+      "loss": 0.6957,
+      "step": 100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.987579152459814e-07,
+      "loss": 0.7115,
+      "step": 101
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.987335606429615e-07,
+      "loss": 0.7222,
+      "step": 102
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.987092060399415e-07,
+      "loss": 0.7276,
+      "step": 103
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.986848514369216e-07,
+      "loss": 0.7507,
+      "step": 104
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.986604968339017e-07,
+      "loss": 0.6819,
+      "step": 105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.986361422308816e-07,
+      "loss": 0.7129,
+      "step": 106
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.986117876278617e-07,
+      "loss": 0.6772,
+      "step": 107
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.985874330248417e-07,
+      "loss": 0.5638,
+      "step": 108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.985630784218218e-07,
+      "loss": 0.6502,
+      "step": 109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.985387238188017e-07,
+      "loss": 0.7086,
+      "step": 110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.985143692157818e-07,
+      "loss": 0.7229,
+      "step": 111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.984900146127618e-07,
+      "loss": 0.7078,
+      "step": 112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.984656600097419e-07,
+      "loss": 0.6904,
+      "step": 113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.984413054067219e-07,
+      "loss": 0.6645,
+      "step": 114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.984169508037019e-07,
+      "loss": 0.7696,
+      "step": 115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.983925962006819e-07,
+      "loss": 0.5584,
+      "step": 116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.98368241597662e-07,
+      "loss": 0.6602,
+      "step": 117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.98343886994642e-07,
+      "loss": 0.6828,
+      "step": 118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.98319532391622e-07,
+      "loss": 0.6914,
+      "step": 119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.98295177788602e-07,
+      "loss": 0.654,
+      "step": 120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.982708231855821e-07,
+      "loss": 0.6952,
+      "step": 121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.982464685825621e-07,
+      "loss": 0.7685,
+      "step": 122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.982221139795422e-07,
+      "loss": 0.6091,
+      "step": 123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.981977593765221e-07,
+      "loss": 0.6674,
+      "step": 124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.981734047735022e-07,
+      "loss": 0.6739,
+      "step": 125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.981490501704822e-07,
+      "loss": 0.6174,
+      "step": 126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.981246955674623e-07,
+      "loss": 0.6218,
+      "step": 127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.981003409644422e-07,
+      "loss": 0.684,
+      "step": 128
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.980759863614223e-07,
+      "loss": 0.6372,
+      "step": 129
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.980516317584023e-07,
+      "loss": 0.6394,
+      "step": 130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.980272771553824e-07,
+      "loss": 0.6312,
+      "step": 131
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.980029225523624e-07,
+      "loss": 0.6453,
+      "step": 132
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.979785679493424e-07,
+      "loss": 0.6461,
+      "step": 133
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.979542133463224e-07,
+      "loss": 0.5552,
+      "step": 134
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.979298587433025e-07,
+      "loss": 0.7402,
+      "step": 135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.979055041402825e-07,
+      "loss": 0.7216,
+      "step": 136
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.978811495372626e-07,
+      "loss": 0.7013,
+      "step": 137
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.978567949342425e-07,
+      "loss": 0.691,
+      "step": 138
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.978324403312226e-07,
+      "loss": 0.7293,
+      "step": 139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.978080857282026e-07,
+      "loss": 0.555,
+      "step": 140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.977837311251827e-07,
+      "loss": 0.676,
+      "step": 141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.977593765221626e-07,
+      "loss": 0.6965,
+      "step": 142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.977350219191427e-07,
+      "loss": 0.6772,
+      "step": 143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.977106673161227e-07,
+      "loss": 0.6768,
+      "step": 144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.976863127131028e-07,
+      "loss": 0.7203,
+      "step": 145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.976619581100828e-07,
+      "loss": 0.7206,
+      "step": 146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.976376035070628e-07,
+      "loss": 0.5678,
+      "step": 147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.976132489040428e-07,
+      "loss": 0.661,
+      "step": 148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.975888943010229e-07,
+      "loss": 0.6551,
+      "step": 149
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.975645396980029e-07,
+      "loss": 0.5613,
+      "step": 150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.975401850949829e-07,
+      "loss": 0.7094,
+      "step": 151
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.975158304919629e-07,
+      "loss": 0.679,
+      "step": 152
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.97491475888943e-07,
+      "loss": 0.6215,
+      "step": 153
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.97467121285923e-07,
+      "loss": 0.565,
+      "step": 154
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.974427666829031e-07,
+      "loss": 0.6944,
+      "step": 155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.97418412079883e-07,
+      "loss": 0.7519,
+      "step": 156
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.973940574768631e-07,
+      "loss": 0.6848,
+      "step": 157
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.973697028738431e-07,
+      "loss": 0.7011,
+      "step": 158
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.973453482708232e-07,
+      "loss": 0.7578,
+      "step": 159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.973209936678031e-07,
+      "loss": 0.6403,
+      "step": 160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.972966390647832e-07,
+      "loss": 0.6808,
+      "step": 161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.972722844617632e-07,
+      "loss": 0.6936,
+      "step": 162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.972479298587433e-07,
+      "loss": 0.7022,
+      "step": 163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.972235752557233e-07,
+      "loss": 0.6379,
+      "step": 164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.971992206527034e-07,
+      "loss": 0.7203,
+      "step": 165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.971748660496834e-07,
+      "loss": 0.5995,
+      "step": 166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.971505114466634e-07,
+      "loss": 0.5753,
+      "step": 167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.971261568436434e-07,
+      "loss": 0.6096,
+      "step": 168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.971018022406235e-07,
+      "loss": 0.6695,
+      "step": 169
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.970774476376035e-07,
+      "loss": 0.6928,
+      "step": 170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.970530930345836e-07,
+      "loss": 0.6236,
+      "step": 171
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.970287384315635e-07,
+      "loss": 0.6387,
+      "step": 172
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.970043838285436e-07,
+      "loss": 0.5463,
+      "step": 173
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.969800292255236e-07,
+      "loss": 0.6646,
+      "step": 174
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.969556746225037e-07,
+      "loss": 0.6682,
+      "step": 175
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.969313200194837e-07,
+      "loss": 0.6691,
+      "step": 176
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.969069654164637e-07,
+      "loss": 0.7389,
+      "step": 177
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.968826108134437e-07,
+      "loss": 0.6967,
+      "step": 178
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.968582562104238e-07,
+      "loss": 0.6772,
+      "step": 179
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.968339016074038e-07,
+      "loss": 0.7312,
+      "step": 180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.968095470043838e-07,
+      "loss": 0.6829,
+      "step": 181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.967851924013638e-07,
+      "loss": 0.7813,
+      "step": 182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.967608377983439e-07,
+      "loss": 0.6732,
+      "step": 183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.967364831953239e-07,
+      "loss": 0.6429,
+      "step": 184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.96712128592304e-07,
+      "loss": 0.6893,
+      "step": 185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.966877739892839e-07,
+      "loss": 0.6557,
+      "step": 186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.96663419386264e-07,
+      "loss": 0.7022,
+      "step": 187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.96639064783244e-07,
+      "loss": 0.6652,
+      "step": 188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.966147101802241e-07,
+      "loss": 0.6099,
+      "step": 189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.96590355577204e-07,
+      "loss": 0.5637,
+      "step": 190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.965660009741841e-07,
+      "loss": 0.6799,
+      "step": 191
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.965416463711641e-07,
+      "loss": 0.6631,
+      "step": 192
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.965172917681442e-07,
+      "loss": 0.6587,
+      "step": 193
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.964929371651242e-07,
+      "loss": 0.6511,
+      "step": 194
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.964685825621042e-07,
+      "loss": 0.729,
+      "step": 195
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.964442279590842e-07,
+      "loss": 0.6818,
+      "step": 196
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.964198733560643e-07,
+      "loss": 0.6773,
+      "step": 197
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.963955187530443e-07,
+      "loss": 0.6313,
+      "step": 198
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.963711641500243e-07,
+      "loss": 0.615,
+      "step": 199
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.963468095470043e-07,
+      "loss": 0.6745,
+      "step": 200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.963224549439844e-07,
+      "loss": 0.6666,
+      "step": 201
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.962981003409644e-07,
+      "loss": 0.6695,
+      "step": 202
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.962737457379445e-07,
+      "loss": 0.7002,
+      "step": 203
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.962493911349244e-07,
+      "loss": 0.6709,
+      "step": 204
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.962250365319045e-07,
+      "loss": 0.6859,
+      "step": 205
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.962006819288845e-07,
+      "loss": 0.636,
+      "step": 206
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.961763273258646e-07,
+      "loss": 0.7163,
+      "step": 207
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.961519727228446e-07,
+      "loss": 0.651,
+      "step": 208
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.961276181198246e-07,
+      "loss": 0.7217,
+      "step": 209
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.961032635168046e-07,
+      "loss": 0.7255,
+      "step": 210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.960789089137847e-07,
+      "loss": 0.7255,
+      "step": 211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.960545543107647e-07,
+      "loss": 0.6171,
+      "step": 212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.960301997077447e-07,
+      "loss": 0.605,
+      "step": 213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.960058451047247e-07,
+      "loss": 0.6724,
+      "step": 214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.959814905017048e-07,
+      "loss": 0.7199,
+      "step": 215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.959571358986848e-07,
+      "loss": 0.6152,
+      "step": 216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.959327812956649e-07,
+      "loss": 0.5937,
+      "step": 217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.959084266926448e-07,
+      "loss": 0.7388,
+      "step": 218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.958840720896249e-07,
+      "loss": 0.7306,
+      "step": 219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.958597174866049e-07,
+      "loss": 0.5986,
+      "step": 220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.95835362883585e-07,
+      "loss": 0.6852,
+      "step": 221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.958110082805649e-07,
+      "loss": 0.7005,
+      "step": 222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.95786653677545e-07,
+      "loss": 0.6188,
+      "step": 223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.95762299074525e-07,
+      "loss": 0.6359,
+      "step": 224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.957379444715051e-07,
+      "loss": 0.6511,
+      "step": 225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.957135898684851e-07,
+      "loss": 0.7179,
+      "step": 226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.956892352654651e-07,
+      "loss": 0.6891,
+      "step": 227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.956648806624451e-07,
+      "loss": 0.5678,
+      "step": 228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.956405260594252e-07,
+      "loss": 0.6408,
+      "step": 229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.956161714564052e-07,
+      "loss": 0.7342,
+      "step": 230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.955918168533853e-07,
+      "loss": 0.6408,
+      "step": 231
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.955674622503652e-07,
+      "loss": 0.7143,
+      "step": 232
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.955431076473453e-07,
+      "loss": 0.681,
+      "step": 233
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.955187530443253e-07,
+      "loss": 0.7453,
+      "step": 234
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.954943984413054e-07,
+      "loss": 0.5502,
+      "step": 235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.954700438382853e-07,
+      "loss": 0.578,
+      "step": 236
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.954456892352654e-07,
+      "loss": 0.693,
+      "step": 237
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.954213346322454e-07,
+      "loss": 0.647,
+      "step": 238
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.953969800292255e-07,
+      "loss": 0.6999,
+      "step": 239
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.953726254262056e-07,
+      "loss": 0.6087,
+      "step": 240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.953482708231855e-07,
+      "loss": 0.671,
+      "step": 241
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.953239162201657e-07,
+      "loss": 0.5916,
+      "step": 242
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.952995616171456e-07,
+      "loss": 0.6711,
+      "step": 243
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.952752070141257e-07,
+      "loss": 0.6436,
+      "step": 244
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.952508524111057e-07,
+      "loss": 0.6605,
+      "step": 245
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.952264978080858e-07,
+      "loss": 0.577,
+      "step": 246
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.952021432050658e-07,
+      "loss": 0.6921,
+      "step": 247
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.951777886020458e-07,
+      "loss": 0.6929,
+      "step": 248
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.951534339990258e-07,
+      "loss": 0.6811,
+      "step": 249
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.951290793960059e-07,
+      "loss": 0.6699,
+      "step": 250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.951047247929859e-07,
+      "loss": 0.6661,
+      "step": 251
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.950803701899659e-07,
+      "loss": 0.716,
+      "step": 252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.950560155869459e-07,
+      "loss": 0.5945,
+      "step": 253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.95031660983926e-07,
+      "loss": 0.6362,
+      "step": 254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.95007306380906e-07,
+      "loss": 0.7142,
+      "step": 255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.949829517778861e-07,
+      "loss": 0.6264,
+      "step": 256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.94958597174866e-07,
+      "loss": 0.6373,
+      "step": 257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.949342425718461e-07,
+      "loss": 0.6462,
+      "step": 258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.949098879688261e-07,
+      "loss": 0.6478,
+      "step": 259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.948855333658062e-07,
+      "loss": 0.7019,
+      "step": 260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.948611787627861e-07,
+      "loss": 0.6335,
+      "step": 261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.948368241597662e-07,
+      "loss": 0.6636,
+      "step": 262
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.948124695567462e-07,
+      "loss": 0.599,
+      "step": 263
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.947881149537263e-07,
+      "loss": 0.6472,
+      "step": 264
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.947637603507063e-07,
+      "loss": 0.6702,
+      "step": 265
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.947394057476863e-07,
+      "loss": 0.6933,
+      "step": 266
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.947150511446663e-07,
+      "loss": 0.6183,
+      "step": 267
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.946906965416464e-07,
+      "loss": 0.654,
+      "step": 268
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.946663419386264e-07,
+      "loss": 0.682,
+      "step": 269
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.946419873356065e-07,
+      "loss": 0.6291,
+      "step": 270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.946176327325864e-07,
+      "loss": 0.5966,
+      "step": 271
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.945932781295665e-07,
+      "loss": 0.6861,
+      "step": 272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.945689235265465e-07,
+      "loss": 0.6573,
+      "step": 273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.945445689235266e-07,
+      "loss": 0.7463,
+      "step": 274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.945202143205065e-07,
+      "loss": 0.7,
+      "step": 275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.944958597174866e-07,
+      "loss": 0.6203,
+      "step": 276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.944715051144666e-07,
+      "loss": 0.6121,
+      "step": 277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.944471505114467e-07,
+      "loss": 0.6924,
+      "step": 278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.944227959084267e-07,
+      "loss": 0.6616,
+      "step": 279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.943984413054067e-07,
+      "loss": 0.6784,
+      "step": 280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.943740867023867e-07,
+      "loss": 0.6459,
+      "step": 281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.943497320993668e-07,
+      "loss": 0.6479,
+      "step": 282
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.943253774963468e-07,
+      "loss": 0.6708,
+      "step": 283
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.943010228933268e-07,
+      "loss": 0.6341,
+      "step": 284
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.942766682903068e-07,
+      "loss": 0.6244,
+      "step": 285
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.942523136872869e-07,
+      "loss": 0.6859,
+      "step": 286
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.942279590842669e-07,
+      "loss": 0.7164,
+      "step": 287
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.94203604481247e-07,
+      "loss": 0.6378,
+      "step": 288
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.941792498782269e-07,
+      "loss": 0.6009,
+      "step": 289
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.94154895275207e-07,
+      "loss": 0.6195,
+      "step": 290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.94130540672187e-07,
+      "loss": 0.6271,
+      "step": 291
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.941061860691671e-07,
+      "loss": 0.6106,
+      "step": 292
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.94081831466147e-07,
+      "loss": 0.6423,
+      "step": 293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.940574768631271e-07,
+      "loss": 0.632,
+      "step": 294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.940331222601071e-07,
+      "loss": 0.6426,
+      "step": 295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.940087676570872e-07,
+      "loss": 0.6801,
+      "step": 296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.939844130540672e-07,
+      "loss": 0.6604,
+      "step": 297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.939600584510472e-07,
+      "loss": 0.5685,
+      "step": 298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.939357038480272e-07,
+      "loss": 0.7121,
+      "step": 299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.939113492450073e-07,
+      "loss": 0.644,
+      "step": 300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.938869946419873e-07,
+      "loss": 0.6461,
+      "step": 301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.938626400389674e-07,
+      "loss": 0.6976,
+      "step": 302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.938382854359473e-07,
+      "loss": 0.6391,
+      "step": 303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.938139308329274e-07,
+      "loss": 0.6417,
+      "step": 304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.937895762299074e-07,
+      "loss": 0.621,
+      "step": 305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.937652216268875e-07,
+      "loss": 0.5525,
+      "step": 306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.937408670238674e-07,
+      "loss": 0.5865,
+      "step": 307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.937165124208475e-07,
+      "loss": 0.6695,
+      "step": 308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.936921578178275e-07,
+      "loss": 0.7184,
+      "step": 309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.936678032148076e-07,
+      "loss": 0.6035,
+      "step": 310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.936434486117876e-07,
+      "loss": 0.6758,
+      "step": 311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.936190940087676e-07,
+      "loss": 0.5984,
+      "step": 312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.935947394057476e-07,
+      "loss": 0.7126,
+      "step": 313
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.935703848027277e-07,
+      "loss": 0.6231,
+      "step": 314
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.935460301997077e-07,
+      "loss": 0.6147,
+      "step": 315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.935216755966877e-07,
+      "loss": 0.6699,
+      "step": 316
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.934973209936677e-07,
+      "loss": 0.6065,
+      "step": 317
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.934729663906478e-07,
+      "loss": 0.6502,
+      "step": 318
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.934486117876278e-07,
+      "loss": 0.6019,
+      "step": 319
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.934242571846079e-07,
+      "loss": 0.6743,
+      "step": 320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.933999025815879e-07,
+      "loss": 0.5822,
+      "step": 321
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.93375547978568e-07,
+      "loss": 0.6516,
+      "step": 322
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.933511933755479e-07,
+      "loss": 0.658,
+      "step": 323
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.93326838772528e-07,
+      "loss": 0.6367,
+      "step": 324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.93302484169508e-07,
+      "loss": 0.6904,
+      "step": 325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.932781295664881e-07,
+      "loss": 0.6146,
+      "step": 326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.932537749634681e-07,
+      "loss": 0.678,
+      "step": 327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.932294203604481e-07,
+      "loss": 0.6568,
+      "step": 328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.932050657574281e-07,
+      "loss": 0.6485,
+      "step": 329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.931807111544082e-07,
+      "loss": 0.6297,
+      "step": 330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.931563565513882e-07,
+      "loss": 0.6081,
+      "step": 331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.931320019483682e-07,
+      "loss": 0.5532,
+      "step": 332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.931076473453482e-07,
+      "loss": 0.6067,
+      "step": 333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.930832927423283e-07,
+      "loss": 0.6134,
+      "step": 334
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.930589381393083e-07,
+      "loss": 0.6253,
+      "step": 335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.930345835362884e-07,
+      "loss": 0.6256,
+      "step": 336
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.930102289332683e-07,
+      "loss": 0.7057,
+      "step": 337
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.929858743302484e-07,
+      "loss": 0.6093,
+      "step": 338
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.929615197272284e-07,
+      "loss": 0.6174,
+      "step": 339
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.929371651242085e-07,
+      "loss": 0.578,
+      "step": 340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.929128105211885e-07,
+      "loss": 0.6413,
+      "step": 341
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.928884559181685e-07,
+      "loss": 0.649,
+      "step": 342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.928641013151485e-07,
+      "loss": 0.6511,
+      "step": 343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.928397467121286e-07,
+      "loss": 0.5886,
+      "step": 344
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.928153921091086e-07,
+      "loss": 0.6701,
+      "step": 345
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.927910375060886e-07,
+      "loss": 0.6476,
+      "step": 346
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.927666829030686e-07,
+      "loss": 0.6524,
+      "step": 347
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.927423283000487e-07,
+      "loss": 0.6115,
+      "step": 348
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.927179736970287e-07,
+      "loss": 0.7012,
+      "step": 349
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.926936190940088e-07,
+      "loss": 0.6604,
+      "step": 350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.926692644909887e-07,
+      "loss": 0.6257,
+      "step": 351
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.926449098879688e-07,
+      "loss": 0.6243,
+      "step": 352
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.926205552849488e-07,
+      "loss": 0.5848,
+      "step": 353
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.925962006819289e-07,
+      "loss": 0.6245,
+      "step": 354
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.925718460789088e-07,
+      "loss": 0.6003,
+      "step": 355
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.925474914758889e-07,
+      "loss": 0.5624,
+      "step": 356
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.92523136872869e-07,
+      "loss": 0.6174,
+      "step": 357
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.92498782269849e-07,
+      "loss": 0.5994,
+      "step": 358
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.92474427666829e-07,
+      "loss": 0.6099,
+      "step": 359
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.92450073063809e-07,
+      "loss": 0.606,
+      "step": 360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.92425718460789e-07,
+      "loss": 0.62,
+      "step": 361
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.924013638577691e-07,
+      "loss": 0.6551,
+      "step": 362
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.923770092547492e-07,
+      "loss": 0.6013,
+      "step": 363
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.923526546517291e-07,
+      "loss": 0.6041,
+      "step": 364
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.923283000487091e-07,
+      "loss": 0.7171,
+      "step": 365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.923039454456892e-07,
+      "loss": 0.6396,
+      "step": 366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.922795908426693e-07,
+      "loss": 0.6793,
+      "step": 367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.922552362396493e-07,
+      "loss": 0.6318,
+      "step": 368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.922308816366292e-07,
+      "loss": 0.6462,
+      "step": 369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.922065270336093e-07,
+      "loss": 0.6447,
+      "step": 370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.921821724305894e-07,
+      "loss": 0.6211,
+      "step": 371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.921578178275694e-07,
+      "loss": 0.6513,
+      "step": 372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.921334632245495e-07,
+      "loss": 0.5957,
+      "step": 373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.921091086215294e-07,
+      "loss": 0.6173,
+      "step": 374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.920847540185095e-07,
+      "loss": 0.5845,
+      "step": 375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.920603994154895e-07,
+      "loss": 0.6452,
+      "step": 376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.920360448124696e-07,
+      "loss": 0.6523,
+      "step": 377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.920116902094495e-07,
+      "loss": 0.5884,
+      "step": 378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.919873356064296e-07,
+      "loss": 0.6447,
+      "step": 379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.919629810034096e-07,
+      "loss": 0.6539,
+      "step": 380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.919386264003897e-07,
+      "loss": 0.6099,
+      "step": 381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.919142717973697e-07,
+      "loss": 0.6254,
+      "step": 382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.918899171943497e-07,
+      "loss": 0.6402,
+      "step": 383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.918655625913297e-07,
+      "loss": 0.5796,
+      "step": 384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.918412079883098e-07,
+      "loss": 0.593,
+      "step": 385
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.918168533852898e-07,
+      "loss": 0.589,
+      "step": 386
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.917924987822698e-07,
+      "loss": 0.6567,
+      "step": 387
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.917681441792498e-07,
+      "loss": 0.6368,
+      "step": 388
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.917437895762299e-07,
+      "loss": 0.6601,
+      "step": 389
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.917194349732099e-07,
+      "loss": 0.6063,
+      "step": 390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.9169508037019e-07,
+      "loss": 0.5992,
+      "step": 391
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.916707257671699e-07,
+      "loss": 0.6268,
+      "step": 392
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.9164637116415e-07,
+      "loss": 0.6294,
+      "step": 393
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.9162201656113e-07,
+      "loss": 0.625,
+      "step": 394
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.915976619581101e-07,
+      "loss": 0.6295,
+      "step": 395
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.915733073550901e-07,
+      "loss": 0.612,
+      "step": 396
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.915489527520702e-07,
+      "loss": 0.6227,
+      "step": 397
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.915245981490502e-07,
+      "loss": 0.6305,
+      "step": 398
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.915002435460302e-07,
+      "loss": 0.6576,
+      "step": 399
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.914758889430102e-07,
+      "loss": 0.638,
+      "step": 400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.914515343399903e-07,
+      "loss": 0.66,
+      "step": 401
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.914271797369703e-07,
+      "loss": 0.5949,
+      "step": 402
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.914028251339503e-07,
+      "loss": 0.5912,
+      "step": 403
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.913784705309303e-07,
+      "loss": 0.6055,
+      "step": 404
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.913541159279104e-07,
+      "loss": 0.5654,
+      "step": 405
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.913297613248904e-07,
+      "loss": 0.608,
+      "step": 406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.913054067218705e-07,
+      "loss": 0.6157,
+      "step": 407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.912810521188504e-07,
+      "loss": 0.6062,
+      "step": 408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.912566975158305e-07,
+      "loss": 0.6128,
+      "step": 409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.912323429128105e-07,
+      "loss": 0.6611,
+      "step": 410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.912079883097906e-07,
+      "loss": 0.6055,
+      "step": 411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.911836337067706e-07,
+      "loss": 0.5684,
+      "step": 412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.911592791037506e-07,
+      "loss": 0.5969,
+      "step": 413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.911349245007306e-07,
+      "loss": 0.5827,
+      "step": 414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.911105698977107e-07,
+      "loss": 0.6213,
+      "step": 415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.910862152946907e-07,
+      "loss": 0.6185,
+      "step": 416
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.910618606916707e-07,
+      "loss": 0.5993,
+      "step": 417
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.910375060886507e-07,
+      "loss": 0.5967,
+      "step": 418
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.910131514856308e-07,
+      "loss": 0.6167,
+      "step": 419
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.909887968826108e-07,
+      "loss": 0.6106,
+      "step": 420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.909644422795909e-07,
+      "loss": 0.5822,
+      "step": 421
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.909400876765708e-07,
+      "loss": 0.6307,
+      "step": 422
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.909157330735509e-07,
+      "loss": 0.5923,
+      "step": 423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.908913784705309e-07,
+      "loss": 0.5719,
+      "step": 424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.90867023867511e-07,
+      "loss": 0.6871,
+      "step": 425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.908426692644909e-07,
+      "loss": 0.5752,
+      "step": 426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.90818314661471e-07,
+      "loss": 0.5918,
+      "step": 427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.90793960058451e-07,
+      "loss": 0.6079,
+      "step": 428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.907696054554311e-07,
+      "loss": 0.6218,
+      "step": 429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.907452508524111e-07,
+      "loss": 0.5545,
+      "step": 430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.907208962493911e-07,
+      "loss": 0.5614,
+      "step": 431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.906965416463711e-07,
+      "loss": 0.6283,
+      "step": 432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.906721870433512e-07,
+      "loss": 0.6231,
+      "step": 433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.906478324403312e-07,
+      "loss": 0.6074,
+      "step": 434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.906234778373113e-07,
+      "loss": 0.6173,
+      "step": 435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.905991232342912e-07,
+      "loss": 0.5759,
+      "step": 436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.905747686312713e-07,
+      "loss": 0.6954,
+      "step": 437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.905504140282513e-07,
+      "loss": 0.6062,
+      "step": 438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.905260594252314e-07,
+      "loss": 0.6307,
+      "step": 439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.905017048222113e-07,
+      "loss": 0.5788,
+      "step": 440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.904773502191914e-07,
+      "loss": 0.5842,
+      "step": 441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.904529956161714e-07,
+      "loss": 0.6071,
+      "step": 442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.904286410131515e-07,
+      "loss": 0.6198,
+      "step": 443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.904042864101315e-07,
+      "loss": 0.6358,
+      "step": 444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.903799318071115e-07,
+      "loss": 0.5525,
+      "step": 445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.903555772040915e-07,
+      "loss": 0.5561,
+      "step": 446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.903312226010716e-07,
+      "loss": 0.5756,
+      "step": 447
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.903068679980516e-07,
+      "loss": 0.5527,
+      "step": 448
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.902825133950316e-07,
+      "loss": 0.5756,
+      "step": 449
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.902581587920116e-07,
+      "loss": 0.5699,
+      "step": 450
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.902338041889917e-07,
+      "loss": 0.5374,
+      "step": 451
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.902094495859717e-07,
+      "loss": 0.5672,
+      "step": 452
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.901850949829518e-07,
+      "loss": 0.5714,
+      "step": 453
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.901607403799317e-07,
+      "loss": 0.5979,
+      "step": 454
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.901363857769118e-07,
+      "loss": 0.621,
+      "step": 455
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.901120311738918e-07,
+      "loss": 0.6123,
+      "step": 456
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.900876765708719e-07,
+      "loss": 0.6249,
+      "step": 457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.900633219678518e-07,
+      "loss": 0.6007,
+      "step": 458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.900389673648319e-07,
+      "loss": 0.6156,
+      "step": 459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.900146127618119e-07,
+      "loss": 0.5539,
+      "step": 460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.89990258158792e-07,
+      "loss": 0.5685,
+      "step": 461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.89965903555772e-07,
+      "loss": 0.5636,
+      "step": 462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.89941548952752e-07,
+      "loss": 0.5856,
+      "step": 463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.89917194349732e-07,
+      "loss": 0.6082,
+      "step": 464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.898928397467121e-07,
+      "loss": 0.5255,
+      "step": 465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.898684851436921e-07,
+      "loss": 0.5786,
+      "step": 466
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.898441305406722e-07,
+      "loss": 0.6236,
+      "step": 467
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.898197759376521e-07,
+      "loss": 0.5223,
+      "step": 468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.897954213346322e-07,
+      "loss": 0.6642,
+      "step": 469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.897710667316122e-07,
+      "loss": 0.5709,
+      "step": 470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.897467121285923e-07,
+      "loss": 0.559,
+      "step": 471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.897223575255723e-07,
+      "loss": 0.6033,
+      "step": 472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.896980029225524e-07,
+      "loss": 0.5763,
+      "step": 473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.896736483195324e-07,
+      "loss": 0.5408,
+      "step": 474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.896492937165124e-07,
+      "loss": 0.5384,
+      "step": 475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.896249391134924e-07,
+      "loss": 0.5618,
+      "step": 476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.896005845104725e-07,
+      "loss": 0.6021,
+      "step": 477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.895762299074525e-07,
+      "loss": 0.6016,
+      "step": 478
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.895518753044325e-07,
+      "loss": 0.5446,
+      "step": 479
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.895275207014125e-07,
+      "loss": 0.5392,
+      "step": 480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.895031660983926e-07,
+      "loss": 0.5296,
+      "step": 481
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.894788114953726e-07,
+      "loss": 0.5236,
+      "step": 482
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.894544568923527e-07,
+      "loss": 0.6439,
+      "step": 483
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.894301022893326e-07,
+      "loss": 0.5581,
+      "step": 484
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.894057476863127e-07,
+      "loss": 0.5122,
+      "step": 485
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.893813930832927e-07,
+      "loss": 0.555,
+      "step": 486
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.893570384802728e-07,
+      "loss": 0.5793,
+      "step": 487
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.893326838772527e-07,
+      "loss": 0.5941,
+      "step": 488
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.893083292742328e-07,
+      "loss": 0.5461,
+      "step": 489
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.892839746712129e-07,
+      "loss": 0.5883,
+      "step": 490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.892596200681929e-07,
+      "loss": 0.5862,
+      "step": 491
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.89235265465173e-07,
+      "loss": 0.5542,
+      "step": 492
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.892109108621529e-07,
+      "loss": 0.5263,
+      "step": 493
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.89186556259133e-07,
+      "loss": 0.5549,
+      "step": 494
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.89162201656113e-07,
+      "loss": 0.574,
+      "step": 495
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.891378470530931e-07,
+      "loss": 0.6004,
+      "step": 496
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.89113492450073e-07,
+      "loss": 0.5387,
+      "step": 497
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.89089137847053e-07,
+      "loss": 0.5564,
+      "step": 498
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.890647832440331e-07,
+      "loss": 0.5603,
+      "step": 499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.890404286410132e-07,
+      "loss": 0.5171,
+      "step": 500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.890160740379932e-07,
+      "loss": 0.5263,
+      "step": 501
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.889917194349732e-07,
+      "loss": 0.5804,
+      "step": 502
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.889673648319532e-07,
+      "loss": 0.544,
+      "step": 503
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.889430102289333e-07,
+      "loss": 0.5751,
+      "step": 504
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.889186556259133e-07,
+      "loss": 0.551,
+      "step": 505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.888943010228934e-07,
+      "loss": 0.5354,
+      "step": 506
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.888699464198733e-07,
+      "loss": 0.5757,
+      "step": 507
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.888455918168534e-07,
+      "loss": 0.5481,
+      "step": 508
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.888212372138334e-07,
+      "loss": 0.6668,
+      "step": 509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.887968826108135e-07,
+      "loss": 0.5799,
+      "step": 510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.887725280077934e-07,
+      "loss": 0.6331,
+      "step": 511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.887481734047735e-07,
+      "loss": 0.517,
+      "step": 512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.887238188017535e-07,
+      "loss": 0.6045,
+      "step": 513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.886994641987336e-07,
+      "loss": 0.5072,
+      "step": 514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.886751095957136e-07,
+      "loss": 0.5258,
+      "step": 515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.886507549926936e-07,
+      "loss": 0.5334,
+      "step": 516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.886264003896736e-07,
+      "loss": 0.5517,
+      "step": 517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.886020457866537e-07,
+      "loss": 0.5767,
+      "step": 518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.885776911836337e-07,
+      "loss": 0.5039,
+      "step": 519
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.885533365806137e-07,
+      "loss": 0.5914,
+      "step": 520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.885289819775937e-07,
+      "loss": 0.5259,
+      "step": 521
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.885046273745738e-07,
+      "loss": 0.5523,
+      "step": 522
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.884802727715538e-07,
+      "loss": 0.502,
+      "step": 523
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.884559181685339e-07,
+      "loss": 0.5116,
+      "step": 524
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.884315635655138e-07,
+      "loss": 0.5317,
+      "step": 525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.884072089624939e-07,
+      "loss": 0.5185,
+      "step": 526
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.883828543594739e-07,
+      "loss": 0.5093,
+      "step": 527
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.88358499756454e-07,
+      "loss": 0.503,
+      "step": 528
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.883341451534339e-07,
+      "loss": 0.4964,
+      "step": 529
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.88309790550414e-07,
+      "loss": 0.5346,
+      "step": 530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.88285435947394e-07,
+      "loss": 0.5393,
+      "step": 531
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.882610813443741e-07,
+      "loss": 0.536,
+      "step": 532
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.882367267413541e-07,
+      "loss": 0.5344,
+      "step": 533
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.882123721383341e-07,
+      "loss": 0.5827,
+      "step": 534
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.881880175353141e-07,
+      "loss": 0.5211,
+      "step": 535
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.881636629322942e-07,
+      "loss": 0.6392,
+      "step": 536
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.881393083292742e-07,
+      "loss": 0.5864,
+      "step": 537
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.881149537262543e-07,
+      "loss": 0.4918,
+      "step": 538
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.880905991232342e-07,
+      "loss": 0.5652,
+      "step": 539
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.880662445202143e-07,
+      "loss": 0.5345,
+      "step": 540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.880418899171943e-07,
+      "loss": 0.5305,
+      "step": 541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.880175353141744e-07,
+      "loss": 0.5715,
+      "step": 542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.879931807111543e-07,
+      "loss": 0.5791,
+      "step": 543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.879688261081344e-07,
+      "loss": 0.529,
+      "step": 544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.879444715051144e-07,
+      "loss": 0.5037,
+      "step": 545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.879201169020945e-07,
+      "loss": 0.5291,
+      "step": 546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.878957622990745e-07,
+      "loss": 0.5304,
+      "step": 547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.878714076960546e-07,
+      "loss": 0.4917,
+      "step": 548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.878470530930346e-07,
+      "loss": 0.4442,
+      "step": 549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.878226984900146e-07,
+      "loss": 0.4705,
+      "step": 550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.877983438869946e-07,
+      "loss": 0.5827,
+      "step": 551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.877739892839747e-07,
+      "loss": 0.4928,
+      "step": 552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.877496346809547e-07,
+      "loss": 0.5636,
+      "step": 553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.877252800779348e-07,
+      "loss": 0.574,
+      "step": 554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.877009254749147e-07,
+      "loss": 0.4939,
+      "step": 555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.876765708718948e-07,
+      "loss": 0.5584,
+      "step": 556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.876522162688748e-07,
+      "loss": 0.5602,
+      "step": 557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.876278616658549e-07,
+      "loss": 0.5328,
+      "step": 558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.876035070628348e-07,
+      "loss": 0.5133,
+      "step": 559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.875791524598149e-07,
+      "loss": 0.5543,
+      "step": 560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.875547978567949e-07,
+      "loss": 0.5148,
+      "step": 561
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.87530443253775e-07,
+      "loss": 0.4337,
+      "step": 562
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.87506088650755e-07,
+      "loss": 0.5072,
+      "step": 563
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.87481734047735e-07,
+      "loss": 0.5621,
+      "step": 564
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.87457379444715e-07,
+      "loss": 0.4615,
+      "step": 565
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.874330248416951e-07,
+      "loss": 0.4832,
+      "step": 566
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.874086702386751e-07,
+      "loss": 0.5822,
+      "step": 567
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.873843156356551e-07,
+      "loss": 0.5997,
+      "step": 568
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.873599610326351e-07,
+      "loss": 0.5046,
+      "step": 569
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.873356064296152e-07,
+      "loss": 0.4994,
+      "step": 570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.873112518265952e-07,
+      "loss": 0.4451,
+      "step": 571
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.872868972235753e-07,
+      "loss": 0.4743,
+      "step": 572
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.872625426205552e-07,
+      "loss": 0.456,
+      "step": 573
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.872381880175353e-07,
+      "loss": 0.5741,
+      "step": 574
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.872138334145153e-07,
+      "loss": 0.4565,
+      "step": 575
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.871894788114954e-07,
+      "loss": 0.4911,
+      "step": 576
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.871651242084754e-07,
+      "loss": 0.4847,
+      "step": 577
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.871407696054554e-07,
+      "loss": 0.5873,
+      "step": 578
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.871164150024354e-07,
+      "loss": 0.444,
+      "step": 579
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.870920603994155e-07,
+      "loss": 0.5274,
+      "step": 580
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.870677057963955e-07,
+      "loss": 0.4931,
+      "step": 581
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.870433511933755e-07,
+      "loss": 0.4775,
+      "step": 582
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.870189965903555e-07,
+      "loss": 0.5897,
+      "step": 583
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.869946419873356e-07,
+      "loss": 0.4278,
+      "step": 584
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.869702873843156e-07,
+      "loss": 0.4715,
+      "step": 585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.869459327812957e-07,
+      "loss": 0.494,
+      "step": 586
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.869215781782756e-07,
+      "loss": 0.4792,
+      "step": 587
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.868972235752557e-07,
+      "loss": 0.4833,
+      "step": 588
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.868728689722357e-07,
+      "loss": 0.5076,
+      "step": 589
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.868485143692158e-07,
+      "loss": 0.5473,
+      "step": 590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.868241597661957e-07,
+      "loss": 0.4408,
+      "step": 591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.867998051631758e-07,
+      "loss": 0.4501,
+      "step": 592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.867754505601558e-07,
+      "loss": 0.4628,
+      "step": 593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.867510959571359e-07,
+      "loss": 0.4818,
+      "step": 594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.867267413541159e-07,
+      "loss": 0.4696,
+      "step": 595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.867023867510959e-07,
+      "loss": 0.4314,
+      "step": 596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.866780321480759e-07,
+      "loss": 0.472,
+      "step": 597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.86653677545056e-07,
+      "loss": 0.465,
+      "step": 598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.86629322942036e-07,
+      "loss": 0.44,
+      "step": 599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.866049683390161e-07,
+      "loss": 0.5,
+      "step": 600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.86580613735996e-07,
+      "loss": 0.5247,
+      "step": 601
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.865562591329761e-07,
+      "loss": 0.4858,
+      "step": 602
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.865319045299561e-07,
+      "loss": 0.4247,
+      "step": 603
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.865075499269362e-07,
+      "loss": 0.499,
+      "step": 604
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.864831953239161e-07,
+      "loss": 0.4966,
+      "step": 605
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.864588407208962e-07,
+      "loss": 0.5026,
+      "step": 606
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.864344861178762e-07,
+      "loss": 0.4366,
+      "step": 607
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.864101315148563e-07,
+      "loss": 0.5915,
+      "step": 608
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.863857769118363e-07,
+      "loss": 0.4378,
+      "step": 609
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.863614223088163e-07,
+      "loss": 0.4186,
+      "step": 610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.863370677057963e-07,
+      "loss": 0.5113,
+      "step": 611
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.863127131027764e-07,
+      "loss": 0.417,
+      "step": 612
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.862883584997564e-07,
+      "loss": 0.4292,
+      "step": 613
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.862640038967364e-07,
+      "loss": 0.4547,
+      "step": 614
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.862396492937164e-07,
+      "loss": 0.4401,
+      "step": 615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.862152946906965e-07,
+      "loss": 0.4422,
+      "step": 616
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.861909400876765e-07,
+      "loss": 0.4568,
+      "step": 617
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.861665854846566e-07,
+      "loss": 0.4362,
+      "step": 618
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.861422308816365e-07,
+      "loss": 0.4114,
+      "step": 619
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.861178762786166e-07,
+      "loss": 0.4112,
+      "step": 620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.860935216755966e-07,
+      "loss": 0.4493,
+      "step": 621
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.860691670725767e-07,
+      "loss": 0.4028,
+      "step": 622
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.860448124695566e-07,
+      "loss": 0.5415,
+      "step": 623
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.860204578665367e-07,
+      "loss": 0.4431,
+      "step": 624
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.859961032635169e-07,
+      "loss": 0.5754,
+      "step": 625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.859717486604968e-07,
+      "loss": 0.5584,
+      "step": 626
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.859473940574769e-07,
+      "loss": 0.5293,
+      "step": 627
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.859230394544569e-07,
+      "loss": 0.4995,
+      "step": 628
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.85898684851437e-07,
+      "loss": 0.6095,
+      "step": 629
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.858743302484169e-07,
+      "loss": 0.3694,
+      "step": 630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.85849975645397e-07,
+      "loss": 0.4586,
+      "step": 631
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.85825621042377e-07,
+      "loss": 0.4887,
+      "step": 632
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.858012664393571e-07,
+      "loss": 0.4428,
+      "step": 633
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.857769118363371e-07,
+      "loss": 0.4294,
+      "step": 634
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.857525572333171e-07,
+      "loss": 0.4367,
+      "step": 635
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.857282026302971e-07,
+      "loss": 0.4854,
+      "step": 636
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.857038480272772e-07,
+      "loss": 0.4042,
+      "step": 637
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.856794934242572e-07,
+      "loss": 0.4236,
+      "step": 638
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.856551388212373e-07,
+      "loss": 0.6426,
+      "step": 639
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.856307842182172e-07,
+      "loss": 0.5753,
+      "step": 640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.856064296151973e-07,
+      "loss": 0.4789,
+      "step": 641
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.855820750121773e-07,
+      "loss": 0.4934,
+      "step": 642
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.855577204091574e-07,
+      "loss": 0.4071,
+      "step": 643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.855333658061373e-07,
+      "loss": 0.4838,
+      "step": 644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.855090112031174e-07,
+      "loss": 0.4224,
+      "step": 645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.854846566000974e-07,
+      "loss": 0.6492,
+      "step": 646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.854603019970775e-07,
+      "loss": 0.391,
+      "step": 647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.854359473940575e-07,
+      "loss": 0.4573,
+      "step": 648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.854115927910375e-07,
+      "loss": 0.4679,
+      "step": 649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.853872381880175e-07,
+      "loss": 0.4431,
+      "step": 650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.853628835849976e-07,
+      "loss": 0.4301,
+      "step": 651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.853385289819776e-07,
+      "loss": 0.433,
+      "step": 652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.853141743789576e-07,
+      "loss": 0.3839,
+      "step": 653
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.852898197759376e-07,
+      "loss": 0.3824,
+      "step": 654
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.852654651729177e-07,
+      "loss": 0.4193,
+      "step": 655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.852411105698977e-07,
+      "loss": 0.5277,
+      "step": 656
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.852167559668778e-07,
+      "loss": 0.4452,
+      "step": 657
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.851924013638577e-07,
+      "loss": 0.4477,
+      "step": 658
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.851680467608378e-07,
+      "loss": 0.3607,
+      "step": 659
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.851436921578178e-07,
+      "loss": 0.4928,
+      "step": 660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.851193375547979e-07,
+      "loss": 0.4484,
+      "step": 661
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.850949829517778e-07,
+      "loss": 0.4241,
+      "step": 662
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.850706283487579e-07,
+      "loss": 0.42,
+      "step": 663
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.850462737457379e-07,
+      "loss": 0.4629,
+      "step": 664
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.85021919142718e-07,
+      "loss": 0.3643,
+      "step": 665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.84997564539698e-07,
+      "loss": 0.4306,
+      "step": 666
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.84973209936678e-07,
+      "loss": 0.47,
+      "step": 667
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.84948855333658e-07,
+      "loss": 0.3704,
+      "step": 668
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.849245007306381e-07,
+      "loss": 0.4427,
+      "step": 669
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.849001461276181e-07,
+      "loss": 0.4045,
+      "step": 670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.848757915245982e-07,
+      "loss": 0.3968,
+      "step": 671
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.848514369215781e-07,
+      "loss": 0.4535,
+      "step": 672
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.848270823185582e-07,
+      "loss": 0.3838,
+      "step": 673
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.848027277155382e-07,
+      "loss": 0.3389,
+      "step": 674
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.847783731125183e-07,
+      "loss": 0.3922,
+      "step": 675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.847540185094982e-07,
+      "loss": 0.351,
+      "step": 676
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.847296639064783e-07,
+      "loss": 0.3822,
+      "step": 677
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.847053093034583e-07,
+      "loss": 0.4664,
+      "step": 678
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.846809547004384e-07,
+      "loss": 0.4047,
+      "step": 679
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.846566000974184e-07,
+      "loss": 0.3536,
+      "step": 680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.846322454943984e-07,
+      "loss": 0.503,
+      "step": 681
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.846078908913784e-07,
+      "loss": 0.3394,
+      "step": 682
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.845835362883585e-07,
+      "loss": 0.3571,
+      "step": 683
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.845591816853385e-07,
+      "loss": 0.3432,
+      "step": 684
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.845348270823185e-07,
+      "loss": 0.3722,
+      "step": 685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.845104724792985e-07,
+      "loss": 0.3819,
+      "step": 686
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.844861178762786e-07,
+      "loss": 0.4576,
+      "step": 687
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.844617632732586e-07,
+      "loss": 0.3537,
+      "step": 688
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.844374086702387e-07,
+      "loss": 0.3716,
+      "step": 689
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.844130540672186e-07,
+      "loss": 0.3325,
+      "step": 690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.843886994641987e-07,
+      "loss": 0.3969,
+      "step": 691
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.843643448611787e-07,
+      "loss": 0.3787,
+      "step": 692
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.843399902581588e-07,
+      "loss": 0.3495,
+      "step": 693
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.843156356551387e-07,
+      "loss": 0.3307,
+      "step": 694
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842912810521188e-07,
+      "loss": 0.4196,
+      "step": 695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842669264490988e-07,
+      "loss": 0.4074,
+      "step": 696
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842425718460789e-07,
+      "loss": 0.4115,
+      "step": 697
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842182172430589e-07,
+      "loss": 0.3237,
+      "step": 698
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.841938626400389e-07,
+      "loss": 0.3306,
+      "step": 699
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.841695080370189e-07,
+      "loss": 0.4045,
+      "step": 700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.84145153433999e-07,
+      "loss": 0.535,
+      "step": 701
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.84120798830979e-07,
+      "loss": 0.3248,
+      "step": 702
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.840964442279591e-07,
+      "loss": 0.5549,
+      "step": 703
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.840720896249391e-07,
+      "loss": 0.4131,
+      "step": 704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.840477350219192e-07,
+      "loss": 0.3345,
+      "step": 705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.840233804188991e-07,
+      "loss": 0.3731,
+      "step": 706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.839990258158792e-07,
+      "loss": 0.3634,
+      "step": 707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.839746712128592e-07,
+      "loss": 0.379,
+      "step": 708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.839503166098393e-07,
+      "loss": 0.3878,
+      "step": 709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.839259620068193e-07,
+      "loss": 0.3415,
+      "step": 710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.839016074037993e-07,
+      "loss": 0.4223,
+      "step": 711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.838772528007793e-07,
+      "loss": 0.3466,
+      "step": 712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.838528981977594e-07,
+      "loss": 0.3596,
+      "step": 713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.838285435947394e-07,
+      "loss": 0.348,
+      "step": 714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.838041889917194e-07,
+      "loss": 0.3381,
+      "step": 715
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.837798343886994e-07,
+      "loss": 0.3373,
+      "step": 716
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.837554797856795e-07,
+      "loss": 0.3329,
+      "step": 717
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.837311251826595e-07,
+      "loss": 0.2872,
+      "step": 718
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.837067705796396e-07,
+      "loss": 0.4766,
+      "step": 719
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.836824159766195e-07,
+      "loss": 0.4278,
+      "step": 720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.836580613735996e-07,
+      "loss": 0.3071,
+      "step": 721
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.836337067705796e-07,
+      "loss": 0.3313,
+      "step": 722
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.836093521675597e-07,
+      "loss": 0.4384,
+      "step": 723
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.835849975645396e-07,
+      "loss": 0.3849,
+      "step": 724
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.835606429615197e-07,
+      "loss": 0.3857,
+      "step": 725
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.835362883584997e-07,
+      "loss": 0.4342,
+      "step": 726
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.835119337554798e-07,
+      "loss": 0.3072,
+      "step": 727
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.834875791524598e-07,
+      "loss": 0.3094,
+      "step": 728
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.834632245494398e-07,
+      "loss": 0.3659,
+      "step": 729
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.834388699464198e-07,
+      "loss": 0.3969,
+      "step": 730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.834145153433999e-07,
+      "loss": 0.2681,
+      "step": 731
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.833901607403799e-07,
+      "loss": 0.3395,
+      "step": 732
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.833658061373599e-07,
+      "loss": 0.3716,
+      "step": 733
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.833414515343399e-07,
+      "loss": 0.3501,
+      "step": 734
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.8331709693132e-07,
+      "loss": 0.2936,
+      "step": 735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.832927423283e-07,
+      "loss": 0.3381,
+      "step": 736
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.832683877252801e-07,
+      "loss": 0.2966,
+      "step": 737
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.8324403312226e-07,
+      "loss": 0.3321,
+      "step": 738
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.832196785192401e-07,
+      "loss": 0.2857,
+      "step": 739
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.831953239162201e-07,
+      "loss": 0.3237,
+      "step": 740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.831709693132002e-07,
+      "loss": 0.3005,
+      "step": 741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.831466147101803e-07,
+      "loss": 0.318,
+      "step": 742
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.831222601071602e-07,
+      "loss": 0.345,
+      "step": 743
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.830979055041402e-07,
+      "loss": 0.3444,
+      "step": 744
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.830735509011203e-07,
+      "loss": 0.4955,
+      "step": 745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.830491962981004e-07,
+      "loss": 0.2912,
+      "step": 746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.830248416950803e-07,
+      "loss": 0.3212,
+      "step": 747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.830004870920603e-07,
+      "loss": 0.3567,
+      "step": 748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.829761324890404e-07,
+      "loss": 0.3486,
+      "step": 749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.829517778860205e-07,
+      "loss": 0.2487,
+      "step": 750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.829274232830005e-07,
+      "loss": 0.2612,
+      "step": 751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.829030686799804e-07,
+      "loss": 0.3416,
+      "step": 752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.828787140769605e-07,
+      "loss": 0.2933,
+      "step": 753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.828543594739406e-07,
+      "loss": 0.3159,
+      "step": 754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.828300048709206e-07,
+      "loss": 0.306,
+      "step": 755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.828056502679005e-07,
+      "loss": 0.2928,
+      "step": 756
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.827812956648806e-07,
+      "loss": 0.336,
+      "step": 757
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.827569410618607e-07,
+      "loss": 0.3069,
+      "step": 758
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.827325864588407e-07,
+      "loss": 0.3193,
+      "step": 759
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.827082318558208e-07,
+      "loss": 0.5077,
+      "step": 760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.826838772528007e-07,
+      "loss": 0.2965,
+      "step": 761
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.826595226497808e-07,
+      "loss": 0.3117,
+      "step": 762
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.826351680467608e-07,
+      "loss": 0.3033,
+      "step": 763
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.826108134437409e-07,
+      "loss": 0.2766,
+      "step": 764
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.825864588407209e-07,
+      "loss": 0.3191,
+      "step": 765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 4.825621042377009e-07,
+      "loss": 0.3682,
+      "step": 766
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.825377496346809e-07,
+      "loss": 0.3389,
+      "step": 767
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.82513395031661e-07,
+      "loss": 0.2774,
+      "step": 768
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.82489040428641e-07,
+      "loss": 0.2858,
+      "step": 769
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.82464685825621e-07,
+      "loss": 0.2991,
+      "step": 770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.82440331222601e-07,
+      "loss": 0.388,
+      "step": 771
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.824159766195811e-07,
+      "loss": 0.4275,
+      "step": 772
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.823916220165611e-07,
+      "loss": 0.2906,
+      "step": 773
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.823672674135412e-07,
+      "loss": 0.2981,
+      "step": 774
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.823429128105211e-07,
+      "loss": 0.3318,
+      "step": 775
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 4.823185582075012e-07,
+      "loss": 0.3222,
+      "step": 776
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.822942036044812e-07,
+      "loss": 0.2516,
+      "step": 777
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.822698490014613e-07,
+      "loss": 0.2828,
+      "step": 778
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.822454943984413e-07,
+      "loss": 0.2535,
+      "step": 779
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.822211397954214e-07,
+      "loss": 0.31,
+      "step": 780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.821967851924014e-07,
+      "loss": 0.4132,
+      "step": 781
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.821724305893814e-07,
+      "loss": 0.2537,
+      "step": 782
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.821480759863614e-07,
+      "loss": 0.3656,
+      "step": 783
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.821237213833415e-07,
+      "loss": 0.319,
+      "step": 784
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.820993667803215e-07,
+      "loss": 0.3567,
+      "step": 785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.820750121773015e-07,
+      "loss": 0.284,
+      "step": 786
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 4.820506575742815e-07,
+      "loss": 0.2284,
+      "step": 787
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.820263029712616e-07,
+      "loss": 0.2672,
+      "step": 788
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.820019483682416e-07,
+      "loss": 0.2839,
+      "step": 789
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.819775937652217e-07,
+      "loss": 0.2761,
+      "step": 790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.819532391622016e-07,
+      "loss": 0.4403,
+      "step": 791
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.819288845591817e-07,
+      "loss": 0.3239,
+      "step": 792
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.819045299561617e-07,
+      "loss": 0.4561,
+      "step": 793
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.818801753531418e-07,
+      "loss": 0.2777,
+      "step": 794
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.818558207501217e-07,
+      "loss": 0.2565,
+      "step": 795
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.818314661471018e-07,
+      "loss": 0.3625,
+      "step": 796
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 4.818071115440818e-07,
+      "loss": 0.3054,
+      "step": 797
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.817827569410619e-07,
+      "loss": 0.2068,
+      "step": 798
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.817584023380419e-07,
+      "loss": 0.4709,
+      "step": 799
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.817340477350219e-07,
+      "loss": 0.2236,
+      "step": 800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.817096931320019e-07,
+      "loss": 0.2308,
+      "step": 801
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.81685338528982e-07,
+      "loss": 0.2264,
+      "step": 802
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.81660983925962e-07,
+      "loss": 0.6715,
+      "step": 803
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.816366293229421e-07,
+      "loss": 0.3524,
+      "step": 804
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.81612274719922e-07,
+      "loss": 0.2413,
+      "step": 805
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.815879201169021e-07,
+      "loss": 0.2601,
+      "step": 806
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 4.815635655138821e-07,
+      "loss": 0.2327,
+      "step": 807
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.815392109108622e-07,
+      "loss": 0.2442,
+      "step": 808
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.815148563078421e-07,
+      "loss": 0.2803,
+      "step": 809
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.814905017048222e-07,
+      "loss": 0.3642,
+      "step": 810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.814661471018022e-07,
+      "loss": 0.225,
+      "step": 811
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.814417924987823e-07,
+      "loss": 0.3098,
+      "step": 812
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.814174378957623e-07,
+      "loss": 0.2184,
+      "step": 813
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.813930832927423e-07,
+      "loss": 0.322,
+      "step": 814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.813687286897223e-07,
+      "loss": 0.2457,
+      "step": 815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.813443740867024e-07,
+      "loss": 0.2971,
+      "step": 816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.813200194836824e-07,
+      "loss": 0.2198,
+      "step": 817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 4.812956648806624e-07,
+      "loss": 0.2386,
+      "step": 818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.812713102776424e-07,
+      "loss": 0.2565,
+      "step": 819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.812469556746225e-07,
+      "loss": 0.4293,
+      "step": 820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.812226010716025e-07,
+      "loss": 0.2521,
+      "step": 821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.811982464685826e-07,
+      "loss": 0.1781,
+      "step": 822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.811738918655625e-07,
+      "loss": 0.3251,
+      "step": 823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.811495372625426e-07,
+      "loss": 0.243,
+      "step": 824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.811251826595226e-07,
+      "loss": 0.2514,
+      "step": 825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.811008280565027e-07,
+      "loss": 0.2527,
+      "step": 826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.810764734534826e-07,
+      "loss": 0.2886,
+      "step": 827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 4.810521188504627e-07,
+      "loss": 0.2209,
+      "step": 828
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.810277642474427e-07,
+      "loss": 0.2138,
+      "step": 829
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.810034096444228e-07,
+      "loss": 0.2208,
+      "step": 830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.809790550414028e-07,
+      "loss": 0.241,
+      "step": 831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.809547004383828e-07,
+      "loss": 0.2735,
+      "step": 832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.809303458353628e-07,
+      "loss": 0.2318,
+      "step": 833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.809059912323429e-07,
+      "loss": 0.2884,
+      "step": 834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.808816366293229e-07,
+      "loss": 0.2953,
+      "step": 835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.80857282026303e-07,
+      "loss": 0.1889,
+      "step": 836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.808329274232829e-07,
+      "loss": 0.2735,
+      "step": 837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 4.80808572820263e-07,
+      "loss": 0.2277,
+      "step": 838
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.80784218217243e-07,
+      "loss": 0.3459,
+      "step": 839
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.807598636142231e-07,
+      "loss": 0.1947,
+      "step": 840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.80735509011203e-07,
+      "loss": 0.2309,
+      "step": 841
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.807111544081831e-07,
+      "loss": 0.275,
+      "step": 842
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.806867998051631e-07,
+      "loss": 0.2777,
+      "step": 843
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.806624452021432e-07,
+      "loss": 0.1923,
+      "step": 844
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.806380905991232e-07,
+      "loss": 0.1788,
+      "step": 845
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.806137359961032e-07,
+      "loss": 0.2951,
+      "step": 846
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.805893813930832e-07,
+      "loss": 0.1841,
+      "step": 847
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 4.805650267900633e-07,
+      "loss": 0.2798,
+      "step": 848
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.805406721870433e-07,
+      "loss": 0.2801,
+      "step": 849
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.805163175840233e-07,
+      "loss": 0.2359,
+      "step": 850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.804919629810033e-07,
+      "loss": 0.2097,
+      "step": 851
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.804676083779834e-07,
+      "loss": 0.2141,
+      "step": 852
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.804432537749634e-07,
+      "loss": 0.2037,
+      "step": 853
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.804188991719435e-07,
+      "loss": 0.1946,
+      "step": 854
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.803945445689235e-07,
+      "loss": 0.2117,
+      "step": 855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.803701899659036e-07,
+      "loss": 0.2773,
+      "step": 856
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.803458353628835e-07,
+      "loss": 0.231,
+      "step": 857
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.803214807598636e-07,
+      "loss": 0.261,
+      "step": 858
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 4.802971261568436e-07,
+      "loss": 0.1961,
+      "step": 859
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.802727715538237e-07,
+      "loss": 0.3588,
+      "step": 860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.802484169508037e-07,
+      "loss": 0.2586,
+      "step": 861
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.802240623477837e-07,
+      "loss": 0.2333,
+      "step": 862
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.801997077447637e-07,
+      "loss": 0.3878,
+      "step": 863
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.801753531417438e-07,
+      "loss": 0.1825,
+      "step": 864
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.801509985387238e-07,
+      "loss": 0.2392,
+      "step": 865
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.801266439357038e-07,
+      "loss": 0.1926,
+      "step": 866
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.801022893326838e-07,
+      "loss": 0.1419,
+      "step": 867
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.800779347296639e-07,
+      "loss": 0.1566,
+      "step": 868
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.80053580126644e-07,
+      "loss": 0.1918,
+      "step": 869
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.80029225523624e-07,
+      "loss": 0.2523,
+      "step": 870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.800048709206039e-07,
+      "loss": 0.2322,
+      "step": 871
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.79980516317584e-07,
+      "loss": 0.2005,
+      "step": 872
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.79956161714564e-07,
+      "loss": 0.1822,
+      "step": 873
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.799318071115441e-07,
+      "loss": 0.1696,
+      "step": 874
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.799074525085242e-07,
+      "loss": 0.1766,
+      "step": 875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.798830979055041e-07,
+      "loss": 0.1783,
+      "step": 876
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.798587433024842e-07,
+      "loss": 0.1739,
+      "step": 877
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.798343886994642e-07,
+      "loss": 0.354,
+      "step": 878
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 4.798100340964443e-07,
+      "loss": 0.218,
+      "step": 879
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.797856794934242e-07,
+      "loss": 0.1672,
+      "step": 880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.797613248904043e-07,
+      "loss": 0.1956,
+      "step": 881
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.797369702873843e-07,
+      "loss": 0.2507,
+      "step": 882
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.797126156843644e-07,
+      "loss": 0.1741,
+      "step": 883
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.796882610813444e-07,
+      "loss": 0.1434,
+      "step": 884
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.796639064783244e-07,
+      "loss": 0.2371,
+      "step": 885
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.796395518753044e-07,
+      "loss": 0.185,
+      "step": 886
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.796151972722845e-07,
+      "loss": 0.3736,
+      "step": 887
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.795908426692645e-07,
+      "loss": 0.1535,
+      "step": 888
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.795664880662445e-07,
+      "loss": 0.1834,
+      "step": 889
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.795421334632245e-07,
+      "loss": 0.1799,
+      "step": 890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.795177788602046e-07,
+      "loss": 0.1808,
+      "step": 891
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.794934242571846e-07,
+      "loss": 0.1767,
+      "step": 892
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.794690696541647e-07,
+      "loss": 0.4135,
+      "step": 893
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.794447150511446e-07,
+      "loss": 0.1651,
+      "step": 894
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.794203604481247e-07,
+      "loss": 0.2266,
+      "step": 895
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.793960058451047e-07,
+      "loss": 0.6145,
+      "step": 896
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.793716512420848e-07,
+      "loss": 0.1528,
+      "step": 897
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.793472966390647e-07,
+      "loss": 0.1596,
+      "step": 898
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.793229420360448e-07,
+      "loss": 0.4896,
+      "step": 899
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 4.792985874330248e-07,
+      "loss": 0.1783,
+      "step": 900
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.792742328300049e-07,
+      "loss": 0.1806,
+      "step": 901
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.792498782269849e-07,
+      "loss": 0.2785,
+      "step": 902
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.792255236239649e-07,
+      "loss": 0.1698,
+      "step": 903
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.792011690209449e-07,
+      "loss": 0.2038,
+      "step": 904
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.79176814417925e-07,
+      "loss": 0.1359,
+      "step": 905
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.79152459814905e-07,
+      "loss": 0.2509,
+      "step": 906
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.791281052118851e-07,
+      "loss": 0.413,
+      "step": 907
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.79103750608865e-07,
+      "loss": 0.4448,
+      "step": 908
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.790793960058451e-07,
+      "loss": 0.126,
+      "step": 909
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.790550414028251e-07,
+      "loss": 0.1803,
+      "step": 910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.790306867998052e-07,
+      "loss": 0.1848,
+      "step": 911
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.790063321967851e-07,
+      "loss": 0.171,
+      "step": 912
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.789819775937652e-07,
+      "loss": 0.5808,
+      "step": 913
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.789576229907452e-07,
+      "loss": 0.447,
+      "step": 914
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.789332683877253e-07,
+      "loss": 0.1195,
+      "step": 915
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.789089137847053e-07,
+      "loss": 0.1914,
+      "step": 916
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.788845591816853e-07,
+      "loss": 0.4621,
+      "step": 917
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.788602045786653e-07,
+      "loss": 0.2153,
+      "step": 918
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.788358499756454e-07,
+      "loss": 0.1938,
+      "step": 919
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.788114953726254e-07,
+      "loss": 0.1455,
+      "step": 920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.787871407696054e-07,
+      "loss": 0.1338,
+      "step": 921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.787627861665854e-07,
+      "loss": 0.2395,
+      "step": 922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.787384315635655e-07,
+      "loss": 0.1469,
+      "step": 923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.787140769605455e-07,
+      "loss": 0.179,
+      "step": 924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.786897223575256e-07,
+      "loss": 0.1621,
+      "step": 925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.786653677545055e-07,
+      "loss": 0.1237,
+      "step": 926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.786410131514856e-07,
+      "loss": 0.1497,
+      "step": 927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.786166585484656e-07,
+      "loss": 0.2844,
+      "step": 928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.785923039454457e-07,
+      "loss": 0.1523,
+      "step": 929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.785679493424256e-07,
+      "loss": 0.6805,
+      "step": 930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.785435947394058e-07,
+      "loss": 0.1547,
+      "step": 931
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.785192401363858e-07,
+      "loss": 0.2732,
+      "step": 932
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.784948855333658e-07,
+      "loss": 0.1519,
+      "step": 933
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.784705309303458e-07,
+      "loss": 0.1341,
+      "step": 934
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.784461763273259e-07,
+      "loss": 0.1798,
+      "step": 935
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.784218217243059e-07,
+      "loss": 0.2526,
+      "step": 936
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.783974671212859e-07,
+      "loss": 0.492,
+      "step": 937
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.783731125182659e-07,
+      "loss": 0.1329,
+      "step": 938
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.78348757915246e-07,
+      "loss": 0.1845,
+      "step": 939
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.78324403312226e-07,
+      "loss": 0.1305,
+      "step": 940
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.783000487092061e-07,
+      "loss": 0.2663,
+      "step": 941
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.78275694106186e-07,
+      "loss": 0.2202,
+      "step": 942
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.782513395031661e-07,
+      "loss": 0.1535,
+      "step": 943
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.782269849001461e-07,
+      "loss": 0.2988,
+      "step": 944
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.782026302971262e-07,
+      "loss": 0.1314,
+      "step": 945
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.781782756941062e-07,
+      "loss": 0.1918,
+      "step": 946
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.781539210910862e-07,
+      "loss": 0.1405,
+      "step": 947
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.781295664880662e-07,
+      "loss": 0.1436,
+      "step": 948
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.781052118850463e-07,
+      "loss": 0.1458,
+      "step": 949
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.780808572820263e-07,
+      "loss": 0.1768,
+      "step": 950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 4.780565026790063e-07,
+      "loss": 0.1681,
+      "step": 951
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.780321480759863e-07,
+      "loss": 0.1568,
+      "step": 952
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.780077934729664e-07,
+      "loss": 0.1185,
+      "step": 953
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.779834388699464e-07,
+      "loss": 0.1091,
+      "step": 954
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.779590842669265e-07,
+      "loss": 0.1153,
+      "step": 955
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.779347296639064e-07,
+      "loss": 0.1416,
+      "step": 956
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.779103750608865e-07,
+      "loss": 0.1857,
+      "step": 957
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.778860204578665e-07,
+      "loss": 0.1241,
+      "step": 958
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.778616658548466e-07,
+      "loss": 0.295,
+      "step": 959
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.778373112518265e-07,
+      "loss": 0.2025,
+      "step": 960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.778129566488066e-07,
+      "loss": 0.1102,
+      "step": 961
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.777886020457866e-07,
+      "loss": 0.1555,
+      "step": 962
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.777642474427667e-07,
+      "loss": 0.1146,
+      "step": 963
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.777398928397467e-07,
+      "loss": 0.157,
+      "step": 964
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.777155382367267e-07,
+      "loss": 0.1162,
+      "step": 965
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.776911836337067e-07,
+      "loss": 0.1003,
+      "step": 966
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.776668290306868e-07,
+      "loss": 0.1242,
+      "step": 967
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.776424744276668e-07,
+      "loss": 0.1555,
+      "step": 968
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.776181198246469e-07,
+      "loss": 0.1542,
+      "step": 969
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.775937652216268e-07,
+      "loss": 0.1078,
+      "step": 970
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.775694106186069e-07,
+      "loss": 0.1409,
+      "step": 971
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 4.775450560155869e-07,
+      "loss": 0.1322,
+      "step": 972
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.77520701412567e-07,
+      "loss": 0.1055,
+      "step": 973
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.774963468095469e-07,
+      "loss": 0.1548,
+      "step": 974
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.77471992206527e-07,
+      "loss": 0.1606,
+      "step": 975
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.77447637603507e-07,
+      "loss": 0.1367,
+      "step": 976
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.774232830004871e-07,
+      "loss": 0.1948,
+      "step": 977
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.773989283974671e-07,
+      "loss": 0.3368,
+      "step": 978
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.773745737944471e-07,
+      "loss": 0.1029,
+      "step": 979
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.773502191914271e-07,
+      "loss": 0.0983,
+      "step": 980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.773258645884072e-07,
+      "loss": 0.1252,
+      "step": 981
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.773015099853872e-07,
+      "loss": 0.0989,
+      "step": 982
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.772771553823672e-07,
+      "loss": 0.2038,
+      "step": 983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.772528007793472e-07,
+      "loss": 0.1371,
+      "step": 984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.772284461763273e-07,
+      "loss": 0.1377,
+      "step": 985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.772040915733073e-07,
+      "loss": 0.1449,
+      "step": 986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.771797369702874e-07,
+      "loss": 0.1178,
+      "step": 987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.771553823672673e-07,
+      "loss": 0.1043,
+      "step": 988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.771310277642474e-07,
+      "loss": 0.1149,
+      "step": 989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.771066731612274e-07,
+      "loss": 0.1071,
+      "step": 990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.770823185582075e-07,
+      "loss": 0.3901,
+      "step": 991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.770579639551874e-07,
+      "loss": 0.1083,
+      "step": 992
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.770336093521675e-07,
+      "loss": 0.1162,
+      "step": 993
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.770092547491475e-07,
+      "loss": 0.1207,
+      "step": 994
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.769849001461276e-07,
+      "loss": 0.2434,
+      "step": 995
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.769605455431076e-07,
+      "loss": 0.1281,
+      "step": 996
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.769361909400876e-07,
+      "loss": 0.4875,
+      "step": 997
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.769118363370676e-07,
+      "loss": 0.1412,
+      "step": 998
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.768874817340477e-07,
+      "loss": 0.1013,
+      "step": 999
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.768631271310277e-07,
+      "loss": 0.0982,
+      "step": 1000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.768387725280078e-07,
+      "loss": 0.5571,
+      "step": 1001
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.768144179249878e-07,
+      "loss": 0.682,
+      "step": 1002
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.767900633219678e-07,
+      "loss": 0.1383,
+      "step": 1003
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.7676570871894785e-07,
+      "loss": 0.1046,
+      "step": 1004
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.7674135411592785e-07,
+      "loss": 0.0892,
+      "step": 1005
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.767169995129079e-07,
+      "loss": 0.1485,
+      "step": 1006
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.766926449098879e-07,
+      "loss": 0.1222,
+      "step": 1007
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.76668290306868e-07,
+      "loss": 0.1237,
+      "step": 1008
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.7664393570384805e-07,
+      "loss": 0.2931,
+      "step": 1009
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.7661958110082805e-07,
+      "loss": 0.1133,
+      "step": 1010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.765952264978081e-07,
+      "loss": 0.1337,
+      "step": 1011
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.765708718947881e-07,
+      "loss": 0.1127,
+      "step": 1012
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.7654651729176816e-07,
+      "loss": 0.1053,
+      "step": 1013
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7652216268874815e-07,
+      "loss": 0.1144,
+      "step": 1014
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.764978080857282e-07,
+      "loss": 0.1198,
+      "step": 1015
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7647345348270826e-07,
+      "loss": 0.1625,
+      "step": 1016
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7644909887968826e-07,
+      "loss": 0.0968,
+      "step": 1017
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.764247442766683e-07,
+      "loss": 0.095,
+      "step": 1018
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.764003896736483e-07,
+      "loss": 0.1052,
+      "step": 1019
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7637603507062836e-07,
+      "loss": 0.1692,
+      "step": 1020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.7635168046760836e-07,
+      "loss": 0.1077,
+      "step": 1021
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.763273258645884e-07,
+      "loss": 0.0932,
+      "step": 1022
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.763029712615684e-07,
+      "loss": 0.0822,
+      "step": 1023
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.7627861665854846e-07,
+      "loss": 0.0886,
+      "step": 1024
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.762542620555285e-07,
+      "loss": 0.113,
+      "step": 1025
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.762299074525085e-07,
+      "loss": 0.1615,
+      "step": 1026
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.7620555284948856e-07,
+      "loss": 0.0945,
+      "step": 1027
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.7618119824646856e-07,
+      "loss": 0.0973,
+      "step": 1028
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.761568436434486e-07,
+      "loss": 0.0759,
+      "step": 1029
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.988818667963053,
+      "eval_loss": 0.1271853744983673,
+      "eval_runtime": 2.8038,
+      "eval_samples_per_second": 733.656,
+      "eval_steps_per_second": 11.77,
+      "step": 1029
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.761324890404286e-07,
+      "loss": 0.1134,
+      "step": 1030
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.7610813443740866e-07,
+      "loss": 0.0798,
+      "step": 1031
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.760837798343887e-07,
+      "loss": 0.2073,
+      "step": 1032
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.760594252313687e-07,
+      "loss": 0.4187,
+      "step": 1033
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.7603507062834876e-07,
+      "loss": 0.3288,
+      "step": 1034
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.7601071602532876e-07,
+      "loss": 0.0819,
+      "step": 1035
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.759863614223088e-07,
+      "loss": 0.4228,
+      "step": 1036
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.759620068192888e-07,
+      "loss": 0.1472,
+      "step": 1037
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.7593765221626886e-07,
+      "loss": 0.0782,
+      "step": 1038
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.7591329761324886e-07,
+      "loss": 0.1031,
+      "step": 1039
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.758889430102289e-07,
+      "loss": 1.072,
+      "step": 1040
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.7586458840720897e-07,
+      "loss": 0.0798,
+      "step": 1041
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.7584023380418896e-07,
+      "loss": 0.0819,
+      "step": 1042
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.75815879201169e-07,
+      "loss": 0.0908,
+      "step": 1043
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 4.75791524598149e-07,
+      "loss": 0.1244,
+      "step": 1044
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.7576716999512907e-07,
+      "loss": 0.2462,
+      "step": 1045
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.7574281539210907e-07,
+      "loss": 0.1656,
+      "step": 1046
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.757184607890891e-07,
+      "loss": 0.0906,
+      "step": 1047
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.7569410618606917e-07,
+      "loss": 0.1075,
+      "step": 1048
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.7566975158304917e-07,
+      "loss": 0.7839,
+      "step": 1049
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.756453969800292e-07,
+      "loss": 0.0786,
+      "step": 1050
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.756210423770092e-07,
+      "loss": 0.17,
+      "step": 1051
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.7559668777398927e-07,
+      "loss": 0.0985,
+      "step": 1052
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.7557233317096927e-07,
+      "loss": 0.0958,
+      "step": 1053
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.755479785679493e-07,
+      "loss": 0.1094,
+      "step": 1054
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.755236239649293e-07,
+      "loss": 0.0902,
+      "step": 1055
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.7549926936190937e-07,
+      "loss": 0.115,
+      "step": 1056
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.754749147588894e-07,
+      "loss": 0.109,
+      "step": 1057
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.754505601558694e-07,
+      "loss": 0.0878,
+      "step": 1058
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.7542620555284947e-07,
+      "loss": 0.0734,
+      "step": 1059
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.7540185094982947e-07,
+      "loss": 0.1664,
+      "step": 1060
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.753774963468095e-07,
+      "loss": 0.0965,
+      "step": 1061
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.753531417437895e-07,
+      "loss": 0.0668,
+      "step": 1062
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.7532878714076957e-07,
+      "loss": 0.1149,
+      "step": 1063
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.753044325377496e-07,
+      "loss": 0.1218,
+      "step": 1064
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 4.752800779347296e-07,
+      "loss": 0.0641,
+      "step": 1065
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.752557233317097e-07,
+      "loss": 0.094,
+      "step": 1066
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.7523136872868967e-07,
+      "loss": 0.0754,
+      "step": 1067
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.752070141256697e-07,
+      "loss": 0.0999,
+      "step": 1068
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.751826595226497e-07,
+      "loss": 0.4159,
+      "step": 1069
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.751583049196298e-07,
+      "loss": 0.0836,
+      "step": 1070
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.751339503166098e-07,
+      "loss": 0.0798,
+      "step": 1071
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.751095957135898e-07,
+      "loss": 0.0675,
+      "step": 1072
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.750852411105699e-07,
+      "loss": 0.1481,
+      "step": 1073
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.750608865075499e-07,
+      "loss": 0.1081,
+      "step": 1074
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.7503653190452993e-07,
+      "loss": 0.1056,
+      "step": 1075
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.750121773015099e-07,
+      "loss": 0.0736,
+      "step": 1076
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7498782269849e-07,
+      "loss": 0.097,
+      "step": 1077
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7496346809547e-07,
+      "loss": 0.0916,
+      "step": 1078
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7493911349245003e-07,
+      "loss": 0.0713,
+      "step": 1079
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.749147588894301e-07,
+      "loss": 0.0633,
+      "step": 1080
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.748904042864101e-07,
+      "loss": 0.0864,
+      "step": 1081
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7486604968339013e-07,
+      "loss": 0.064,
+      "step": 1082
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7484169508037013e-07,
+      "loss": 0.0722,
+      "step": 1083
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.748173404773502e-07,
+      "loss": 0.1959,
+      "step": 1084
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.7479298587433023e-07,
+      "loss": 0.0665,
+      "step": 1085
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.747686312713103e-07,
+      "loss": 0.0758,
+      "step": 1086
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7474427666829034e-07,
+      "loss": 0.0721,
+      "step": 1087
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7471992206527033e-07,
+      "loss": 0.0797,
+      "step": 1088
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.746955674622504e-07,
+      "loss": 0.3567,
+      "step": 1089
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.746712128592304e-07,
+      "loss": 0.0908,
+      "step": 1090
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7464685825621044e-07,
+      "loss": 0.3269,
+      "step": 1091
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7462250365319043e-07,
+      "loss": 0.1046,
+      "step": 1092
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.745981490501705e-07,
+      "loss": 0.1529,
+      "step": 1093
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.745737944471505e-07,
+      "loss": 0.0821,
+      "step": 1094
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.7454943984413054e-07,
+      "loss": 0.0966,
+      "step": 1095
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.745250852411106e-07,
+      "loss": 0.0579,
+      "step": 1096
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.745007306380906e-07,
+      "loss": 0.0643,
+      "step": 1097
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.7447637603507064e-07,
+      "loss": 0.0688,
+      "step": 1098
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.7445202143205064e-07,
+      "loss": 0.0731,
+      "step": 1099
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.744276668290307e-07,
+      "loss": 0.0663,
+      "step": 1100
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.744033122260107e-07,
+      "loss": 0.062,
+      "step": 1101
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.7437895762299074e-07,
+      "loss": 0.1218,
+      "step": 1102
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.743546030199708e-07,
+      "loss": 0.0881,
+      "step": 1103
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.743302484169508e-07,
+      "loss": 0.1375,
+      "step": 1104
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.7430589381393084e-07,
+      "loss": 0.0645,
+      "step": 1105
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.7428153921091084e-07,
+      "loss": 0.0673,
+      "step": 1106
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.742571846078909e-07,
+      "loss": 0.7849,
+      "step": 1107
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.742328300048709e-07,
+      "loss": 0.0536,
+      "step": 1108
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.7420847540185094e-07,
+      "loss": 0.1138,
+      "step": 1109
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.7418412079883094e-07,
+      "loss": 0.0551,
+      "step": 1110
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.74159766195811e-07,
+      "loss": 0.2023,
+      "step": 1111
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.7413541159279104e-07,
+      "loss": 0.0626,
+      "step": 1112
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.7411105698977104e-07,
+      "loss": 0.073,
+      "step": 1113
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.740867023867511e-07,
+      "loss": 0.0719,
+      "step": 1114
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.740623477837311e-07,
+      "loss": 0.1266,
+      "step": 1115
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 4.7403799318071114e-07,
+      "loss": 0.1014,
+      "step": 1116
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.7401363857769114e-07,
+      "loss": 0.0738,
+      "step": 1117
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.739892839746712e-07,
+      "loss": 0.0528,
+      "step": 1118
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.7396492937165125e-07,
+      "loss": 0.068,
+      "step": 1119
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.7394057476863125e-07,
+      "loss": 0.1352,
+      "step": 1120
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.739162201656113e-07,
+      "loss": 0.0995,
+      "step": 1121
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.738918655625913e-07,
+      "loss": 0.071,
+      "step": 1122
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.7386751095957135e-07,
+      "loss": 0.0613,
+      "step": 1123
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.7384315635655135e-07,
+      "loss": 0.0523,
+      "step": 1124
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.738188017535314e-07,
+      "loss": 0.0679,
+      "step": 1125
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.737944471505114e-07,
+      "loss": 0.0806,
+      "step": 1126
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.7377009254749145e-07,
+      "loss": 0.0543,
+      "step": 1127
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.737457379444715e-07,
+      "loss": 0.0584,
+      "step": 1128
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.737213833414515e-07,
+      "loss": 0.1078,
+      "step": 1129
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.7369702873843155e-07,
+      "loss": 0.067,
+      "step": 1130
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.7367267413541155e-07,
+      "loss": 0.0721,
+      "step": 1131
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.736483195323916e-07,
+      "loss": 0.2852,
+      "step": 1132
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.736239649293716e-07,
+      "loss": 0.4176,
+      "step": 1133
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.7359961032635165e-07,
+      "loss": 0.0597,
+      "step": 1134
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.735752557233317e-07,
+      "loss": 0.0928,
+      "step": 1135
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.735509011203117e-07,
+      "loss": 0.0519,
+      "step": 1136
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 4.7352654651729175e-07,
+      "loss": 0.0444,
+      "step": 1137
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.7350219191427175e-07,
+      "loss": 0.0595,
+      "step": 1138
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.734778373112518e-07,
+      "loss": 0.0622,
+      "step": 1139
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.734534827082318e-07,
+      "loss": 0.0759,
+      "step": 1140
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.7342912810521185e-07,
+      "loss": 0.0507,
+      "step": 1141
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.734047735021919e-07,
+      "loss": 0.0856,
+      "step": 1142
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.733804188991719e-07,
+      "loss": 0.0593,
+      "step": 1143
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.7335606429615196e-07,
+      "loss": 0.0597,
+      "step": 1144
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.7333170969313195e-07,
+      "loss": 0.0744,
+      "step": 1145
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.73307355090112e-07,
+      "loss": 0.0439,
+      "step": 1146
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.73283000487092e-07,
+      "loss": 0.1262,
+      "step": 1147
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.7325864588407206e-07,
+      "loss": 0.0615,
+      "step": 1148
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.7323429128105205e-07,
+      "loss": 0.0499,
+      "step": 1149
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.732099366780321e-07,
+      "loss": 0.0534,
+      "step": 1150
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.7318558207501216e-07,
+      "loss": 0.2083,
+      "step": 1151
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.7316122747199216e-07,
+      "loss": 0.0523,
+      "step": 1152
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.731368728689722e-07,
+      "loss": 0.0672,
+      "step": 1153
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.731125182659522e-07,
+      "loss": 0.5861,
+      "step": 1154
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.7308816366293226e-07,
+      "loss": 0.0643,
+      "step": 1155
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.7306380905991226e-07,
+      "loss": 0.0629,
+      "step": 1156
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.730394544568923e-07,
+      "loss": 0.0504,
+      "step": 1157
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.7301509985387236e-07,
+      "loss": 0.0545,
+      "step": 1158
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.7299074525085236e-07,
+      "loss": 0.0697,
+      "step": 1159
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.729663906478324e-07,
+      "loss": 0.0578,
+      "step": 1160
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.7294203604481246e-07,
+      "loss": 0.3832,
+      "step": 1161
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.729176814417925e-07,
+      "loss": 0.0523,
+      "step": 1162
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.728933268387725e-07,
+      "loss": 0.0531,
+      "step": 1163
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.7286897223575256e-07,
+      "loss": 0.0622,
+      "step": 1164
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.728446176327326e-07,
+      "loss": 0.0594,
+      "step": 1165
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.728202630297126e-07,
+      "loss": 0.0589,
+      "step": 1166
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 4.7279590842669267e-07,
+      "loss": 0.0546,
+      "step": 1167
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.7277155382367266e-07,
+      "loss": 0.0449,
+      "step": 1168
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.727471992206527e-07,
+      "loss": 0.0524,
+      "step": 1169
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.727228446176327e-07,
+      "loss": 0.0631,
+      "step": 1170
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.7269849001461277e-07,
+      "loss": 0.0971,
+      "step": 1171
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.7267413541159277e-07,
+      "loss": 0.0586,
+      "step": 1172
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.726497808085728e-07,
+      "loss": 0.0455,
+      "step": 1173
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.7262542620555287e-07,
+      "loss": 0.0522,
+      "step": 1174
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.7260107160253287e-07,
+      "loss": 0.0615,
+      "step": 1175
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.725767169995129e-07,
+      "loss": 0.0443,
+      "step": 1176
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.725523623964929e-07,
+      "loss": 0.0553,
+      "step": 1177
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.7252800779347297e-07,
+      "loss": 0.0648,
+      "step": 1178
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.7250365319045297e-07,
+      "loss": 0.1447,
+      "step": 1179
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.72479298587433e-07,
+      "loss": 0.0441,
+      "step": 1180
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.7245494398441307e-07,
+      "loss": 0.0355,
+      "step": 1181
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.7243058938139307e-07,
+      "loss": 0.0543,
+      "step": 1182
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.724062347783731e-07,
+      "loss": 0.0516,
+      "step": 1183
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.723818801753531e-07,
+      "loss": 0.0388,
+      "step": 1184
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.7235752557233317e-07,
+      "loss": 0.0524,
+      "step": 1185
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.7233317096931317e-07,
+      "loss": 0.0587,
+      "step": 1186
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.723088163662932e-07,
+      "loss": 0.0573,
+      "step": 1187
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 4.722844617632732e-07,
+      "loss": 0.0463,
+      "step": 1188
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.7226010716025327e-07,
+      "loss": 0.0985,
+      "step": 1189
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.722357525572333e-07,
+      "loss": 0.049,
+      "step": 1190
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.722113979542133e-07,
+      "loss": 0.1078,
+      "step": 1191
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.721870433511934e-07,
+      "loss": 0.2208,
+      "step": 1192
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.7216268874817337e-07,
+      "loss": 0.0541,
+      "step": 1193
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.721383341451534e-07,
+      "loss": 0.1434,
+      "step": 1194
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.721139795421334e-07,
+      "loss": 0.0542,
+      "step": 1195
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.720896249391135e-07,
+      "loss": 0.0438,
+      "step": 1196
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.7206527033609353e-07,
+      "loss": 0.0449,
+      "step": 1197
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.720409157330735e-07,
+      "loss": 0.0387,
+      "step": 1198
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.720165611300536e-07,
+      "loss": 0.0401,
+      "step": 1199
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.719922065270336e-07,
+      "loss": 0.0458,
+      "step": 1200
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.7196785192401363e-07,
+      "loss": 0.0415,
+      "step": 1201
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.719434973209936e-07,
+      "loss": 0.0538,
+      "step": 1202
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.719191427179737e-07,
+      "loss": 0.0496,
+      "step": 1203
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.718947881149537e-07,
+      "loss": 0.192,
+      "step": 1204
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.7187043351193373e-07,
+      "loss": 0.0544,
+      "step": 1205
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.718460789089138e-07,
+      "loss": 0.0532,
+      "step": 1206
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.718217243058938e-07,
+      "loss": 0.3312,
+      "step": 1207
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.7179736970287383e-07,
+      "loss": 0.1172,
+      "step": 1208
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 4.7177301509985383e-07,
+      "loss": 0.0391,
+      "step": 1209
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.717486604968339e-07,
+      "loss": 0.0689,
+      "step": 1210
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.717243058938139e-07,
+      "loss": 0.0383,
+      "step": 1211
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.7169995129079393e-07,
+      "loss": 0.06,
+      "step": 1212
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.71675596687774e-07,
+      "loss": 0.0334,
+      "step": 1213
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.71651242084754e-07,
+      "loss": 0.1308,
+      "step": 1214
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.7162688748173403e-07,
+      "loss": 0.0405,
+      "step": 1215
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.7160253287871403e-07,
+      "loss": 0.0605,
+      "step": 1216
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.715781782756941e-07,
+      "loss": 0.2454,
+      "step": 1217
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.715538236726741e-07,
+      "loss": 0.0605,
+      "step": 1218
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.7152946906965413e-07,
+      "loss": 0.038,
+      "step": 1219
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.7150511446663413e-07,
+      "loss": 0.0451,
+      "step": 1220
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.714807598636142e-07,
+      "loss": 0.0399,
+      "step": 1221
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.7145640526059424e-07,
+      "loss": 0.0407,
+      "step": 1222
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.7143205065757423e-07,
+      "loss": 0.1322,
+      "step": 1223
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.714076960545543e-07,
+      "loss": 0.1732,
+      "step": 1224
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.713833414515343e-07,
+      "loss": 0.0935,
+      "step": 1225
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.7135898684851434e-07,
+      "loss": 0.0478,
+      "step": 1226
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.7133463224549433e-07,
+      "loss": 0.0435,
+      "step": 1227
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.713102776424744e-07,
+      "loss": 0.0362,
+      "step": 1228
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 4.7128592303945444e-07,
+      "loss": 0.4473,
+      "step": 1229
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.7126156843643444e-07,
+      "loss": 0.8933,
+      "step": 1230
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.712372138334145e-07,
+      "loss": 0.5194,
+      "step": 1231
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.712128592303945e-07,
+      "loss": 0.0464,
+      "step": 1232
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.7118850462737454e-07,
+      "loss": 0.031,
+      "step": 1233
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.7116415002435454e-07,
+      "loss": 0.0686,
+      "step": 1234
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.711397954213346e-07,
+      "loss": 0.0579,
+      "step": 1235
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.711154408183146e-07,
+      "loss": 0.0416,
+      "step": 1236
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.7109108621529464e-07,
+      "loss": 0.4974,
+      "step": 1237
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.7106673161227474e-07,
+      "loss": 0.0589,
+      "step": 1238
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.7104237700925474e-07,
+      "loss": 0.1233,
+      "step": 1239
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.710180224062348e-07,
+      "loss": 0.0853,
+      "step": 1240
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.709936678032148e-07,
+      "loss": 0.039,
+      "step": 1241
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.7096931320019484e-07,
+      "loss": 0.2152,
+      "step": 1242
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.7094495859717484e-07,
+      "loss": 0.0812,
+      "step": 1243
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.709206039941549e-07,
+      "loss": 0.1226,
+      "step": 1244
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.7089624939113495e-07,
+      "loss": 0.0321,
+      "step": 1245
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.7087189478811495e-07,
+      "loss": 0.0309,
+      "step": 1246
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.70847540185095e-07,
+      "loss": 0.0524,
+      "step": 1247
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.70823185582075e-07,
+      "loss": 0.0464,
+      "step": 1248
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.7079883097905505e-07,
+      "loss": 0.0946,
+      "step": 1249
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.7077447637603505e-07,
+      "loss": 0.1517,
+      "step": 1250
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.707501217730151e-07,
+      "loss": 0.1057,
+      "step": 1251
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.7072576716999515e-07,
+      "loss": 0.0355,
+      "step": 1252
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.7070141256697515e-07,
+      "loss": 0.0342,
+      "step": 1253
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.706770579639552e-07,
+      "loss": 0.0435,
+      "step": 1254
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.706527033609352e-07,
+      "loss": 0.051,
+      "step": 1255
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.7062834875791525e-07,
+      "loss": 0.0344,
+      "step": 1256
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.7060399415489525e-07,
+      "loss": 0.0437,
+      "step": 1257
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.705796395518753e-07,
+      "loss": 0.0324,
+      "step": 1258
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.705552849488553e-07,
+      "loss": 0.8506,
+      "step": 1259
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 4.7053093034583535e-07,
+      "loss": 0.0382,
+      "step": 1260
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.705065757428154e-07,
+      "loss": 0.0368,
+      "step": 1261
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.704822211397954e-07,
+      "loss": 0.0737,
+      "step": 1262
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.7045786653677545e-07,
+      "loss": 0.0435,
+      "step": 1263
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.7043351193375545e-07,
+      "loss": 0.0419,
+      "step": 1264
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.704091573307355e-07,
+      "loss": 0.0394,
+      "step": 1265
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.703848027277155e-07,
+      "loss": 0.075,
+      "step": 1266
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.7036044812469555e-07,
+      "loss": 0.0696,
+      "step": 1267
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.703360935216756e-07,
+      "loss": 0.0369,
+      "step": 1268
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.703117389186556e-07,
+      "loss": 1.1684,
+      "step": 1269
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.7028738431563566e-07,
+      "loss": 0.0293,
+      "step": 1270
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.7026302971261565e-07,
+      "loss": 0.052,
+      "step": 1271
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.702386751095957e-07,
+      "loss": 0.0372,
+      "step": 1272
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.702143205065757e-07,
+      "loss": 0.2004,
+      "step": 1273
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.7018996590355576e-07,
+      "loss": 0.0417,
+      "step": 1274
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.7016561130053575e-07,
+      "loss": 0.0339,
+      "step": 1275
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.701412566975158e-07,
+      "loss": 0.0458,
+      "step": 1276
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.7011690209449586e-07,
+      "loss": 0.0343,
+      "step": 1277
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.7009254749147586e-07,
+      "loss": 0.0981,
+      "step": 1278
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.700681928884559e-07,
+      "loss": 0.0582,
+      "step": 1279
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.700438382854359e-07,
+      "loss": 0.0405,
+      "step": 1280
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 4.7001948368241596e-07,
+      "loss": 0.0813,
+      "step": 1281
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.6999512907939596e-07,
+      "loss": 0.0957,
+      "step": 1282
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.69970774476376e-07,
+      "loss": 0.0444,
+      "step": 1283
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.6994641987335606e-07,
+      "loss": 0.035,
+      "step": 1284
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.6992206527033606e-07,
+      "loss": 0.0291,
+      "step": 1285
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.698977106673161e-07,
+      "loss": 0.0654,
+      "step": 1286
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.698733560642961e-07,
+      "loss": 0.0399,
+      "step": 1287
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.6984900146127616e-07,
+      "loss": 0.0349,
+      "step": 1288
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.6982464685825616e-07,
+      "loss": 0.0841,
+      "step": 1289
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.698002922552362e-07,
+      "loss": 0.033,
+      "step": 1290
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.697759376522162e-07,
+      "loss": 0.039,
+      "step": 1291
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.6975158304919626e-07,
+      "loss": 0.0369,
+      "step": 1292
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.697272284461763e-07,
+      "loss": 0.0387,
+      "step": 1293
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.697028738431563e-07,
+      "loss": 0.0446,
+      "step": 1294
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.6967851924013636e-07,
+      "loss": 0.0312,
+      "step": 1295
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.6965416463711636e-07,
+      "loss": 0.1767,
+      "step": 1296
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.696298100340964e-07,
+      "loss": 0.0532,
+      "step": 1297
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.696054554310764e-07,
+      "loss": 0.03,
+      "step": 1298
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.6958110082805646e-07,
+      "loss": 0.3833,
+      "step": 1299
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.695567462250365e-07,
+      "loss": 0.0305,
+      "step": 1300
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 4.695323916220165e-07,
+      "loss": 0.0392,
+      "step": 1301
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.6950803701899657e-07,
+      "loss": 0.0535,
+      "step": 1302
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.6948368241597657e-07,
+      "loss": 0.0597,
+      "step": 1303
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.694593278129566e-07,
+      "loss": 0.0349,
+      "step": 1304
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.694349732099366e-07,
+      "loss": 0.0334,
+      "step": 1305
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.6941061860691667e-07,
+      "loss": 0.3313,
+      "step": 1306
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.693862640038967e-07,
+      "loss": 0.1155,
+      "step": 1307
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.693619094008767e-07,
+      "loss": 0.0705,
+      "step": 1308
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.6933755479785677e-07,
+      "loss": 0.037,
+      "step": 1309
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.6931320019483677e-07,
+      "loss": 0.0413,
+      "step": 1310
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.692888455918168e-07,
+      "loss": 1.0847,
+      "step": 1311
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.692644909887968e-07,
+      "loss": 0.0973,
+      "step": 1312
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.6924013638577687e-07,
+      "loss": 0.0323,
+      "step": 1313
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.692157817827569e-07,
+      "loss": 0.0393,
+      "step": 1314
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.6919142717973697e-07,
+      "loss": 0.0791,
+      "step": 1315
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.69167072576717e-07,
+      "loss": 0.0322,
+      "step": 1316
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.69142717973697e-07,
+      "loss": 0.0571,
+      "step": 1317
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.691183633706771e-07,
+      "loss": 0.031,
+      "step": 1318
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.690940087676571e-07,
+      "loss": 0.0408,
+      "step": 1319
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.690696541646371e-07,
+      "loss": 0.0346,
+      "step": 1320
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.690452995616171e-07,
+      "loss": 0.0273,
+      "step": 1321
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 4.690209449585972e-07,
+      "loss": 0.0247,
+      "step": 1322
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.6899659035557723e-07,
+      "loss": 0.6561,
+      "step": 1323
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.689722357525572e-07,
+      "loss": 0.0498,
+      "step": 1324
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.689478811495373e-07,
+      "loss": 0.2256,
+      "step": 1325
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.689235265465173e-07,
+      "loss": 0.0402,
+      "step": 1326
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.6889917194349733e-07,
+      "loss": 0.0438,
+      "step": 1327
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.6887481734047733e-07,
+      "loss": 0.0293,
+      "step": 1328
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.688504627374574e-07,
+      "loss": 0.5048,
+      "step": 1329
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.6882610813443743e-07,
+      "loss": 0.0306,
+      "step": 1330
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.6880175353141743e-07,
+      "loss": 0.0317,
+      "step": 1331
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.687773989283975e-07,
+      "loss": 0.0372,
+      "step": 1332
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.687530443253775e-07,
+      "loss": 0.0423,
+      "step": 1333
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.6872868972235753e-07,
+      "loss": 0.031,
+      "step": 1334
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.6870433511933753e-07,
+      "loss": 0.0324,
+      "step": 1335
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.686799805163176e-07,
+      "loss": 0.0869,
+      "step": 1336
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.686556259132976e-07,
+      "loss": 0.0267,
+      "step": 1337
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.6863127131027763e-07,
+      "loss": 0.0358,
+      "step": 1338
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.686069167072577e-07,
+      "loss": 0.205,
+      "step": 1339
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.685825621042377e-07,
+      "loss": 0.0519,
+      "step": 1340
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.6855820750121773e-07,
+      "loss": 0.039,
+      "step": 1341
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.6853385289819773e-07,
+      "loss": 0.0266,
+      "step": 1342
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.685094982951778e-07,
+      "loss": 0.0349,
+      "step": 1343
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.684851436921578e-07,
+      "loss": 0.0324,
+      "step": 1344
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.6846078908913783e-07,
+      "loss": 0.7774,
+      "step": 1345
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.684364344861179e-07,
+      "loss": 0.0349,
+      "step": 1346
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.684120798830979e-07,
+      "loss": 0.3103,
+      "step": 1347
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.6838772528007794e-07,
+      "loss": 0.0301,
+      "step": 1348
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.6836337067705793e-07,
+      "loss": 0.0325,
+      "step": 1349
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.68339016074038e-07,
+      "loss": 0.0505,
+      "step": 1350
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.68314661471018e-07,
+      "loss": 0.0304,
+      "step": 1351
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.6829030686799804e-07,
+      "loss": 0.0323,
+      "step": 1352
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 4.6826595226497804e-07,
+      "loss": 0.0267,
+      "step": 1353
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.682415976619581e-07,
+      "loss": 0.0422,
+      "step": 1354
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.6821724305893814e-07,
+      "loss": 0.2623,
+      "step": 1355
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.6819288845591814e-07,
+      "loss": 0.0391,
+      "step": 1356
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.681685338528982e-07,
+      "loss": 0.0909,
+      "step": 1357
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.681441792498782e-07,
+      "loss": 0.1304,
+      "step": 1358
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.6811982464685824e-07,
+      "loss": 0.0271,
+      "step": 1359
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.6809547004383824e-07,
+      "loss": 0.0312,
+      "step": 1360
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.680711154408183e-07,
+      "loss": 0.0401,
+      "step": 1361
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.6804676083779834e-07,
+      "loss": 0.029,
+      "step": 1362
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.6802240623477834e-07,
+      "loss": 0.0275,
+      "step": 1363
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.679980516317584e-07,
+      "loss": 0.0297,
+      "step": 1364
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.679736970287384e-07,
+      "loss": 0.0292,
+      "step": 1365
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.6794934242571844e-07,
+      "loss": 0.2948,
+      "step": 1366
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.6792498782269844e-07,
+      "loss": 0.0253,
+      "step": 1367
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.679006332196785e-07,
+      "loss": 0.031,
+      "step": 1368
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.678762786166585e-07,
+      "loss": 0.03,
+      "step": 1369
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.6785192401363854e-07,
+      "loss": 0.147,
+      "step": 1370
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.678275694106186e-07,
+      "loss": 0.0291,
+      "step": 1371
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.678032148075986e-07,
+      "loss": 0.0285,
+      "step": 1372
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 4.6777886020457864e-07,
+      "loss": 0.2831,
+      "step": 1373
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.6775450560155864e-07,
+      "loss": 0.7491,
+      "step": 1374
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.677301509985387e-07,
+      "loss": 0.0292,
+      "step": 1375
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.677057963955187e-07,
+      "loss": 0.054,
+      "step": 1376
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.6768144179249875e-07,
+      "loss": 0.1,
+      "step": 1377
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.676570871894788e-07,
+      "loss": 0.044,
+      "step": 1378
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.676327325864588e-07,
+      "loss": 0.052,
+      "step": 1379
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.6760837798343885e-07,
+      "loss": 0.0286,
+      "step": 1380
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.6758402338041885e-07,
+      "loss": 0.0259,
+      "step": 1381
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.675596687773989e-07,
+      "loss": 0.0291,
+      "step": 1382
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.675353141743789e-07,
+      "loss": 0.0526,
+      "step": 1383
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.6751095957135895e-07,
+      "loss": 0.0284,
+      "step": 1384
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.6748660496833895e-07,
+      "loss": 0.4269,
+      "step": 1385
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.67462250365319e-07,
+      "loss": 0.2591,
+      "step": 1386
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.6743789576229905e-07,
+      "loss": 0.0469,
+      "step": 1387
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.6741354115927905e-07,
+      "loss": 0.173,
+      "step": 1388
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.673891865562591e-07,
+      "loss": 0.0262,
+      "step": 1389
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.673648319532391e-07,
+      "loss": 0.0228,
+      "step": 1390
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.673404773502192e-07,
+      "loss": 0.0321,
+      "step": 1391
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.673161227471992e-07,
+      "loss": 0.0301,
+      "step": 1392
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.6729176814417925e-07,
+      "loss": 0.0264,
+      "step": 1393
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.672674135411593e-07,
+      "loss": 0.0249,
+      "step": 1394
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.672430589381393e-07,
+      "loss": 0.0286,
+      "step": 1395
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.6721870433511936e-07,
+      "loss": 0.0279,
+      "step": 1396
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.6719434973209935e-07,
+      "loss": 0.02,
+      "step": 1397
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.671699951290794e-07,
+      "loss": 0.025,
+      "step": 1398
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.671456405260594e-07,
+      "loss": 0.0459,
+      "step": 1399
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.6712128592303946e-07,
+      "loss": 0.0203,
+      "step": 1400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.670969313200195e-07,
+      "loss": 0.0239,
+      "step": 1401
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.670725767169995e-07,
+      "loss": 0.031,
+      "step": 1402
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.6704822211397956e-07,
+      "loss": 0.0278,
+      "step": 1403
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.6702386751095956e-07,
+      "loss": 0.0312,
+      "step": 1404
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.669995129079396e-07,
+      "loss": 0.0295,
+      "step": 1405
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.669751583049196e-07,
+      "loss": 0.0859,
+      "step": 1406
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.6695080370189966e-07,
+      "loss": 0.0391,
+      "step": 1407
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.6692644909887966e-07,
+      "loss": 0.0241,
+      "step": 1408
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.669020944958597e-07,
+      "loss": 0.0343,
+      "step": 1409
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.6687773989283976e-07,
+      "loss": 0.0275,
+      "step": 1410
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.6685338528981976e-07,
+      "loss": 0.0224,
+      "step": 1411
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.668290306867998e-07,
+      "loss": 0.04,
+      "step": 1412
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.668046760837798e-07,
+      "loss": 0.0265,
+      "step": 1413
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.6678032148075986e-07,
+      "loss": 0.0375,
+      "step": 1414
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6675596687773986e-07,
+      "loss": 0.0373,
+      "step": 1415
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.667316122747199e-07,
+      "loss": 0.0271,
+      "step": 1416
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6670725767169996e-07,
+      "loss": 0.191,
+      "step": 1417
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6668290306867996e-07,
+      "loss": 0.0284,
+      "step": 1418
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6665854846566e-07,
+      "loss": 0.0689,
+      "step": 1419
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6663419386264e-07,
+      "loss": 0.0239,
+      "step": 1420
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6660983925962006e-07,
+      "loss": 0.0232,
+      "step": 1421
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6658548465660006e-07,
+      "loss": 0.0289,
+      "step": 1422
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.665611300535801e-07,
+      "loss": 0.0242,
+      "step": 1423
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.665367754505601e-07,
+      "loss": 0.0205,
+      "step": 1424
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.6651242084754016e-07,
+      "loss": 0.032,
+      "step": 1425
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.664880662445202e-07,
+      "loss": 0.0239,
+      "step": 1426
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.664637116415002e-07,
+      "loss": 0.0234,
+      "step": 1427
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.6643935703848027e-07,
+      "loss": 0.0231,
+      "step": 1428
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.6641500243546027e-07,
+      "loss": 0.0264,
+      "step": 1429
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.663906478324403e-07,
+      "loss": 0.0174,
+      "step": 1430
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.663662932294203e-07,
+      "loss": 0.0241,
+      "step": 1431
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.6634193862640037e-07,
+      "loss": 0.0285,
+      "step": 1432
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.663175840233804e-07,
+      "loss": 0.0203,
+      "step": 1433
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.662932294203604e-07,
+      "loss": 0.0288,
+      "step": 1434
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.6626887481734047e-07,
+      "loss": 0.0246,
+      "step": 1435
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.6624452021432047e-07,
+      "loss": 0.3738,
+      "step": 1436
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.662201656113005e-07,
+      "loss": 0.0239,
+      "step": 1437
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.661958110082805e-07,
+      "loss": 0.0235,
+      "step": 1438
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.6617145640526057e-07,
+      "loss": 0.0655,
+      "step": 1439
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.6614710180224057e-07,
+      "loss": 0.0334,
+      "step": 1440
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.661227471992206e-07,
+      "loss": 0.0366,
+      "step": 1441
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.6609839259620067e-07,
+      "loss": 0.099,
+      "step": 1442
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.6607403799318067e-07,
+      "loss": 0.0206,
+      "step": 1443
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.660496833901607e-07,
+      "loss": 0.0254,
+      "step": 1444
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 4.660253287871407e-07,
+      "loss": 0.1561,
+      "step": 1445
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.6600097418412077e-07,
+      "loss": 0.165,
+      "step": 1446
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.6597661958110077e-07,
+      "loss": 0.0212,
+      "step": 1447
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.659522649780808e-07,
+      "loss": 0.0261,
+      "step": 1448
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.659279103750609e-07,
+      "loss": 0.5673,
+      "step": 1449
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.6590355577204087e-07,
+      "loss": 0.0199,
+      "step": 1450
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.658792011690209e-07,
+      "loss": 0.0251,
+      "step": 1451
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.658548465660009e-07,
+      "loss": 0.0255,
+      "step": 1452
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.65830491962981e-07,
+      "loss": 0.0247,
+      "step": 1453
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.65806137359961e-07,
+      "loss": 0.0737,
+      "step": 1454
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.65781782756941e-07,
+      "loss": 0.0218,
+      "step": 1455
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.65757428153921e-07,
+      "loss": 0.0243,
+      "step": 1456
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.657330735509011e-07,
+      "loss": 0.0251,
+      "step": 1457
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.6570871894788113e-07,
+      "loss": 0.0201,
+      "step": 1458
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.656843643448611e-07,
+      "loss": 0.0213,
+      "step": 1459
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.656600097418412e-07,
+      "loss": 0.0247,
+      "step": 1460
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.656356551388212e-07,
+      "loss": 0.0402,
+      "step": 1461
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.6561130053580123e-07,
+      "loss": 0.0198,
+      "step": 1462
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.6558694593278123e-07,
+      "loss": 0.0163,
+      "step": 1463
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.655625913297613e-07,
+      "loss": 0.0505,
+      "step": 1464
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.6553823672674133e-07,
+      "loss": 0.0191,
+      "step": 1465
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 4.6551388212372133e-07,
+      "loss": 0.1133,
+      "step": 1466
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.654895275207014e-07,
+      "loss": 0.0219,
+      "step": 1467
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.6546517291768143e-07,
+      "loss": 0.0223,
+      "step": 1468
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.654408183146615e-07,
+      "loss": 0.0204,
+      "step": 1469
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.654164637116415e-07,
+      "loss": 0.0202,
+      "step": 1470
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.6539210910862153e-07,
+      "loss": 0.0219,
+      "step": 1471
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.653677545056016e-07,
+      "loss": 0.0235,
+      "step": 1472
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.653433999025816e-07,
+      "loss": 0.0238,
+      "step": 1473
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.6531904529956164e-07,
+      "loss": 0.0349,
+      "step": 1474
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.6529469069654163e-07,
+      "loss": 0.0811,
+      "step": 1475
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.652703360935217e-07,
+      "loss": 0.0183,
+      "step": 1476
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.652459814905017e-07,
+      "loss": 0.0188,
+      "step": 1477
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.6522162688748174e-07,
+      "loss": 0.0234,
+      "step": 1478
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.6519727228446174e-07,
+      "loss": 0.0237,
+      "step": 1479
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.651729176814418e-07,
+      "loss": 0.0245,
+      "step": 1480
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.6514856307842184e-07,
+      "loss": 0.0268,
+      "step": 1481
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.6512420847540184e-07,
+      "loss": 0.0257,
+      "step": 1482
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.650998538723819e-07,
+      "loss": 0.0312,
+      "step": 1483
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.650754992693619e-07,
+      "loss": 0.0253,
+      "step": 1484
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.6505114466634194e-07,
+      "loss": 0.1282,
+      "step": 1485
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 4.6502679006332194e-07,
+      "loss": 0.0228,
+      "step": 1486
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.65002435460302e-07,
+      "loss": 0.0222,
+      "step": 1487
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.6497808085728204e-07,
+      "loss": 0.0176,
+      "step": 1488
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.6495372625426204e-07,
+      "loss": 0.0239,
+      "step": 1489
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.649293716512421e-07,
+      "loss": 0.0209,
+      "step": 1490
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.649050170482221e-07,
+      "loss": 0.0399,
+      "step": 1491
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.6488066244520214e-07,
+      "loss": 0.3384,
+      "step": 1492
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.6485630784218214e-07,
+      "loss": 0.02,
+      "step": 1493
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.648319532391622e-07,
+      "loss": 0.0209,
+      "step": 1494
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.648075986361422e-07,
+      "loss": 0.0204,
+      "step": 1495
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.6478324403312224e-07,
+      "loss": 0.0216,
+      "step": 1496
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.647588894301023e-07,
+      "loss": 0.0554,
+      "step": 1497
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.647345348270823e-07,
+      "loss": 0.0196,
+      "step": 1498
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.6471018022406234e-07,
+      "loss": 0.0248,
+      "step": 1499
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.6468582562104234e-07,
+      "loss": 0.0866,
+      "step": 1500
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.646614710180224e-07,
+      "loss": 0.021,
+      "step": 1501
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.646371164150024e-07,
+      "loss": 0.0836,
+      "step": 1502
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.6461276181198245e-07,
+      "loss": 0.0184,
+      "step": 1503
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.645884072089625e-07,
+      "loss": 0.022,
+      "step": 1504
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.645640526059425e-07,
+      "loss": 0.0317,
+      "step": 1505
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.6453969800292255e-07,
+      "loss": 0.026,
+      "step": 1506
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.6451534339990255e-07,
+      "loss": 0.0528,
+      "step": 1507
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.644909887968826e-07,
+      "loss": 0.0434,
+      "step": 1508
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.644666341938626e-07,
+      "loss": 0.0191,
+      "step": 1509
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.6444227959084265e-07,
+      "loss": 0.02,
+      "step": 1510
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.644179249878227e-07,
+      "loss": 0.0282,
+      "step": 1511
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.643935703848027e-07,
+      "loss": 0.0746,
+      "step": 1512
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.6436921578178275e-07,
+      "loss": 0.0199,
+      "step": 1513
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.6434486117876275e-07,
+      "loss": 0.0232,
+      "step": 1514
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.643205065757428e-07,
+      "loss": 0.022,
+      "step": 1515
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.642961519727228e-07,
+      "loss": 0.0506,
+      "step": 1516
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 4.6427179736970285e-07,
+      "loss": 0.1425,
+      "step": 1517
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.6424744276668285e-07,
+      "loss": 0.0442,
+      "step": 1518
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.642230881636629e-07,
+      "loss": 0.0223,
+      "step": 1519
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.6419873356064295e-07,
+      "loss": 0.0306,
+      "step": 1520
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.6417437895762295e-07,
+      "loss": 0.0211,
+      "step": 1521
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.64150024354603e-07,
+      "loss": 0.0397,
+      "step": 1522
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.64125669751583e-07,
+      "loss": 0.0223,
+      "step": 1523
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.6410131514856305e-07,
+      "loss": 0.019,
+      "step": 1524
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.6407696054554305e-07,
+      "loss": 0.02,
+      "step": 1525
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.640526059425231e-07,
+      "loss": 0.0349,
+      "step": 1526
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.6402825133950316e-07,
+      "loss": 0.0187,
+      "step": 1527
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 4.6400389673648315e-07,
+      "loss": 0.9809,
+      "step": 1528
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.639795421334632e-07,
+      "loss": 0.0206,
+      "step": 1529
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.639551875304432e-07,
+      "loss": 0.0201,
+      "step": 1530
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.6393083292742326e-07,
+      "loss": 1.0265,
+      "step": 1531
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.6390647832440325e-07,
+      "loss": 0.0454,
+      "step": 1532
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.638821237213833e-07,
+      "loss": 0.0327,
+      "step": 1533
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.638577691183633e-07,
+      "loss": 0.0175,
+      "step": 1534
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.6383341451534336e-07,
+      "loss": 0.0172,
+      "step": 1535
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.638090599123234e-07,
+      "loss": 0.0197,
+      "step": 1536
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.637847053093034e-07,
+      "loss": 0.0321,
+      "step": 1537
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.6376035070628346e-07,
+      "loss": 0.0185,
+      "step": 1538
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.6373599610326346e-07,
+      "loss": 0.0211,
+      "step": 1539
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.637116415002435e-07,
+      "loss": 0.0328,
+      "step": 1540
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.636872868972235e-07,
+      "loss": 0.018,
+      "step": 1541
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.6366293229420356e-07,
+      "loss": 0.0221,
+      "step": 1542
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.636385776911836e-07,
+      "loss": 0.0222,
+      "step": 1543
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.6361422308816366e-07,
+      "loss": 0.0147,
+      "step": 1544
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.635898684851437e-07,
+      "loss": 0.0168,
+      "step": 1545
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.635655138821237e-07,
+      "loss": 0.0457,
+      "step": 1546
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.6354115927910376e-07,
+      "loss": 0.2734,
+      "step": 1547
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.6351680467608376e-07,
+      "loss": 0.0179,
+      "step": 1548
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.634924500730638e-07,
+      "loss": 0.027,
+      "step": 1549
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.6346809547004387e-07,
+      "loss": 0.0161,
+      "step": 1550
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.6344374086702386e-07,
+      "loss": 0.0171,
+      "step": 1551
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.634193862640039e-07,
+      "loss": 0.5192,
+      "step": 1552
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.633950316609839e-07,
+      "loss": 0.0188,
+      "step": 1553
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.6337067705796397e-07,
+      "loss": 0.0226,
+      "step": 1554
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.6334632245494397e-07,
+      "loss": 0.0182,
+      "step": 1555
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.63321967851924e-07,
+      "loss": 0.0181,
+      "step": 1556
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.63297613248904e-07,
+      "loss": 0.0219,
+      "step": 1557
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 4.6327325864588407e-07,
+      "loss": 0.0173,
+      "step": 1558
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.632489040428641e-07,
+      "loss": 0.0133,
+      "step": 1559
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.632245494398441e-07,
+      "loss": 0.0197,
+      "step": 1560
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.6320019483682417e-07,
+      "loss": 0.0709,
+      "step": 1561
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.6317584023380417e-07,
+      "loss": 0.0168,
+      "step": 1562
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.631514856307842e-07,
+      "loss": 0.0181,
+      "step": 1563
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.631271310277642e-07,
+      "loss": 0.0171,
+      "step": 1564
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.6310277642474427e-07,
+      "loss": 0.0205,
+      "step": 1565
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.630784218217243e-07,
+      "loss": 0.0163,
+      "step": 1566
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.630540672187043e-07,
+      "loss": 0.0171,
+      "step": 1567
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.6302971261568437e-07,
+      "loss": 0.0174,
+      "step": 1568
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.6300535801266437e-07,
+      "loss": 0.0164,
+      "step": 1569
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.629810034096444e-07,
+      "loss": 0.0187,
+      "step": 1570
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.629566488066244e-07,
+      "loss": 0.0201,
+      "step": 1571
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.6293229420360447e-07,
+      "loss": 0.0164,
+      "step": 1572
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.6290793960058447e-07,
+      "loss": 0.0181,
+      "step": 1573
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.628835849975645e-07,
+      "loss": 0.0187,
+      "step": 1574
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.628592303945446e-07,
+      "loss": 0.0162,
+      "step": 1575
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.6283487579152457e-07,
+      "loss": 0.0197,
+      "step": 1576
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.628105211885046e-07,
+      "loss": 0.0169,
+      "step": 1577
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.627861665854846e-07,
+      "loss": 0.0167,
+      "step": 1578
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.627618119824647e-07,
+      "loss": 0.0193,
+      "step": 1579
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.627374573794447e-07,
+      "loss": 0.0261,
+      "step": 1580
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.627131027764247e-07,
+      "loss": 0.0235,
+      "step": 1581
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.626887481734048e-07,
+      "loss": 0.0194,
+      "step": 1582
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.626643935703848e-07,
+      "loss": 0.0161,
+      "step": 1583
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.6264003896736483e-07,
+      "loss": 0.018,
+      "step": 1584
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.626156843643448e-07,
+      "loss": 0.7002,
+      "step": 1585
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.625913297613249e-07,
+      "loss": 0.0168,
+      "step": 1586
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.625669751583049e-07,
+      "loss": 0.0182,
+      "step": 1587
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.6254262055528493e-07,
+      "loss": 0.0183,
+      "step": 1588
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.6251826595226493e-07,
+      "loss": 0.0335,
+      "step": 1589
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.62493911349245e-07,
+      "loss": 0.0166,
+      "step": 1590
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6246955674622503e-07,
+      "loss": 0.0285,
+      "step": 1591
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6244520214320503e-07,
+      "loss": 0.0177,
+      "step": 1592
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.624208475401851e-07,
+      "loss": 0.0176,
+      "step": 1593
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.623964929371651e-07,
+      "loss": 0.0176,
+      "step": 1594
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6237213833414513e-07,
+      "loss": 0.0586,
+      "step": 1595
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6234778373112513e-07,
+      "loss": 0.0181,
+      "step": 1596
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.623234291281052e-07,
+      "loss": 0.0172,
+      "step": 1597
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6229907452508523e-07,
+      "loss": 0.0155,
+      "step": 1598
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6227471992206523e-07,
+      "loss": 0.0199,
+      "step": 1599
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.622503653190453e-07,
+      "loss": 0.016,
+      "step": 1600
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.622260107160253e-07,
+      "loss": 0.0282,
+      "step": 1601
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.6220165611300533e-07,
+      "loss": 0.0386,
+      "step": 1602
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.6217730150998533e-07,
+      "loss": 0.0245,
+      "step": 1603
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.621529469069654e-07,
+      "loss": 0.0213,
+      "step": 1604
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.621285923039454e-07,
+      "loss": 0.0164,
+      "step": 1605
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.6210423770092543e-07,
+      "loss": 0.0259,
+      "step": 1606
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.620798830979055e-07,
+      "loss": 0.0149,
+      "step": 1607
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.620555284948855e-07,
+      "loss": 0.0153,
+      "step": 1608
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.6203117389186554e-07,
+      "loss": 0.0175,
+      "step": 1609
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.6200681928884553e-07,
+      "loss": 0.0144,
+      "step": 1610
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.619824646858256e-07,
+      "loss": 0.017,
+      "step": 1611
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.619581100828056e-07,
+      "loss": 0.016,
+      "step": 1612
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.6193375547978564e-07,
+      "loss": 0.0156,
+      "step": 1613
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.619094008767657e-07,
+      "loss": 0.0203,
+      "step": 1614
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.618850462737457e-07,
+      "loss": 0.0157,
+      "step": 1615
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.6186069167072574e-07,
+      "loss": 0.02,
+      "step": 1616
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.6183633706770574e-07,
+      "loss": 0.0161,
+      "step": 1617
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.618119824646858e-07,
+      "loss": 0.1314,
+      "step": 1618
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.617876278616658e-07,
+      "loss": 0.2882,
+      "step": 1619
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.6176327325864584e-07,
+      "loss": 0.0182,
+      "step": 1620
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.6173891865562594e-07,
+      "loss": 0.0164,
+      "step": 1621
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.6171456405260594e-07,
+      "loss": 0.0179,
+      "step": 1622
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.61690209449586e-07,
+      "loss": 0.013,
+      "step": 1623
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.61665854846566e-07,
+      "loss": 0.022,
+      "step": 1624
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.6164150024354604e-07,
+      "loss": 0.0169,
+      "step": 1625
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.6161714564052604e-07,
+      "loss": 0.0242,
+      "step": 1626
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.615927910375061e-07,
+      "loss": 0.0163,
+      "step": 1627
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.615684364344861e-07,
+      "loss": 0.0165,
+      "step": 1628
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.6154408183146615e-07,
+      "loss": 0.015,
+      "step": 1629
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.615197272284462e-07,
+      "loss": 0.1613,
+      "step": 1630
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.614953726254262e-07,
+      "loss": 0.0157,
+      "step": 1631
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.6147101802240625e-07,
+      "loss": 0.0145,
+      "step": 1632
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.6144666341938625e-07,
+      "loss": 0.0172,
+      "step": 1633
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.614223088163663e-07,
+      "loss": 0.0148,
+      "step": 1634
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.613979542133463e-07,
+      "loss": 0.0137,
+      "step": 1635
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.6137359961032635e-07,
+      "loss": 0.017,
+      "step": 1636
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.613492450073064e-07,
+      "loss": 0.0245,
+      "step": 1637
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.613248904042864e-07,
+      "loss": 0.0152,
+      "step": 1638
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.6130053580126645e-07,
+      "loss": 0.0177,
+      "step": 1639
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.6127618119824645e-07,
+      "loss": 0.0139,
+      "step": 1640
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.612518265952265e-07,
+      "loss": 0.0168,
+      "step": 1641
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.612274719922065e-07,
+      "loss": 0.0139,
+      "step": 1642
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.6120311738918655e-07,
+      "loss": 0.023,
+      "step": 1643
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.6117876278616655e-07,
+      "loss": 0.0139,
+      "step": 1644
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.611544081831466e-07,
+      "loss": 0.0144,
+      "step": 1645
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.6113005358012665e-07,
+      "loss": 0.0141,
+      "step": 1646
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.6110569897710665e-07,
+      "loss": 0.0157,
+      "step": 1647
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.610813443740867e-07,
+      "loss": 0.0535,
+      "step": 1648
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.610569897710667e-07,
+      "loss": 0.0127,
+      "step": 1649
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.6103263516804675e-07,
+      "loss": 0.0149,
+      "step": 1650
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.6100828056502675e-07,
+      "loss": 0.0173,
+      "step": 1651
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.609839259620068e-07,
+      "loss": 0.0386,
+      "step": 1652
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.6095957135898686e-07,
+      "loss": 0.0174,
+      "step": 1653
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.6093521675596685e-07,
+      "loss": 1.0275,
+      "step": 1654
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.609108621529469e-07,
+      "loss": 0.0174,
+      "step": 1655
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.608865075499269e-07,
+      "loss": 0.0779,
+      "step": 1656
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.6086215294690696e-07,
+      "loss": 0.0142,
+      "step": 1657
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.6083779834388695e-07,
+      "loss": 0.0176,
+      "step": 1658
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.60813443740867e-07,
+      "loss": 0.0188,
+      "step": 1659
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.60789089137847e-07,
+      "loss": 0.0662,
+      "step": 1660
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.6076473453482706e-07,
+      "loss": 0.0131,
+      "step": 1661
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.607403799318071e-07,
+      "loss": 0.0165,
+      "step": 1662
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.607160253287871e-07,
+      "loss": 0.0251,
+      "step": 1663
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.6069167072576716e-07,
+      "loss": 0.0133,
+      "step": 1664
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.6066731612274716e-07,
+      "loss": 0.0139,
+      "step": 1665
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.606429615197272e-07,
+      "loss": 0.0205,
+      "step": 1666
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.606186069167072e-07,
+      "loss": 0.0142,
+      "step": 1667
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.6059425231368726e-07,
+      "loss": 0.014,
+      "step": 1668
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.605698977106673e-07,
+      "loss": 0.0155,
+      "step": 1669
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.605455431076473e-07,
+      "loss": 0.7662,
+      "step": 1670
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.6052118850462736e-07,
+      "loss": 0.0149,
+      "step": 1671
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 4.6049683390160736e-07,
+      "loss": 0.0153,
+      "step": 1672
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.604724792985874e-07,
+      "loss": 0.0137,
+      "step": 1673
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.604481246955674e-07,
+      "loss": 0.0125,
+      "step": 1674
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.6042377009254746e-07,
+      "loss": 0.8999,
+      "step": 1675
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.603994154895275e-07,
+      "loss": 0.0139,
+      "step": 1676
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.603750608865075e-07,
+      "loss": 0.0155,
+      "step": 1677
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.6035070628348756e-07,
+      "loss": 0.0144,
+      "step": 1678
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.6032635168046756e-07,
+      "loss": 0.0234,
+      "step": 1679
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.603019970774476e-07,
+      "loss": 1.073,
+      "step": 1680
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.602776424744276e-07,
+      "loss": 0.0142,
+      "step": 1681
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 4.6025328787140766e-07,
+      "loss": 0.0265,
+      "step": 1682
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.6022893326838766e-07,
+      "loss": 0.0127,
+      "step": 1683
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.602045786653677e-07,
+      "loss": 0.012,
+      "step": 1684
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.6018022406234777e-07,
+      "loss": 0.0165,
+      "step": 1685
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.6015586945932777e-07,
+      "loss": 0.017,
+      "step": 1686
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.601315148563078e-07,
+      "loss": 0.0209,
+      "step": 1687
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.601071602532878e-07,
+      "loss": 1.5472,
+      "step": 1688
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.6008280565026787e-07,
+      "loss": 0.0177,
+      "step": 1689
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.6005845104724787e-07,
+      "loss": 0.0348,
+      "step": 1690
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.600340964442279e-07,
+      "loss": 0.0133,
+      "step": 1691
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 4.6000974184120797e-07,
+      "loss": 0.0175,
+      "step": 1692
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.5998538723818797e-07,
+      "loss": 0.0194,
+      "step": 1693
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.59961032635168e-07,
+      "loss": 0.0128,
+      "step": 1694
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.59936678032148e-07,
+      "loss": 0.0149,
+      "step": 1695
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.5991232342912807e-07,
+      "loss": 0.0135,
+      "step": 1696
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.598879688261081e-07,
+      "loss": 0.1628,
+      "step": 1697
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.5986361422308817e-07,
+      "loss": 0.0152,
+      "step": 1698
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.598392596200682e-07,
+      "loss": 0.0119,
+      "step": 1699
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.598149050170482e-07,
+      "loss": 0.0126,
+      "step": 1700
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.597905504140283e-07,
+      "loss": 0.0131,
+      "step": 1701
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 4.597661958110083e-07,
+      "loss": 0.0135,
+      "step": 1702
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.597418412079883e-07,
+      "loss": 0.0195,
+      "step": 1703
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.597174866049683e-07,
+      "loss": 0.0421,
+      "step": 1704
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.596931320019484e-07,
+      "loss": 0.7751,
+      "step": 1705
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.596687773989284e-07,
+      "loss": 0.0147,
+      "step": 1706
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.596444227959084e-07,
+      "loss": 0.9781,
+      "step": 1707
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.596200681928885e-07,
+      "loss": 0.0152,
+      "step": 1708
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.595957135898685e-07,
+      "loss": 0.0113,
+      "step": 1709
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.5957135898684853e-07,
+      "loss": 0.0275,
+      "step": 1710
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.5954700438382853e-07,
+      "loss": 0.0149,
+      "step": 1711
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.595226497808086e-07,
+      "loss": 0.0236,
+      "step": 1712
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 4.594982951777886e-07,
+      "loss": 0.0131,
+      "step": 1713
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.5947394057476863e-07,
+      "loss": 0.0155,
+      "step": 1714
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.594495859717487e-07,
+      "loss": 0.0138,
+      "step": 1715
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.594252313687287e-07,
+      "loss": 0.0114,
+      "step": 1716
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.5940087676570873e-07,
+      "loss": 0.0133,
+      "step": 1717
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.5937652216268873e-07,
+      "loss": 0.0223,
+      "step": 1718
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.593521675596688e-07,
+      "loss": 0.0256,
+      "step": 1719
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.593278129566488e-07,
+      "loss": 0.0251,
+      "step": 1720
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.5930345835362883e-07,
+      "loss": 0.015,
+      "step": 1721
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.5927910375060883e-07,
+      "loss": 0.0168,
+      "step": 1722
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 4.592547491475889e-07,
+      "loss": 0.0198,
+      "step": 1723
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.5923039454456893e-07,
+      "loss": 0.0142,
+      "step": 1724
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.5920603994154893e-07,
+      "loss": 0.014,
+      "step": 1725
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.59181685338529e-07,
+      "loss": 0.0116,
+      "step": 1726
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.59157330735509e-07,
+      "loss": 0.2271,
+      "step": 1727
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.5913297613248903e-07,
+      "loss": 0.0115,
+      "step": 1728
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.5910862152946903e-07,
+      "loss": 0.0143,
+      "step": 1729
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.590842669264491e-07,
+      "loss": 0.0173,
+      "step": 1730
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.5905991232342914e-07,
+      "loss": 0.0124,
+      "step": 1731
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.5903555772040913e-07,
+      "loss": 0.0099,
+      "step": 1732
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 4.590112031173892e-07,
+      "loss": 0.013,
+      "step": 1733
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.589868485143692e-07,
+      "loss": 0.0132,
+      "step": 1734
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.5896249391134924e-07,
+      "loss": 0.012,
+      "step": 1735
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.5893813930832924e-07,
+      "loss": 0.0134,
+      "step": 1736
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.589137847053093e-07,
+      "loss": 0.0144,
+      "step": 1737
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.588894301022893e-07,
+      "loss": 0.0133,
+      "step": 1738
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.5886507549926934e-07,
+      "loss": 0.0829,
+      "step": 1739
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.588407208962494e-07,
+      "loss": 0.0161,
+      "step": 1740
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.588163662932294e-07,
+      "loss": 0.0168,
+      "step": 1741
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.5879201169020944e-07,
+      "loss": 0.0137,
+      "step": 1742
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.5876765708718944e-07,
+      "loss": 0.0139,
+      "step": 1743
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 4.587433024841695e-07,
+      "loss": 0.0169,
+      "step": 1744
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.587189478811495e-07,
+      "loss": 0.0123,
+      "step": 1745
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5869459327812954e-07,
+      "loss": 0.0124,
+      "step": 1746
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.586702386751096e-07,
+      "loss": 0.0151,
+      "step": 1747
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.586458840720896e-07,
+      "loss": 0.0153,
+      "step": 1748
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5862152946906964e-07,
+      "loss": 0.0275,
+      "step": 1749
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5859717486604964e-07,
+      "loss": 0.0148,
+      "step": 1750
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.585728202630297e-07,
+      "loss": 0.0133,
+      "step": 1751
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.585484656600097e-07,
+      "loss": 0.0261,
+      "step": 1752
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5852411105698974e-07,
+      "loss": 0.0723,
+      "step": 1753
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 4.5849975645396974e-07,
+      "loss": 0.0148,
+      "step": 1754
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.584754018509498e-07,
+      "loss": 0.0116,
+      "step": 1755
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.5845104724792984e-07,
+      "loss": 0.0167,
+      "step": 1756
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.5842669264490984e-07,
+      "loss": 0.0092,
+      "step": 1757
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.584023380418899e-07,
+      "loss": 0.012,
+      "step": 1758
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.583779834388699e-07,
+      "loss": 0.0134,
+      "step": 1759
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.5835362883584995e-07,
+      "loss": 0.0129,
+      "step": 1760
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.5832927423282994e-07,
+      "loss": 0.0119,
+      "step": 1761
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.5830491962981e-07,
+      "loss": 0.0176,
+      "step": 1762
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.5828056502679005e-07,
+      "loss": 0.0146,
+      "step": 1763
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.5825621042377005e-07,
+      "loss": 0.0146,
+      "step": 1764
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.582318558207501e-07,
+      "loss": 0.0135,
+      "step": 1765
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.582075012177301e-07,
+      "loss": 0.0113,
+      "step": 1766
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.5818314661471015e-07,
+      "loss": 0.0125,
+      "step": 1767
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.5815879201169015e-07,
+      "loss": 0.0112,
+      "step": 1768
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.581344374086702e-07,
+      "loss": 0.0147,
+      "step": 1769
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.581100828056502e-07,
+      "loss": 0.0132,
+      "step": 1770
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.5808572820263025e-07,
+      "loss": 0.0124,
+      "step": 1771
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.580613735996103e-07,
+      "loss": 0.2455,
+      "step": 1772
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.580370189965903e-07,
+      "loss": 0.0134,
+      "step": 1773
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.580126643935704e-07,
+      "loss": 0.0121,
+      "step": 1774
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.579883097905504e-07,
+      "loss": 0.0154,
+      "step": 1775
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.5796395518753045e-07,
+      "loss": 0.0095,
+      "step": 1776
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.5793960058451045e-07,
+      "loss": 0.0113,
+      "step": 1777
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.579152459814905e-07,
+      "loss": 0.0119,
+      "step": 1778
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.5789089137847056e-07,
+      "loss": 0.0127,
+      "step": 1779
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.5786653677545055e-07,
+      "loss": 0.0128,
+      "step": 1780
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.578421821724306e-07,
+      "loss": 0.0115,
+      "step": 1781
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.578178275694106e-07,
+      "loss": 0.0122,
+      "step": 1782
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.5779347296639066e-07,
+      "loss": 0.0156,
+      "step": 1783
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.5776911836337065e-07,
+      "loss": 0.0113,
+      "step": 1784
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.577447637603507e-07,
+      "loss": 0.0126,
+      "step": 1785
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.5772040915733076e-07,
+      "loss": 0.0123,
+      "step": 1786
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.5769605455431076e-07,
+      "loss": 0.0143,
+      "step": 1787
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.576716999512908e-07,
+      "loss": 0.0134,
+      "step": 1788
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.576473453482708e-07,
+      "loss": 0.019,
+      "step": 1789
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.5762299074525086e-07,
+      "loss": 0.013,
+      "step": 1790
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.5759863614223086e-07,
+      "loss": 0.0107,
+      "step": 1791
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.575742815392109e-07,
+      "loss": 0.0144,
+      "step": 1792
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.575499269361909e-07,
+      "loss": 0.0101,
+      "step": 1793
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.5752557233317096e-07,
+      "loss": 0.0134,
+      "step": 1794
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.57501217730151e-07,
+      "loss": 0.0116,
+      "step": 1795
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.57476863127131e-07,
+      "loss": 0.0125,
+      "step": 1796
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.5745250852411106e-07,
+      "loss": 0.0097,
+      "step": 1797
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.5742815392109106e-07,
+      "loss": 0.011,
+      "step": 1798
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.574037993180711e-07,
+      "loss": 0.0106,
+      "step": 1799
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.573794447150511e-07,
+      "loss": 0.0132,
+      "step": 1800
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.5735509011203116e-07,
+      "loss": 0.0106,
+      "step": 1801
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.573307355090112e-07,
+      "loss": 0.0103,
+      "step": 1802
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.573063809059912e-07,
+      "loss": 0.0114,
+      "step": 1803
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.5728202630297126e-07,
+      "loss": 1.0867,
+      "step": 1804
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.5725767169995126e-07,
+      "loss": 0.0127,
+      "step": 1805
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.572333170969313e-07,
+      "loss": 0.0114,
+      "step": 1806
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.572089624939113e-07,
+      "loss": 0.0182,
+      "step": 1807
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.5718460789089136e-07,
+      "loss": 0.0123,
+      "step": 1808
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.5716025328787136e-07,
+      "loss": 0.0125,
+      "step": 1809
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.571358986848514e-07,
+      "loss": 0.0504,
+      "step": 1810
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.5711154408183147e-07,
+      "loss": 0.0123,
+      "step": 1811
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.5708718947881147e-07,
+      "loss": 0.0099,
+      "step": 1812
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.570628348757915e-07,
+      "loss": 0.0518,
+      "step": 1813
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.570384802727715e-07,
+      "loss": 0.011,
+      "step": 1814
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.5701412566975157e-07,
+      "loss": 0.0105,
+      "step": 1815
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.5698977106673157e-07,
+      "loss": 0.0121,
+      "step": 1816
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.569654164637116e-07,
+      "loss": 0.0094,
+      "step": 1817
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.5694106186069167e-07,
+      "loss": 0.0135,
+      "step": 1818
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.5691670725767167e-07,
+      "loss": 0.7509,
+      "step": 1819
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.568923526546517e-07,
+      "loss": 0.02,
+      "step": 1820
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.568679980516317e-07,
+      "loss": 0.8006,
+      "step": 1821
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.5684364344861177e-07,
+      "loss": 0.0121,
+      "step": 1822
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.5681928884559177e-07,
+      "loss": 0.9279,
+      "step": 1823
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.567949342425718e-07,
+      "loss": 0.0123,
+      "step": 1824
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.567705796395518e-07,
+      "loss": 0.0158,
+      "step": 1825
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.5674622503653187e-07,
+      "loss": 0.0106,
+      "step": 1826
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.567218704335119e-07,
+      "loss": 0.0127,
+      "step": 1827
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.566975158304919e-07,
+      "loss": 0.0095,
+      "step": 1828
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.5667316122747197e-07,
+      "loss": 0.01,
+      "step": 1829
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.5664880662445197e-07,
+      "loss": 0.0118,
+      "step": 1830
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.56624452021432e-07,
+      "loss": 0.0151,
+      "step": 1831
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.56600097418412e-07,
+      "loss": 0.048,
+      "step": 1832
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.5657574281539207e-07,
+      "loss": 0.0109,
+      "step": 1833
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.565513882123721e-07,
+      "loss": 0.0173,
+      "step": 1834
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.565270336093521e-07,
+      "loss": 0.0114,
+      "step": 1835
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.565026790063322e-07,
+      "loss": 0.0151,
+      "step": 1836
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.564783244033122e-07,
+      "loss": 0.0137,
+      "step": 1837
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.564539698002922e-07,
+      "loss": 0.0127,
+      "step": 1838
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.564296151972722e-07,
+      "loss": 0.0532,
+      "step": 1839
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.564052605942523e-07,
+      "loss": 0.683,
+      "step": 1840
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.5638090599123233e-07,
+      "loss": 0.0096,
+      "step": 1841
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.563565513882123e-07,
+      "loss": 0.0148,
+      "step": 1842
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.563321967851924e-07,
+      "loss": 0.009,
+      "step": 1843
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.563078421821724e-07,
+      "loss": 0.0117,
+      "step": 1844
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.5628348757915243e-07,
+      "loss": 0.0101,
+      "step": 1845
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.5625913297613243e-07,
+      "loss": 0.0143,
+      "step": 1846
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.562347783731125e-07,
+      "loss": 0.0991,
+      "step": 1847
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.562104237700925e-07,
+      "loss": 0.0117,
+      "step": 1848
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.5618606916707253e-07,
+      "loss": 0.0166,
+      "step": 1849
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.561617145640526e-07,
+      "loss": 0.0419,
+      "step": 1850
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.5613735996103263e-07,
+      "loss": 0.0097,
+      "step": 1851
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.561130053580127e-07,
+      "loss": 0.0337,
+      "step": 1852
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.560886507549927e-07,
+      "loss": 0.0117,
+      "step": 1853
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.5606429615197273e-07,
+      "loss": 0.0116,
+      "step": 1854
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.5603994154895273e-07,
+      "loss": 0.0239,
+      "step": 1855
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.560155869459328e-07,
+      "loss": 0.0097,
+      "step": 1856
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.5599123234291284e-07,
+      "loss": 0.0161,
+      "step": 1857
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5596687773989283e-07,
+      "loss": 0.0363,
+      "step": 1858
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.559425231368729e-07,
+      "loss": 0.0192,
+      "step": 1859
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.559181685338529e-07,
+      "loss": 0.0102,
+      "step": 1860
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5589381393083294e-07,
+      "loss": 0.0124,
+      "step": 1861
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5586945932781294e-07,
+      "loss": 0.0099,
+      "step": 1862
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.55845104724793e-07,
+      "loss": 0.0488,
+      "step": 1863
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5582075012177304e-07,
+      "loss": 0.016,
+      "step": 1864
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.5579639551875304e-07,
+      "loss": 0.209,
+      "step": 1865
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.557720409157331e-07,
+      "loss": 0.0123,
+      "step": 1866
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.557476863127131e-07,
+      "loss": 0.0135,
+      "step": 1867
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.5572333170969314e-07,
+      "loss": 0.0134,
+      "step": 1868
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.5569897710667314e-07,
+      "loss": 0.0088,
+      "step": 1869
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.556746225036532e-07,
+      "loss": 0.011,
+      "step": 1870
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.556502679006332e-07,
+      "loss": 0.0101,
+      "step": 1871
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.5562591329761324e-07,
+      "loss": 0.0105,
+      "step": 1872
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.556015586945933e-07,
+      "loss": 0.0117,
+      "step": 1873
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.555772040915733e-07,
+      "loss": 0.0108,
+      "step": 1874
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.5555284948855334e-07,
+      "loss": 0.7945,
+      "step": 1875
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.5552849488553334e-07,
+      "loss": 0.012,
+      "step": 1876
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.555041402825134e-07,
+      "loss": 0.0091,
+      "step": 1877
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.554797856794934e-07,
+      "loss": 0.011,
+      "step": 1878
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.5545543107647344e-07,
+      "loss": 0.0533,
+      "step": 1879
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.554310764734535e-07,
+      "loss": 0.0141,
+      "step": 1880
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.554067218704335e-07,
+      "loss": 0.0124,
+      "step": 1881
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.5538236726741354e-07,
+      "loss": 0.0122,
+      "step": 1882
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.5535801266439354e-07,
+      "loss": 0.0113,
+      "step": 1883
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.553336580613736e-07,
+      "loss": 0.0108,
+      "step": 1884
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.553093034583536e-07,
+      "loss": 0.2307,
+      "step": 1885
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.5528494885533365e-07,
+      "loss": 0.0101,
+      "step": 1886
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.5526059425231364e-07,
+      "loss": 0.011,
+      "step": 1887
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.552362396492937e-07,
+      "loss": 0.0232,
+      "step": 1888
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.5521188504627375e-07,
+      "loss": 0.0083,
+      "step": 1889
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.5518753044325375e-07,
+      "loss": 0.0086,
+      "step": 1890
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.551631758402338e-07,
+      "loss": 0.0097,
+      "step": 1891
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.551388212372138e-07,
+      "loss": 0.0097,
+      "step": 1892
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.5511446663419385e-07,
+      "loss": 0.0086,
+      "step": 1893
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.5509011203117385e-07,
+      "loss": 0.0132,
+      "step": 1894
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.550657574281539e-07,
+      "loss": 0.0126,
+      "step": 1895
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.5504140282513395e-07,
+      "loss": 0.0097,
+      "step": 1896
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.5501704822211395e-07,
+      "loss": 0.0109,
+      "step": 1897
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.54992693619094e-07,
+      "loss": 0.0117,
+      "step": 1898
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.54968339016074e-07,
+      "loss": 0.0119,
+      "step": 1899
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.5494398441305405e-07,
+      "loss": 0.0094,
+      "step": 1900
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.5491962981003405e-07,
+      "loss": 0.0101,
+      "step": 1901
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.548952752070141e-07,
+      "loss": 0.0098,
+      "step": 1902
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.548709206039941e-07,
+      "loss": 0.0084,
+      "step": 1903
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.5484656600097415e-07,
+      "loss": 0.0105,
+      "step": 1904
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.548222113979542e-07,
+      "loss": 0.0169,
+      "step": 1905
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.547978567949342e-07,
+      "loss": 0.0106,
+      "step": 1906
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.5477350219191425e-07,
+      "loss": 0.0122,
+      "step": 1907
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.5474914758889425e-07,
+      "loss": 0.0108,
+      "step": 1908
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.547247929858743e-07,
+      "loss": 0.0118,
+      "step": 1909
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.547004383828543e-07,
+      "loss": 0.0118,
+      "step": 1910
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.5467608377983435e-07,
+      "loss": 0.0113,
+      "step": 1911
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.546517291768144e-07,
+      "loss": 0.0167,
+      "step": 1912
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.546273745737944e-07,
+      "loss": 0.0107,
+      "step": 1913
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.5460301997077446e-07,
+      "loss": 0.0087,
+      "step": 1914
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.5457866536775445e-07,
+      "loss": 0.0728,
+      "step": 1915
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.545543107647345e-07,
+      "loss": 0.0119,
+      "step": 1916
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.545299561617145e-07,
+      "loss": 0.0113,
+      "step": 1917
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.5450560155869456e-07,
+      "loss": 1.1748,
+      "step": 1918
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.5448124695567456e-07,
+      "loss": 0.0093,
+      "step": 1919
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.544568923526546e-07,
+      "loss": 0.0593,
+      "step": 1920
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.5443253774963466e-07,
+      "loss": 0.0098,
+      "step": 1921
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.5440818314661466e-07,
+      "loss": 0.0118,
+      "step": 1922
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.543838285435947e-07,
+      "loss": 0.0085,
+      "step": 1923
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.543594739405747e-07,
+      "loss": 0.0093,
+      "step": 1924
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.5433511933755476e-07,
+      "loss": 0.0105,
+      "step": 1925
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.5431076473453476e-07,
+      "loss": 0.0093,
+      "step": 1926
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.5428641013151486e-07,
+      "loss": 0.0119,
+      "step": 1927
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.542620555284949e-07,
+      "loss": 0.0082,
+      "step": 1928
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.542377009254749e-07,
+      "loss": 0.0226,
+      "step": 1929
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.5421334632245496e-07,
+      "loss": 0.01,
+      "step": 1930
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.5418899171943496e-07,
+      "loss": 0.012,
+      "step": 1931
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.54164637116415e-07,
+      "loss": 0.0112,
+      "step": 1932
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.54140282513395e-07,
+      "loss": 0.0122,
+      "step": 1933
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.5411592791037506e-07,
+      "loss": 0.0132,
+      "step": 1934
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.540915733073551e-07,
+      "loss": 0.0093,
+      "step": 1935
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.540672187043351e-07,
+      "loss": 0.0094,
+      "step": 1936
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.5404286410131517e-07,
+      "loss": 0.0096,
+      "step": 1937
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.5401850949829517e-07,
+      "loss": 0.0094,
+      "step": 1938
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.539941548952752e-07,
+      "loss": 0.4423,
+      "step": 1939
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.539698002922552e-07,
+      "loss": 0.0092,
+      "step": 1940
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.5394544568923527e-07,
+      "loss": 0.0113,
+      "step": 1941
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.5392109108621527e-07,
+      "loss": 0.0086,
+      "step": 1942
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.538967364831953e-07,
+      "loss": 1.2233,
+      "step": 1943
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.5387238188017537e-07,
+      "loss": 0.0115,
+      "step": 1944
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.5384802727715537e-07,
+      "loss": 0.0684,
+      "step": 1945
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.538236726741354e-07,
+      "loss": 0.0092,
+      "step": 1946
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.537993180711154e-07,
+      "loss": 0.0659,
+      "step": 1947
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.5377496346809547e-07,
+      "loss": 0.0085,
+      "step": 1948
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.5375060886507547e-07,
+      "loss": 0.0159,
+      "step": 1949
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.537262542620555e-07,
+      "loss": 0.0097,
+      "step": 1950
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.5370189965903557e-07,
+      "loss": 0.0096,
+      "step": 1951
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.5367754505601557e-07,
+      "loss": 0.0172,
+      "step": 1952
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.536531904529956e-07,
+      "loss": 0.1047,
+      "step": 1953
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.536288358499756e-07,
+      "loss": 0.0102,
+      "step": 1954
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.5360448124695567e-07,
+      "loss": 0.0099,
+      "step": 1955
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.5358012664393567e-07,
+      "loss": 0.0122,
+      "step": 1956
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.535557720409157e-07,
+      "loss": 0.0085,
+      "step": 1957
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.535314174378957e-07,
+      "loss": 0.0107,
+      "step": 1958
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.5350706283487577e-07,
+      "loss": 0.0094,
+      "step": 1959
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.534827082318558e-07,
+      "loss": 0.0088,
+      "step": 1960
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.534583536288358e-07,
+      "loss": 0.011,
+      "step": 1961
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.534339990258159e-07,
+      "loss": 0.0088,
+      "step": 1962
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.534096444227959e-07,
+      "loss": 0.0107,
+      "step": 1963
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.533852898197759e-07,
+      "loss": 0.0095,
+      "step": 1964
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.533609352167559e-07,
+      "loss": 0.0103,
+      "step": 1965
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.53336580613736e-07,
+      "loss": 0.0104,
+      "step": 1966
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.5331222601071603e-07,
+      "loss": 0.0085,
+      "step": 1967
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.53287871407696e-07,
+      "loss": 0.0107,
+      "step": 1968
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.532635168046761e-07,
+      "loss": 0.009,
+      "step": 1969
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.532391622016561e-07,
+      "loss": 0.0096,
+      "step": 1970
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.5321480759863613e-07,
+      "loss": 0.0084,
+      "step": 1971
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.5319045299561613e-07,
+      "loss": 0.0784,
+      "step": 1972
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.531660983925962e-07,
+      "loss": 0.008,
+      "step": 1973
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.531417437895762e-07,
+      "loss": 0.0098,
+      "step": 1974
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.5311738918655623e-07,
+      "loss": 0.7634,
+      "step": 1975
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.530930345835363e-07,
+      "loss": 0.0087,
+      "step": 1976
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.530686799805163e-07,
+      "loss": 0.0105,
+      "step": 1977
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.5304432537749633e-07,
+      "loss": 0.0122,
+      "step": 1978
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.5301997077447633e-07,
+      "loss": 0.0094,
+      "step": 1979
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.529956161714564e-07,
+      "loss": 0.01,
+      "step": 1980
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.529712615684364e-07,
+      "loss": 0.0096,
+      "step": 1981
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.5294690696541643e-07,
+      "loss": 0.0148,
+      "step": 1982
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.529225523623965e-07,
+      "loss": 0.0121,
+      "step": 1983
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.528981977593765e-07,
+      "loss": 0.0095,
+      "step": 1984
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.5287384315635653e-07,
+      "loss": 0.0083,
+      "step": 1985
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.5284948855333653e-07,
+      "loss": 0.0648,
+      "step": 1986
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.528251339503166e-07,
+      "loss": 0.0085,
+      "step": 1987
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.528007793472966e-07,
+      "loss": 0.0098,
+      "step": 1988
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.5277642474427663e-07,
+      "loss": 0.0121,
+      "step": 1989
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.5275207014125663e-07,
+      "loss": 0.0095,
+      "step": 1990
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.527277155382367e-07,
+      "loss": 0.0082,
+      "step": 1991
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.5270336093521674e-07,
+      "loss": 0.1658,
+      "step": 1992
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.5267900633219674e-07,
+      "loss": 0.0105,
+      "step": 1993
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.526546517291768e-07,
+      "loss": 0.0137,
+      "step": 1994
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.526302971261568e-07,
+      "loss": 0.0108,
+      "step": 1995
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.5260594252313684e-07,
+      "loss": 0.0089,
+      "step": 1996
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.5258158792011684e-07,
+      "loss": 0.0086,
+      "step": 1997
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.525572333170969e-07,
+      "loss": 0.011,
+      "step": 1998
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.5253287871407694e-07,
+      "loss": 0.0085,
+      "step": 1999
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.5250852411105694e-07,
+      "loss": 0.0091,
+      "step": 2000
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.52484169508037e-07,
+      "loss": 0.0095,
+      "step": 2001
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.52459814905017e-07,
+      "loss": 0.0087,
+      "step": 2002
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.5243546030199704e-07,
+      "loss": 0.0074,
+      "step": 2003
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.524111056989771e-07,
+      "loss": 0.0101,
+      "step": 2004
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.5238675109595714e-07,
+      "loss": 0.0129,
+      "step": 2005
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.523623964929372e-07,
+      "loss": 0.0096,
+      "step": 2006
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.523380418899172e-07,
+      "loss": 0.0089,
+      "step": 2007
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.5231368728689724e-07,
+      "loss": 0.007,
+      "step": 2008
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.5228933268387724e-07,
+      "loss": 0.0254,
+      "step": 2009
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.522649780808573e-07,
+      "loss": 0.0084,
+      "step": 2010
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.522406234778373e-07,
+      "loss": 0.0097,
+      "step": 2011
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.5221626887481735e-07,
+      "loss": 0.0105,
+      "step": 2012
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.5219191427179734e-07,
+      "loss": 0.0106,
+      "step": 2013
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.521675596687774e-07,
+      "loss": 0.0168,
+      "step": 2014
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.5214320506575745e-07,
+      "loss": 0.0106,
+      "step": 2015
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.5211885046273745e-07,
+      "loss": 0.0083,
+      "step": 2016
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.520944958597175e-07,
+      "loss": 0.0088,
+      "step": 2017
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.520701412566975e-07,
+      "loss": 0.0086,
+      "step": 2018
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.5204578665367755e-07,
+      "loss": 0.3261,
+      "step": 2019
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.5202143205065755e-07,
+      "loss": 0.0096,
+      "step": 2020
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.519970774476376e-07,
+      "loss": 0.0089,
+      "step": 2021
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.5197272284461765e-07,
+      "loss": 0.0078,
+      "step": 2022
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.5194836824159765e-07,
+      "loss": 0.0098,
+      "step": 2023
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.519240136385777e-07,
+      "loss": 0.0085,
+      "step": 2024
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.518996590355577e-07,
+      "loss": 0.0087,
+      "step": 2025
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.5187530443253775e-07,
+      "loss": 0.0084,
+      "step": 2026
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.5185094982951775e-07,
+      "loss": 0.0081,
+      "step": 2027
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.518265952264978e-07,
+      "loss": 0.0117,
+      "step": 2028
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.518022406234778e-07,
+      "loss": 0.0076,
+      "step": 2029
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.5177788602045785e-07,
+      "loss": 0.0102,
+      "step": 2030
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.517535314174379e-07,
+      "loss": 0.01,
+      "step": 2031
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.517291768144179e-07,
+      "loss": 0.0096,
+      "step": 2032
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.5170482221139795e-07,
+      "loss": 0.0087,
+      "step": 2033
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.5168046760837795e-07,
+      "loss": 0.0079,
+      "step": 2034
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.51656113005358e-07,
+      "loss": 0.0079,
+      "step": 2035
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.51631758402338e-07,
+      "loss": 0.008,
+      "step": 2036
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.5160740379931805e-07,
+      "loss": 0.0096,
+      "step": 2037
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.515830491962981e-07,
+      "loss": 0.0087,
+      "step": 2038
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.515586945932781e-07,
+      "loss": 0.0125,
+      "step": 2039
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.5153433999025816e-07,
+      "loss": 0.0078,
+      "step": 2040
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.5150998538723815e-07,
+      "loss": 0.0071,
+      "step": 2041
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.514856307842182e-07,
+      "loss": 0.0106,
+      "step": 2042
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.514612761811982e-07,
+      "loss": 0.0088,
+      "step": 2043
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.5143692157817826e-07,
+      "loss": 0.01,
+      "step": 2044
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.514125669751583e-07,
+      "loss": 0.0088,
+      "step": 2045
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.513882123721383e-07,
+      "loss": 0.0139,
+      "step": 2046
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.5136385776911836e-07,
+      "loss": 0.0073,
+      "step": 2047
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.5133950316609836e-07,
+      "loss": 0.008,
+      "step": 2048
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.513151485630784e-07,
+      "loss": 0.014,
+      "step": 2049
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.512907939600584e-07,
+      "loss": 0.0093,
+      "step": 2050
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.5126643935703846e-07,
+      "loss": 0.0082,
+      "step": 2051
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.5124208475401846e-07,
+      "loss": 0.0074,
+      "step": 2052
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.512177301509985e-07,
+      "loss": 0.0099,
+      "step": 2053
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.5119337554797856e-07,
+      "loss": 0.0169,
+      "step": 2054
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.5116902094495856e-07,
+      "loss": 0.0082,
+      "step": 2055
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.511446663419386e-07,
+      "loss": 0.0087,
+      "step": 2056
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.511203117389186e-07,
+      "loss": 0.0087,
+      "step": 2057
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.5109595713589866e-07,
+      "loss": 0.0093,
+      "step": 2058
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.9946524064171123,
+      "eval_loss": 0.03693099319934845,
+      "eval_runtime": 2.7762,
+      "eval_samples_per_second": 740.952,
+      "eval_steps_per_second": 11.887,
+      "step": 2058
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.5107160253287866e-07,
+      "loss": 0.0452,
+      "step": 2059
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.510472479298587e-07,
+      "loss": 0.0084,
+      "step": 2060
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.5102289332683876e-07,
+      "loss": 0.039,
+      "step": 2061
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.5099853872381876e-07,
+      "loss": 0.0101,
+      "step": 2062
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.509741841207988e-07,
+      "loss": 0.0091,
+      "step": 2063
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.509498295177788e-07,
+      "loss": 0.0086,
+      "step": 2064
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.5092547491475886e-07,
+      "loss": 0.0093,
+      "step": 2065
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.5090112031173886e-07,
+      "loss": 0.0078,
+      "step": 2066
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.508767657087189e-07,
+      "loss": 0.0093,
+      "step": 2067
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.508524111056989e-07,
+      "loss": 0.0073,
+      "step": 2068
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.5082805650267897e-07,
+      "loss": 0.501,
+      "step": 2069
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.50803701899659e-07,
+      "loss": 0.0066,
+      "step": 2070
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.50779347296639e-07,
+      "loss": 0.0074,
+      "step": 2071
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.5075499269361907e-07,
+      "loss": 0.0109,
+      "step": 2072
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.5073063809059907e-07,
+      "loss": 0.0082,
+      "step": 2073
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.507062834875791e-07,
+      "loss": 0.0071,
+      "step": 2074
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.506819288845591e-07,
+      "loss": 0.0184,
+      "step": 2075
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5065757428153917e-07,
+      "loss": 0.0108,
+      "step": 2076
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.506332196785192e-07,
+      "loss": 0.0131,
+      "step": 2077
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.506088650754992e-07,
+      "loss": 0.0093,
+      "step": 2078
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5058451047247927e-07,
+      "loss": 0.0811,
+      "step": 2079
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5056015586945927e-07,
+      "loss": 0.011,
+      "step": 2080
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5053580126643937e-07,
+      "loss": 0.0078,
+      "step": 2081
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5051144666341937e-07,
+      "loss": 0.0072,
+      "step": 2082
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.504870920603994e-07,
+      "loss": 0.008,
+      "step": 2083
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.504627374573795e-07,
+      "loss": 0.0081,
+      "step": 2084
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.504383828543595e-07,
+      "loss": 0.0091,
+      "step": 2085
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.504140282513395e-07,
+      "loss": 0.0084,
+      "step": 2086
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.503896736483195e-07,
+      "loss": 0.0563,
+      "step": 2087
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.503653190452996e-07,
+      "loss": 0.3385,
+      "step": 2088
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.503409644422796e-07,
+      "loss": 0.0087,
+      "step": 2089
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.503166098392596e-07,
+      "loss": 0.0072,
+      "step": 2090
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.502922552362396e-07,
+      "loss": 0.0065,
+      "step": 2091
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.502679006332197e-07,
+      "loss": 0.7589,
+      "step": 2092
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5024354603019973e-07,
+      "loss": 0.0085,
+      "step": 2093
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5021919142717973e-07,
+      "loss": 0.0093,
+      "step": 2094
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.501948368241598e-07,
+      "loss": 0.0073,
+      "step": 2095
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.501704822211398e-07,
+      "loss": 0.0088,
+      "step": 2096
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5014612761811983e-07,
+      "loss": 0.0334,
+      "step": 2097
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5012177301509983e-07,
+      "loss": 0.0083,
+      "step": 2098
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.500974184120799e-07,
+      "loss": 0.0228,
+      "step": 2099
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5007306380905993e-07,
+      "loss": 0.0155,
+      "step": 2100
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5004870920603993e-07,
+      "loss": 0.0103,
+      "step": 2101
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5002435460302e-07,
+      "loss": 0.008,
+      "step": 2102
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.5e-07,
+      "loss": 0.0087,
+      "step": 2103
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4997564539698003e-07,
+      "loss": 0.0087,
+      "step": 2104
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4995129079396003e-07,
+      "loss": 0.7354,
+      "step": 2105
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.499269361909401e-07,
+      "loss": 0.0135,
+      "step": 2106
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.499025815879201e-07,
+      "loss": 0.0078,
+      "step": 2107
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4987822698490013e-07,
+      "loss": 0.0075,
+      "step": 2108
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.498538723818802e-07,
+      "loss": 0.0099,
+      "step": 2109
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.498295177788602e-07,
+      "loss": 0.0079,
+      "step": 2110
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4980516317584023e-07,
+      "loss": 0.0083,
+      "step": 2111
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.4978080857282023e-07,
+      "loss": 0.0082,
+      "step": 2112
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.497564539698003e-07,
+      "loss": 0.0101,
+      "step": 2113
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.497320993667803e-07,
+      "loss": 0.0089,
+      "step": 2114
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4970774476376033e-07,
+      "loss": 0.0087,
+      "step": 2115
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.496833901607404e-07,
+      "loss": 0.008,
+      "step": 2116
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.496590355577204e-07,
+      "loss": 0.0088,
+      "step": 2117
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4963468095470044e-07,
+      "loss": 0.0086,
+      "step": 2118
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4961032635168044e-07,
+      "loss": 0.0073,
+      "step": 2119
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.495859717486605e-07,
+      "loss": 0.0105,
+      "step": 2120
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.495616171456405e-07,
+      "loss": 0.0097,
+      "step": 2121
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4953726254262054e-07,
+      "loss": 0.0091,
+      "step": 2122
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.4951290793960054e-07,
+      "loss": 0.0084,
+      "step": 2123
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.494885533365806e-07,
+      "loss": 0.0071,
+      "step": 2124
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4946419873356064e-07,
+      "loss": 0.0068,
+      "step": 2125
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4943984413054064e-07,
+      "loss": 0.0059,
+      "step": 2126
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.494154895275207e-07,
+      "loss": 0.0078,
+      "step": 2127
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.493911349245007e-07,
+      "loss": 0.0071,
+      "step": 2128
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4936678032148074e-07,
+      "loss": 0.0059,
+      "step": 2129
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4934242571846074e-07,
+      "loss": 0.0714,
+      "step": 2130
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.493180711154408e-07,
+      "loss": 0.0082,
+      "step": 2131
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4929371651242084e-07,
+      "loss": 0.0069,
+      "step": 2132
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.4926936190940084e-07,
+      "loss": 0.0086,
+      "step": 2133
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.492450073063809e-07,
+      "loss": 0.0071,
+      "step": 2134
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.492206527033609e-07,
+      "loss": 0.0091,
+      "step": 2135
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.4919629810034094e-07,
+      "loss": 0.0073,
+      "step": 2136
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.4917194349732094e-07,
+      "loss": 0.0083,
+      "step": 2137
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.49147588894301e-07,
+      "loss": 0.0096,
+      "step": 2138
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.49123234291281e-07,
+      "loss": 0.0085,
+      "step": 2139
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.4909887968826104e-07,
+      "loss": 0.0095,
+      "step": 2140
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.490745250852411e-07,
+      "loss": 0.0111,
+      "step": 2141
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.490501704822211e-07,
+      "loss": 0.0177,
+      "step": 2142
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.4902581587920115e-07,
+      "loss": 0.0086,
+      "step": 2143
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.4900146127618114e-07,
+      "loss": 0.0072,
+      "step": 2144
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.489771066731612e-07,
+      "loss": 0.0099,
+      "step": 2145
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.489527520701412e-07,
+      "loss": 0.0086,
+      "step": 2146
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.4892839746712125e-07,
+      "loss": 0.02,
+      "step": 2147
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.489040428641013e-07,
+      "loss": 0.0078,
+      "step": 2148
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.488796882610813e-07,
+      "loss": 0.0112,
+      "step": 2149
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.4885533365806135e-07,
+      "loss": 0.0117,
+      "step": 2150
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.4883097905504135e-07,
+      "loss": 0.0077,
+      "step": 2151
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.488066244520214e-07,
+      "loss": 0.0092,
+      "step": 2152
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.487822698490014e-07,
+      "loss": 0.0068,
+      "step": 2153
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.4875791524598145e-07,
+      "loss": 0.0074,
+      "step": 2154
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.4873356064296145e-07,
+      "loss": 0.0061,
+      "step": 2155
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.487092060399415e-07,
+      "loss": 0.0076,
+      "step": 2156
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.486848514369216e-07,
+      "loss": 0.0075,
+      "step": 2157
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.486604968339016e-07,
+      "loss": 0.0075,
+      "step": 2158
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.4863614223088165e-07,
+      "loss": 0.0085,
+      "step": 2159
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.4861178762786165e-07,
+      "loss": 0.007,
+      "step": 2160
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.485874330248417e-07,
+      "loss": 0.0074,
+      "step": 2161
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.485630784218217e-07,
+      "loss": 0.0116,
+      "step": 2162
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.4853872381880175e-07,
+      "loss": 0.0087,
+      "step": 2163
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.485143692157818e-07,
+      "loss": 0.0092,
+      "step": 2164
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.484900146127618e-07,
+      "loss": 0.007,
+      "step": 2165
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.4846566000974186e-07,
+      "loss": 0.611,
+      "step": 2166
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.4844130540672185e-07,
+      "loss": 0.0091,
+      "step": 2167
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.484169508037019e-07,
+      "loss": 0.0308,
+      "step": 2168
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.483925962006819e-07,
+      "loss": 0.0083,
+      "step": 2169
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.4836824159766196e-07,
+      "loss": 0.0095,
+      "step": 2170
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.48343886994642e-07,
+      "loss": 0.0073,
+      "step": 2171
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.48319532391622e-07,
+      "loss": 0.006,
+      "step": 2172
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.4829517778860206e-07,
+      "loss": 0.0082,
+      "step": 2173
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.4827082318558206e-07,
+      "loss": 0.0067,
+      "step": 2174
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.482464685825621e-07,
+      "loss": 0.0081,
+      "step": 2175
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.482221139795421e-07,
+      "loss": 0.0084,
+      "step": 2176
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.4819775937652216e-07,
+      "loss": 0.0062,
+      "step": 2177
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.4817340477350216e-07,
+      "loss": 0.0086,
+      "step": 2178
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.481490501704822e-07,
+      "loss": 0.0063,
+      "step": 2179
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.4812469556746226e-07,
+      "loss": 0.0087,
+      "step": 2180
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.4810034096444226e-07,
+      "loss": 0.5995,
+      "step": 2181
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.480759863614223e-07,
+      "loss": 0.0061,
+      "step": 2182
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.480516317584023e-07,
+      "loss": 0.0064,
+      "step": 2183
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.4802727715538236e-07,
+      "loss": 0.0062,
+      "step": 2184
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.4800292255236236e-07,
+      "loss": 0.0061,
+      "step": 2185
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.479785679493424e-07,
+      "loss": 0.0143,
+      "step": 2186
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.4795421334632246e-07,
+      "loss": 0.0079,
+      "step": 2187
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.4792985874330246e-07,
+      "loss": 0.0072,
+      "step": 2188
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.479055041402825e-07,
+      "loss": 0.0082,
+      "step": 2189
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.478811495372625e-07,
+      "loss": 0.9406,
+      "step": 2190
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.4785679493424256e-07,
+      "loss": 0.0083,
+      "step": 2191
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.4783244033122256e-07,
+      "loss": 0.0198,
+      "step": 2192
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.478080857282026e-07,
+      "loss": 0.0062,
+      "step": 2193
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.477837311251826e-07,
+      "loss": 0.0065,
+      "step": 2194
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.4775937652216267e-07,
+      "loss": 0.0223,
+      "step": 2195
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.477350219191427e-07,
+      "loss": 0.0063,
+      "step": 2196
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.477106673161227e-07,
+      "loss": 0.0075,
+      "step": 2197
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.4768631271310277e-07,
+      "loss": 0.0099,
+      "step": 2198
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.4766195811008277e-07,
+      "loss": 0.0176,
+      "step": 2199
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.476376035070628e-07,
+      "loss": 0.0066,
+      "step": 2200
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.476132489040428e-07,
+      "loss": 0.0075,
+      "step": 2201
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.4758889430102287e-07,
+      "loss": 0.008,
+      "step": 2202
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.475645396980029e-07,
+      "loss": 0.012,
+      "step": 2203
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.475401850949829e-07,
+      "loss": 0.0073,
+      "step": 2204
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.4751583049196297e-07,
+      "loss": 0.0075,
+      "step": 2205
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.4749147588894297e-07,
+      "loss": 0.0073,
+      "step": 2206
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.47467121285923e-07,
+      "loss": 0.0068,
+      "step": 2207
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.47442766682903e-07,
+      "loss": 0.0065,
+      "step": 2208
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.4741841207988307e-07,
+      "loss": 0.0091,
+      "step": 2209
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.473940574768631e-07,
+      "loss": 0.009,
+      "step": 2210
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.473697028738431e-07,
+      "loss": 0.017,
+      "step": 2211
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.4734534827082317e-07,
+      "loss": 0.007,
+      "step": 2212
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.4732099366780317e-07,
+      "loss": 0.0168,
+      "step": 2213
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.472966390647832e-07,
+      "loss": 0.0065,
+      "step": 2214
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.472722844617632e-07,
+      "loss": 0.0089,
+      "step": 2215
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.4724792985874327e-07,
+      "loss": 0.0065,
+      "step": 2216
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.4722357525572327e-07,
+      "loss": 0.0075,
+      "step": 2217
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.471992206527033e-07,
+      "loss": 0.007,
+      "step": 2218
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.471748660496834e-07,
+      "loss": 0.006,
+      "step": 2219
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.471505114466634e-07,
+      "loss": 0.0069,
+      "step": 2220
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.471261568436434e-07,
+      "loss": 0.0071,
+      "step": 2221
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.471018022406234e-07,
+      "loss": 0.0091,
+      "step": 2222
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.470774476376035e-07,
+      "loss": 0.0069,
+      "step": 2223
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.470530930345835e-07,
+      "loss": 0.0075,
+      "step": 2224
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.470287384315635e-07,
+      "loss": 0.0068,
+      "step": 2225
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.470043838285436e-07,
+      "loss": 0.0077,
+      "step": 2226
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.469800292255236e-07,
+      "loss": 0.0065,
+      "step": 2227
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.4695567462250363e-07,
+      "loss": 0.0052,
+      "step": 2228
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.4693132001948363e-07,
+      "loss": 0.006,
+      "step": 2229
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.469069654164637e-07,
+      "loss": 0.0088,
+      "step": 2230
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.468826108134437e-07,
+      "loss": 1.2535,
+      "step": 2231
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.4685825621042373e-07,
+      "loss": 1.437,
+      "step": 2232
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.4683390160740373e-07,
+      "loss": 0.0071,
+      "step": 2233
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.4680954700438383e-07,
+      "loss": 0.0075,
+      "step": 2234
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.467851924013639e-07,
+      "loss": 0.0074,
+      "step": 2235
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.467608377983439e-07,
+      "loss": 0.0062,
+      "step": 2236
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.4673648319532393e-07,
+      "loss": 0.0074,
+      "step": 2237
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.4671212859230393e-07,
+      "loss": 0.0109,
+      "step": 2238
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.46687773989284e-07,
+      "loss": 0.0068,
+      "step": 2239
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.46663419386264e-07,
+      "loss": 0.007,
+      "step": 2240
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.4663906478324403e-07,
+      "loss": 0.0066,
+      "step": 2241
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.466147101802241e-07,
+      "loss": 0.0093,
+      "step": 2242
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.465903555772041e-07,
+      "loss": 0.0065,
+      "step": 2243
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.4656600097418414e-07,
+      "loss": 0.0062,
+      "step": 2244
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.4654164637116414e-07,
+      "loss": 0.0069,
+      "step": 2245
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.465172917681442e-07,
+      "loss": 0.0092,
+      "step": 2246
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.464929371651242e-07,
+      "loss": 0.007,
+      "step": 2247
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.4646858256210424e-07,
+      "loss": 0.0087,
+      "step": 2248
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.464442279590843e-07,
+      "loss": 0.0066,
+      "step": 2249
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.464198733560643e-07,
+      "loss": 0.0064,
+      "step": 2250
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.4639551875304434e-07,
+      "loss": 0.013,
+      "step": 2251
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.4637116415002434e-07,
+      "loss": 0.0088,
+      "step": 2252
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.463468095470044e-07,
+      "loss": 0.0056,
+      "step": 2253
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.463224549439844e-07,
+      "loss": 0.0108,
+      "step": 2254
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.4629810034096444e-07,
+      "loss": 0.0076,
+      "step": 2255
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.4627374573794444e-07,
+      "loss": 0.0071,
+      "step": 2256
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.462493911349245e-07,
+      "loss": 0.0063,
+      "step": 2257
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.4622503653190454e-07,
+      "loss": 0.0077,
+      "step": 2258
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.4620068192888454e-07,
+      "loss": 0.0062,
+      "step": 2259
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.461763273258646e-07,
+      "loss": 0.008,
+      "step": 2260
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.461519727228446e-07,
+      "loss": 0.0073,
+      "step": 2261
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.4612761811982464e-07,
+      "loss": 0.2115,
+      "step": 2262
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.4610326351680464e-07,
+      "loss": 0.0059,
+      "step": 2263
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.460789089137847e-07,
+      "loss": 0.0092,
+      "step": 2264
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.4605455431076474e-07,
+      "loss": 0.0063,
+      "step": 2265
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.4603019970774474e-07,
+      "loss": 0.0057,
+      "step": 2266
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.460058451047248e-07,
+      "loss": 0.0077,
+      "step": 2267
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.459814905017048e-07,
+      "loss": 0.0065,
+      "step": 2268
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.4595713589868485e-07,
+      "loss": 0.0104,
+      "step": 2269
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.4593278129566484e-07,
+      "loss": 0.0072,
+      "step": 2270
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.459084266926449e-07,
+      "loss": 0.0085,
+      "step": 2271
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.458840720896249e-07,
+      "loss": 0.0071,
+      "step": 2272
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.4585971748660495e-07,
+      "loss": 0.9809,
+      "step": 2273
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.45835362883585e-07,
+      "loss": 0.008,
+      "step": 2274
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.45811008280565e-07,
+      "loss": 0.0063,
+      "step": 2275
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.4578665367754505e-07,
+      "loss": 0.0058,
+      "step": 2276
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.4576229907452505e-07,
+      "loss": 0.3457,
+      "step": 2277
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.457379444715051e-07,
+      "loss": 0.0107,
+      "step": 2278
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.457135898684851e-07,
+      "loss": 0.0068,
+      "step": 2279
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.4568923526546515e-07,
+      "loss": 0.0073,
+      "step": 2280
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.456648806624452e-07,
+      "loss": 0.0065,
+      "step": 2281
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.456405260594252e-07,
+      "loss": 0.0074,
+      "step": 2282
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.4561617145640525e-07,
+      "loss": 0.006,
+      "step": 2283
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.4559181685338525e-07,
+      "loss": 0.0063,
+      "step": 2284
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.455674622503653e-07,
+      "loss": 0.0096,
+      "step": 2285
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.455431076473453e-07,
+      "loss": 0.0059,
+      "step": 2286
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.4551875304432535e-07,
+      "loss": 0.0085,
+      "step": 2287
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.4549439844130535e-07,
+      "loss": 0.0162,
+      "step": 2288
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.454700438382854e-07,
+      "loss": 0.0066,
+      "step": 2289
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.4544568923526545e-07,
+      "loss": 0.0072,
+      "step": 2290
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.4542133463224545e-07,
+      "loss": 0.007,
+      "step": 2291
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.453969800292255e-07,
+      "loss": 0.0058,
+      "step": 2292
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.453726254262055e-07,
+      "loss": 0.0078,
+      "step": 2293
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.4534827082318555e-07,
+      "loss": 0.0071,
+      "step": 2294
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.4532391622016555e-07,
+      "loss": 0.0103,
+      "step": 2295
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.452995616171456e-07,
+      "loss": 0.0344,
+      "step": 2296
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.4527520701412566e-07,
+      "loss": 0.007,
+      "step": 2297
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.4525085241110565e-07,
+      "loss": 0.005,
+      "step": 2298
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.452264978080857e-07,
+      "loss": 0.006,
+      "step": 2299
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.452021432050657e-07,
+      "loss": 0.0071,
+      "step": 2300
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.4517778860204576e-07,
+      "loss": 0.0053,
+      "step": 2301
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.4515343399902576e-07,
+      "loss": 0.006,
+      "step": 2302
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.451290793960058e-07,
+      "loss": 0.0068,
+      "step": 2303
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.451047247929858e-07,
+      "loss": 0.0078,
+      "step": 2304
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.4508037018996586e-07,
+      "loss": 0.007,
+      "step": 2305
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.450560155869459e-07,
+      "loss": 0.0074,
+      "step": 2306
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.450316609839259e-07,
+      "loss": 0.6067,
+      "step": 2307
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.4500730638090596e-07,
+      "loss": 0.0073,
+      "step": 2308
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.4498295177788596e-07,
+      "loss": 0.9671,
+      "step": 2309
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.4495859717486606e-07,
+      "loss": 0.0067,
+      "step": 2310
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.4493424257184606e-07,
+      "loss": 0.008,
+      "step": 2311
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.449098879688261e-07,
+      "loss": 0.0054,
+      "step": 2312
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.4488553336580616e-07,
+      "loss": 0.0075,
+      "step": 2313
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.4486117876278616e-07,
+      "loss": 0.0519,
+      "step": 2314
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.448368241597662e-07,
+      "loss": 0.0068,
+      "step": 2315
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.448124695567462e-07,
+      "loss": 0.0062,
+      "step": 2316
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.4478811495372626e-07,
+      "loss": 0.0065,
+      "step": 2317
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.4476376035070626e-07,
+      "loss": 0.0072,
+      "step": 2318
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.447394057476863e-07,
+      "loss": 0.0072,
+      "step": 2319
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.4471505114466637e-07,
+      "loss": 0.0083,
+      "step": 2320
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.4469069654164637e-07,
+      "loss": 0.0077,
+      "step": 2321
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.446663419386264e-07,
+      "loss": 0.0062,
+      "step": 2322
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.446419873356064e-07,
+      "loss": 0.0052,
+      "step": 2323
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.4461763273258647e-07,
+      "loss": 0.0058,
+      "step": 2324
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.4459327812956647e-07,
+      "loss": 0.0067,
+      "step": 2325
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.445689235265465e-07,
+      "loss": 0.0068,
+      "step": 2326
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.445445689235265e-07,
+      "loss": 0.0182,
+      "step": 2327
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.4452021432050657e-07,
+      "loss": 0.0065,
+      "step": 2328
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.444958597174866e-07,
+      "loss": 0.0084,
+      "step": 2329
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.444715051144666e-07,
+      "loss": 0.0052,
+      "step": 2330
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.4444715051144667e-07,
+      "loss": 0.0091,
+      "step": 2331
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.4442279590842667e-07,
+      "loss": 0.0065,
+      "step": 2332
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.443984413054067e-07,
+      "loss": 0.0145,
+      "step": 2333
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.443740867023867e-07,
+      "loss": 0.0058,
+      "step": 2334
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.4434973209936677e-07,
+      "loss": 0.0057,
+      "step": 2335
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.443253774963468e-07,
+      "loss": 0.0069,
+      "step": 2336
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.443010228933268e-07,
+      "loss": 0.0057,
+      "step": 2337
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.4427666829030687e-07,
+      "loss": 0.0075,
+      "step": 2338
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.4425231368728687e-07,
+      "loss": 0.0047,
+      "step": 2339
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.442279590842669e-07,
+      "loss": 0.0073,
+      "step": 2340
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.442036044812469e-07,
+      "loss": 0.0062,
+      "step": 2341
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.4417924987822697e-07,
+      "loss": 0.0071,
+      "step": 2342
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.4415489527520697e-07,
+      "loss": 0.0075,
+      "step": 2343
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.44130540672187e-07,
+      "loss": 0.007,
+      "step": 2344
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.441061860691671e-07,
+      "loss": 0.0294,
+      "step": 2345
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.440818314661471e-07,
+      "loss": 0.0163,
+      "step": 2346
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.440574768631271e-07,
+      "loss": 0.0076,
+      "step": 2347
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.440331222601071e-07,
+      "loss": 0.0062,
+      "step": 2348
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.440087676570872e-07,
+      "loss": 0.0125,
+      "step": 2349
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.439844130540672e-07,
+      "loss": 0.0063,
+      "step": 2350
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.4396005845104723e-07,
+      "loss": 0.0068,
+      "step": 2351
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.439357038480273e-07,
+      "loss": 0.0339,
+      "step": 2352
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.439113492450073e-07,
+      "loss": 0.0103,
+      "step": 2353
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.4388699464198733e-07,
+      "loss": 0.0086,
+      "step": 2354
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.4386264003896733e-07,
+      "loss": 0.0073,
+      "step": 2355
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.438382854359474e-07,
+      "loss": 0.0064,
+      "step": 2356
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.438139308329274e-07,
+      "loss": 0.0068,
+      "step": 2357
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.4378957622990743e-07,
+      "loss": 0.008,
+      "step": 2358
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.4376522162688743e-07,
+      "loss": 0.005,
+      "step": 2359
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.437408670238675e-07,
+      "loss": 0.0057,
+      "step": 2360
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.4371651242084753e-07,
+      "loss": 0.0053,
+      "step": 2361
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.4369215781782753e-07,
+      "loss": 0.0059,
+      "step": 2362
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.436678032148076e-07,
+      "loss": 0.0062,
+      "step": 2363
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.436434486117876e-07,
+      "loss": 0.0067,
+      "step": 2364
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.4361909400876763e-07,
+      "loss": 0.0057,
+      "step": 2365
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.4359473940574763e-07,
+      "loss": 0.0059,
+      "step": 2366
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.435703848027277e-07,
+      "loss": 0.006,
+      "step": 2367
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.4354603019970773e-07,
+      "loss": 0.0077,
+      "step": 2368
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.4352167559668773e-07,
+      "loss": 0.0065,
+      "step": 2369
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.434973209936678e-07,
+      "loss": 0.0156,
+      "step": 2370
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.434729663906478e-07,
+      "loss": 0.0065,
+      "step": 2371
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.4344861178762783e-07,
+      "loss": 0.0062,
+      "step": 2372
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.4342425718460783e-07,
+      "loss": 0.0073,
+      "step": 2373
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.433999025815879e-07,
+      "loss": 0.0082,
+      "step": 2374
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.433755479785679e-07,
+      "loss": 0.0091,
+      "step": 2375
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.4335119337554794e-07,
+      "loss": 0.0052,
+      "step": 2376
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.43326838772528e-07,
+      "loss": 0.0061,
+      "step": 2377
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.43302484169508e-07,
+      "loss": 0.0056,
+      "step": 2378
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.4327812956648804e-07,
+      "loss": 0.0073,
+      "step": 2379
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.4325377496346804e-07,
+      "loss": 0.0195,
+      "step": 2380
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.432294203604481e-07,
+      "loss": 0.0058,
+      "step": 2381
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.432050657574281e-07,
+      "loss": 0.0068,
+      "step": 2382
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.4318071115440814e-07,
+      "loss": 0.0058,
+      "step": 2383
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.431563565513882e-07,
+      "loss": 0.0062,
+      "step": 2384
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.431320019483682e-07,
+      "loss": 0.0072,
+      "step": 2385
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.4310764734534824e-07,
+      "loss": 0.0045,
+      "step": 2386
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.430832927423283e-07,
+      "loss": 0.0083,
+      "step": 2387
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.4305893813930834e-07,
+      "loss": 0.0063,
+      "step": 2388
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.4303458353628834e-07,
+      "loss": 0.1428,
+      "step": 2389
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.430102289332684e-07,
+      "loss": 0.0062,
+      "step": 2390
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.4298587433024844e-07,
+      "loss": 0.006,
+      "step": 2391
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.4296151972722844e-07,
+      "loss": 0.006,
+      "step": 2392
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.429371651242085e-07,
+      "loss": 0.0074,
+      "step": 2393
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.429128105211885e-07,
+      "loss": 0.0075,
+      "step": 2394
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4288845591816855e-07,
+      "loss": 0.008,
+      "step": 2395
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4286410131514854e-07,
+      "loss": 0.0056,
+      "step": 2396
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.428397467121286e-07,
+      "loss": 0.0058,
+      "step": 2397
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.428153921091086e-07,
+      "loss": 0.0062,
+      "step": 2398
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4279103750608865e-07,
+      "loss": 0.0062,
+      "step": 2399
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.427666829030687e-07,
+      "loss": 0.0073,
+      "step": 2400
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.427423283000487e-07,
+      "loss": 0.0176,
+      "step": 2401
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4271797369702875e-07,
+      "loss": 0.0058,
+      "step": 2402
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4269361909400875e-07,
+      "loss": 0.0591,
+      "step": 2403
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.426692644909888e-07,
+      "loss": 0.8404,
+      "step": 2404
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.426449098879688e-07,
+      "loss": 0.006,
+      "step": 2405
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4262055528494885e-07,
+      "loss": 0.0072,
+      "step": 2406
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.425962006819289e-07,
+      "loss": 0.0051,
+      "step": 2407
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.425718460789089e-07,
+      "loss": 0.0054,
+      "step": 2408
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4254749147588895e-07,
+      "loss": 0.0066,
+      "step": 2409
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4252313687286895e-07,
+      "loss": 0.0061,
+      "step": 2410
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.42498782269849e-07,
+      "loss": 0.0066,
+      "step": 2411
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.42474427666829e-07,
+      "loss": 0.0066,
+      "step": 2412
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4245007306380905e-07,
+      "loss": 0.004,
+      "step": 2413
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.424257184607891e-07,
+      "loss": 0.0071,
+      "step": 2414
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.424013638577691e-07,
+      "loss": 0.0056,
+      "step": 2415
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.4237700925474915e-07,
+      "loss": 0.0589,
+      "step": 2416
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.4235265465172915e-07,
+      "loss": 0.0107,
+      "step": 2417
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.423283000487092e-07,
+      "loss": 0.0105,
+      "step": 2418
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.423039454456892e-07,
+      "loss": 0.0124,
+      "step": 2419
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.4227959084266925e-07,
+      "loss": 1.4365,
+      "step": 2420
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.4225523623964925e-07,
+      "loss": 0.0087,
+      "step": 2421
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.422308816366293e-07,
+      "loss": 0.0068,
+      "step": 2422
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.4220652703360936e-07,
+      "loss": 0.006,
+      "step": 2423
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.4218217243058935e-07,
+      "loss": 0.0059,
+      "step": 2424
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.421578178275694e-07,
+      "loss": 0.007,
+      "step": 2425
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.421334632245494e-07,
+      "loss": 0.0061,
+      "step": 2426
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.4210910862152946e-07,
+      "loss": 0.0093,
+      "step": 2427
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.4208475401850946e-07,
+      "loss": 0.0056,
+      "step": 2428
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.420603994154895e-07,
+      "loss": 0.006,
+      "step": 2429
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.4203604481246956e-07,
+      "loss": 1.262,
+      "step": 2430
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.4201169020944956e-07,
+      "loss": 0.0069,
+      "step": 2431
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.419873356064296e-07,
+      "loss": 0.0054,
+      "step": 2432
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.419629810034096e-07,
+      "loss": 0.0061,
+      "step": 2433
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.4193862640038966e-07,
+      "loss": 0.011,
+      "step": 2434
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.4191427179736966e-07,
+      "loss": 0.0056,
+      "step": 2435
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.418899171943497e-07,
+      "loss": 0.0055,
+      "step": 2436
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.418655625913297e-07,
+      "loss": 0.0042,
+      "step": 2437
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.4184120798830976e-07,
+      "loss": 0.005,
+      "step": 2438
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.418168533852898e-07,
+      "loss": 0.0057,
+      "step": 2439
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.417924987822698e-07,
+      "loss": 0.0057,
+      "step": 2440
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.4176814417924986e-07,
+      "loss": 0.0062,
+      "step": 2441
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.4174378957622986e-07,
+      "loss": 0.0063,
+      "step": 2442
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.417194349732099e-07,
+      "loss": 0.0056,
+      "step": 2443
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.416950803701899e-07,
+      "loss": 0.006,
+      "step": 2444
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.4167072576716996e-07,
+      "loss": 0.0043,
+      "step": 2445
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.4164637116415e-07,
+      "loss": 0.0058,
+      "step": 2446
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.4162201656113e-07,
+      "loss": 0.0059,
+      "step": 2447
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.4159766195811006e-07,
+      "loss": 0.0056,
+      "step": 2448
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.4157330735509006e-07,
+      "loss": 0.0063,
+      "step": 2449
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.415489527520701e-07,
+      "loss": 0.0075,
+      "step": 2450
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.415245981490501e-07,
+      "loss": 0.007,
+      "step": 2451
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.4150024354603017e-07,
+      "loss": 0.0054,
+      "step": 2452
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.4147588894301016e-07,
+      "loss": 1.8193,
+      "step": 2453
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.414515343399902e-07,
+      "loss": 0.0054,
+      "step": 2454
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.4142717973697027e-07,
+      "loss": 0.0059,
+      "step": 2455
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.4140282513395027e-07,
+      "loss": 0.0064,
+      "step": 2456
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.413784705309303e-07,
+      "loss": 0.0045,
+      "step": 2457
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.413541159279103e-07,
+      "loss": 0.0058,
+      "step": 2458
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.4132976132489037e-07,
+      "loss": 0.005,
+      "step": 2459
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.4130540672187037e-07,
+      "loss": 0.0053,
+      "step": 2460
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.412810521188504e-07,
+      "loss": 0.0059,
+      "step": 2461
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.4125669751583047e-07,
+      "loss": 0.007,
+      "step": 2462
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.4123234291281047e-07,
+      "loss": 0.0062,
+      "step": 2463
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.4120798830979057e-07,
+      "loss": 0.0089,
+      "step": 2464
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.4118363370677057e-07,
+      "loss": 0.0047,
+      "step": 2465
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.411592791037506e-07,
+      "loss": 0.0097,
+      "step": 2466
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.411349245007306e-07,
+      "loss": 0.0065,
+      "step": 2467
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.411105698977107e-07,
+      "loss": 0.0062,
+      "step": 2468
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.410862152946907e-07,
+      "loss": 0.0047,
+      "step": 2469
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.410618606916707e-07,
+      "loss": 0.0062,
+      "step": 2470
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.410375060886508e-07,
+      "loss": 0.0105,
+      "step": 2471
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.410131514856308e-07,
+      "loss": 0.0049,
+      "step": 2472
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.409887968826108e-07,
+      "loss": 0.0053,
+      "step": 2473
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.409644422795908e-07,
+      "loss": 0.007,
+      "step": 2474
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.409400876765709e-07,
+      "loss": 0.0361,
+      "step": 2475
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.409157330735509e-07,
+      "loss": 0.0059,
+      "step": 2476
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.4089137847053093e-07,
+      "loss": 0.0051,
+      "step": 2477
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.40867023867511e-07,
+      "loss": 0.0086,
+      "step": 2478
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.40842669264491e-07,
+      "loss": 0.0058,
+      "step": 2479
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.4081831466147103e-07,
+      "loss": 0.0064,
+      "step": 2480
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.4079396005845103e-07,
+      "loss": 0.0077,
+      "step": 2481
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.407696054554311e-07,
+      "loss": 0.0048,
+      "step": 2482
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.407452508524111e-07,
+      "loss": 0.0778,
+      "step": 2483
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.4072089624939113e-07,
+      "loss": 0.0053,
+      "step": 2484
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 4.406965416463712e-07,
+      "loss": 0.0098,
+      "step": 2485
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.406721870433512e-07,
+      "loss": 0.0139,
+      "step": 2486
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.4064783244033123e-07,
+      "loss": 0.0054,
+      "step": 2487
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.4062347783731123e-07,
+      "loss": 0.0065,
+      "step": 2488
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.405991232342913e-07,
+      "loss": 0.0047,
+      "step": 2489
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.405747686312713e-07,
+      "loss": 0.0059,
+      "step": 2490
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.4055041402825133e-07,
+      "loss": 0.0053,
+      "step": 2491
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.4052605942523133e-07,
+      "loss": 0.0062,
+      "step": 2492
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.405017048222114e-07,
+      "loss": 0.0052,
+      "step": 2493
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.4047735021919143e-07,
+      "loss": 0.0054,
+      "step": 2494
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.4045299561617143e-07,
+      "loss": 0.0054,
+      "step": 2495
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.404286410131515e-07,
+      "loss": 0.0049,
+      "step": 2496
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.404042864101315e-07,
+      "loss": 0.006,
+      "step": 2497
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.4037993180711153e-07,
+      "loss": 0.0062,
+      "step": 2498
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.4035557720409153e-07,
+      "loss": 0.0059,
+      "step": 2499
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.403312226010716e-07,
+      "loss": 0.0208,
+      "step": 2500
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.4030686799805164e-07,
+      "loss": 0.0045,
+      "step": 2501
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.4028251339503164e-07,
+      "loss": 0.0046,
+      "step": 2502
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.402581587920117e-07,
+      "loss": 0.0531,
+      "step": 2503
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.402338041889917e-07,
+      "loss": 0.0082,
+      "step": 2504
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 4.4020944958597174e-07,
+      "loss": 0.005,
+      "step": 2505
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.4018509498295174e-07,
+      "loss": 0.0121,
+      "step": 2506
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.401607403799318e-07,
+      "loss": 0.0049,
+      "step": 2507
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.401363857769118e-07,
+      "loss": 0.0065,
+      "step": 2508
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.4011203117389184e-07,
+      "loss": 0.0055,
+      "step": 2509
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.400876765708719e-07,
+      "loss": 0.0064,
+      "step": 2510
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.400633219678519e-07,
+      "loss": 0.0059,
+      "step": 2511
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.4003896736483194e-07,
+      "loss": 0.4848,
+      "step": 2512
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.4001461276181194e-07,
+      "loss": 0.0051,
+      "step": 2513
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.39990258158792e-07,
+      "loss": 0.005,
+      "step": 2514
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 4.39965903555772e-07,
+      "loss": 0.0061,
+      "step": 2515
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.3994154895275204e-07,
+      "loss": 0.0043,
+      "step": 2516
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.399171943497321e-07,
+      "loss": 1.03,
+      "step": 2517
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.398928397467121e-07,
+      "loss": 0.0069,
+      "step": 2518
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.3986848514369214e-07,
+      "loss": 0.0055,
+      "step": 2519
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.3984413054067214e-07,
+      "loss": 0.0055,
+      "step": 2520
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.398197759376522e-07,
+      "loss": 0.0053,
+      "step": 2521
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.397954213346322e-07,
+      "loss": 0.0057,
+      "step": 2522
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.3977106673161224e-07,
+      "loss": 0.0064,
+      "step": 2523
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.3974671212859224e-07,
+      "loss": 0.004,
+      "step": 2524
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.397223575255723e-07,
+      "loss": 0.0058,
+      "step": 2525
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 4.3969800292255235e-07,
+      "loss": 0.0057,
+      "step": 2526
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.3967364831953234e-07,
+      "loss": 0.0054,
+      "step": 2527
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.396492937165124e-07,
+      "loss": 0.0112,
+      "step": 2528
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.396249391134924e-07,
+      "loss": 0.0102,
+      "step": 2529
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.3960058451047245e-07,
+      "loss": 0.0076,
+      "step": 2530
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.3957622990745244e-07,
+      "loss": 0.0195,
+      "step": 2531
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.395518753044325e-07,
+      "loss": 0.0054,
+      "step": 2532
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.3952752070141255e-07,
+      "loss": 0.0062,
+      "step": 2533
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.3950316609839255e-07,
+      "loss": 0.0049,
+      "step": 2534
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.394788114953726e-07,
+      "loss": 0.0053,
+      "step": 2535
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 4.394544568923526e-07,
+      "loss": 0.0175,
+      "step": 2536
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.3943010228933265e-07,
+      "loss": 0.0054,
+      "step": 2537
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.3940574768631265e-07,
+      "loss": 0.0054,
+      "step": 2538
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.393813930832927e-07,
+      "loss": 1.0889,
+      "step": 2539
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.393570384802728e-07,
+      "loss": 0.0053,
+      "step": 2540
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.393326838772528e-07,
+      "loss": 0.0357,
+      "step": 2541
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.3930832927423285e-07,
+      "loss": 0.0042,
+      "step": 2542
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.3928397467121285e-07,
+      "loss": 0.0065,
+      "step": 2543
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.392596200681929e-07,
+      "loss": 0.0044,
+      "step": 2544
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.392352654651729e-07,
+      "loss": 0.0071,
+      "step": 2545
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 4.3921091086215295e-07,
+      "loss": 0.005,
+      "step": 2546
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.3918655625913295e-07,
+      "loss": 0.0058,
+      "step": 2547
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.39162201656113e-07,
+      "loss": 0.0056,
+      "step": 2548
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.3913784705309306e-07,
+      "loss": 0.0054,
+      "step": 2549
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.3911349245007305e-07,
+      "loss": 0.0051,
+      "step": 2550
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.390891378470531e-07,
+      "loss": 0.0045,
+      "step": 2551
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.390647832440331e-07,
+      "loss": 0.0052,
+      "step": 2552
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.3904042864101316e-07,
+      "loss": 0.0054,
+      "step": 2553
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.3901607403799316e-07,
+      "loss": 0.0108,
+      "step": 2554
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.389917194349732e-07,
+      "loss": 0.0049,
+      "step": 2555
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.3896736483195326e-07,
+      "loss": 0.005,
+      "step": 2556
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 4.3894301022893326e-07,
+      "loss": 0.0055,
+      "step": 2557
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.389186556259133e-07,
+      "loss": 0.0048,
+      "step": 2558
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.388943010228933e-07,
+      "loss": 0.0077,
+      "step": 2559
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.3886994641987336e-07,
+      "loss": 0.0066,
+      "step": 2560
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.3884559181685336e-07,
+      "loss": 0.0049,
+      "step": 2561
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.388212372138334e-07,
+      "loss": 0.005,
+      "step": 2562
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.387968826108134e-07,
+      "loss": 0.0047,
+      "step": 2563
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.3877252800779346e-07,
+      "loss": 0.005,
+      "step": 2564
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.387481734047735e-07,
+      "loss": 0.0051,
+      "step": 2565
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.387238188017535e-07,
+      "loss": 0.0058,
+      "step": 2566
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 4.3869946419873356e-07,
+      "loss": 0.0056,
+      "step": 2567
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.3867510959571356e-07,
+      "loss": 0.0049,
+      "step": 2568
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.386507549926936e-07,
+      "loss": 0.0042,
+      "step": 2569
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.386264003896736e-07,
+      "loss": 0.0047,
+      "step": 2570
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.3860204578665366e-07,
+      "loss": 0.0044,
+      "step": 2571
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.385776911836337e-07,
+      "loss": 0.0066,
+      "step": 2572
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.385533365806137e-07,
+      "loss": 0.0069,
+      "step": 2573
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.3852898197759376e-07,
+      "loss": 0.0053,
+      "step": 2574
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.3850462737457376e-07,
+      "loss": 1.2109,
+      "step": 2575
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.384802727715538e-07,
+      "loss": 0.0176,
+      "step": 2576
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 4.384559181685338e-07,
+      "loss": 0.0049,
+      "step": 2577
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.3843156356551387e-07,
+      "loss": 0.0061,
+      "step": 2578
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.384072089624939e-07,
+      "loss": 0.008,
+      "step": 2579
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.383828543594739e-07,
+      "loss": 0.0041,
+      "step": 2580
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.3835849975645397e-07,
+      "loss": 0.0882,
+      "step": 2581
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.3833414515343397e-07,
+      "loss": 0.0046,
+      "step": 2582
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.38309790550414e-07,
+      "loss": 0.0046,
+      "step": 2583
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.38285435947394e-07,
+      "loss": 0.0047,
+      "step": 2584
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.3826108134437407e-07,
+      "loss": 0.0055,
+      "step": 2585
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.3823672674135407e-07,
+      "loss": 0.7578,
+      "step": 2586
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 4.382123721383341e-07,
+      "loss": 0.0048,
+      "step": 2587
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.3818801753531417e-07,
+      "loss": 0.0048,
+      "step": 2588
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.3816366293229417e-07,
+      "loss": 0.0048,
+      "step": 2589
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.381393083292742e-07,
+      "loss": 0.0041,
+      "step": 2590
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.381149537262542e-07,
+      "loss": 0.0058,
+      "step": 2591
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.3809059912323427e-07,
+      "loss": 0.0062,
+      "step": 2592
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.3806624452021427e-07,
+      "loss": 0.0048,
+      "step": 2593
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.380418899171943e-07,
+      "loss": 0.008,
+      "step": 2594
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.3801753531417437e-07,
+      "loss": 0.0049,
+      "step": 2595
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.3799318071115437e-07,
+      "loss": 0.0047,
+      "step": 2596
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.379688261081344e-07,
+      "loss": 0.0047,
+      "step": 2597
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 4.379444715051144e-07,
+      "loss": 0.0052,
+      "step": 2598
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.3792011690209447e-07,
+      "loss": 0.0057,
+      "step": 2599
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.3789576229907447e-07,
+      "loss": 0.0049,
+      "step": 2600
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.378714076960545e-07,
+      "loss": 0.0155,
+      "step": 2601
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.378470530930345e-07,
+      "loss": 0.005,
+      "step": 2602
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.378226984900146e-07,
+      "loss": 0.0047,
+      "step": 2603
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.377983438869946e-07,
+      "loss": 0.0048,
+      "step": 2604
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.377739892839746e-07,
+      "loss": 0.0048,
+      "step": 2605
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.377496346809547e-07,
+      "loss": 0.0057,
+      "step": 2606
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.377252800779347e-07,
+      "loss": 0.0051,
+      "step": 2607
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 4.377009254749147e-07,
+      "loss": 0.006,
+      "step": 2608
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.376765708718947e-07,
+      "loss": 0.0071,
+      "step": 2609
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.376522162688748e-07,
+      "loss": 0.0051,
+      "step": 2610
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.3762786166585483e-07,
+      "loss": 0.0055,
+      "step": 2611
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.3760350706283483e-07,
+      "loss": 0.0068,
+      "step": 2612
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.375791524598149e-07,
+      "loss": 0.0054,
+      "step": 2613
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.375547978567949e-07,
+      "loss": 0.0232,
+      "step": 2614
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.3753044325377493e-07,
+      "loss": 0.0046,
+      "step": 2615
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.3750608865075493e-07,
+      "loss": 0.0307,
+      "step": 2616
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.3748173404773503e-07,
+      "loss": 0.0049,
+      "step": 2617
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 4.374573794447151e-07,
+      "loss": 0.0047,
+      "step": 2618
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.374330248416951e-07,
+      "loss": 0.0046,
+      "step": 2619
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.3740867023867513e-07,
+      "loss": 0.0054,
+      "step": 2620
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.3738431563565513e-07,
+      "loss": 0.0045,
+      "step": 2621
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.373599610326352e-07,
+      "loss": 0.0044,
+      "step": 2622
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.373356064296152e-07,
+      "loss": 0.0042,
+      "step": 2623
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.3731125182659523e-07,
+      "loss": 0.0046,
+      "step": 2624
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.3728689722357523e-07,
+      "loss": 0.005,
+      "step": 2625
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.372625426205553e-07,
+      "loss": 0.0054,
+      "step": 2626
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.3723818801753534e-07,
+      "loss": 0.0049,
+      "step": 2627
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.3721383341451534e-07,
+      "loss": 0.0059,
+      "step": 2628
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 4.371894788114954e-07,
+      "loss": 0.0053,
+      "step": 2629
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.371651242084754e-07,
+      "loss": 0.0158,
+      "step": 2630
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.3714076960545544e-07,
+      "loss": 0.0041,
+      "step": 2631
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.3711641500243544e-07,
+      "loss": 0.0047,
+      "step": 2632
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.370920603994155e-07,
+      "loss": 0.0055,
+      "step": 2633
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.3706770579639554e-07,
+      "loss": 0.0041,
+      "step": 2634
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.3704335119337554e-07,
+      "loss": 0.0052,
+      "step": 2635
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.370189965903556e-07,
+      "loss": 0.0046,
+      "step": 2636
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.369946419873356e-07,
+      "loss": 0.0045,
+      "step": 2637
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.3697028738431564e-07,
+      "loss": 0.0077,
+      "step": 2638
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 4.3694593278129564e-07,
+      "loss": 0.0038,
+      "step": 2639
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.369215781782757e-07,
+      "loss": 0.0038,
+      "step": 2640
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.368972235752557e-07,
+      "loss": 0.0047,
+      "step": 2641
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.3687286897223574e-07,
+      "loss": 0.0053,
+      "step": 2642
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.368485143692158e-07,
+      "loss": 0.0045,
+      "step": 2643
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.368241597661958e-07,
+      "loss": 0.0054,
+      "step": 2644
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.3679980516317584e-07,
+      "loss": 0.0049,
+      "step": 2645
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.3677545056015584e-07,
+      "loss": 0.1257,
+      "step": 2646
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.367510959571359e-07,
+      "loss": 0.0039,
+      "step": 2647
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.367267413541159e-07,
+      "loss": 0.0045,
+      "step": 2648
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 4.3670238675109594e-07,
+      "loss": 0.0057,
+      "step": 2649
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.36678032148076e-07,
+      "loss": 0.0044,
+      "step": 2650
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.36653677545056e-07,
+      "loss": 0.0052,
+      "step": 2651
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.3662932294203605e-07,
+      "loss": 0.0044,
+      "step": 2652
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.3660496833901604e-07,
+      "loss": 0.0068,
+      "step": 2653
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.365806137359961e-07,
+      "loss": 0.0053,
+      "step": 2654
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.365562591329761e-07,
+      "loss": 0.0035,
+      "step": 2655
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.3653190452995615e-07,
+      "loss": 0.0051,
+      "step": 2656
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.3650754992693614e-07,
+      "loss": 0.0052,
+      "step": 2657
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.364831953239162e-07,
+      "loss": 0.0054,
+      "step": 2658
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 4.3645884072089625e-07,
+      "loss": 0.0076,
+      "step": 2659
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.3643448611787625e-07,
+      "loss": 0.0045,
+      "step": 2660
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.364101315148563e-07,
+      "loss": 0.0057,
+      "step": 2661
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.363857769118363e-07,
+      "loss": 0.0037,
+      "step": 2662
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.3636142230881635e-07,
+      "loss": 0.0046,
+      "step": 2663
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.3633706770579635e-07,
+      "loss": 0.0056,
+      "step": 2664
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.363127131027764e-07,
+      "loss": 0.0043,
+      "step": 2665
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.3628835849975645e-07,
+      "loss": 0.0047,
+      "step": 2666
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.3626400389673645e-07,
+      "loss": 0.0053,
+      "step": 2667
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.362396492937165e-07,
+      "loss": 0.0049,
+      "step": 2668
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.362152946906965e-07,
+      "loss": 0.0045,
+      "step": 2669
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 4.3619094008767655e-07,
+      "loss": 0.0056,
+      "step": 2670
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.3616658548465655e-07,
+      "loss": 0.582,
+      "step": 2671
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.361422308816366e-07,
+      "loss": 0.0047,
+      "step": 2672
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.361178762786166e-07,
+      "loss": 0.0049,
+      "step": 2673
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.3609352167559665e-07,
+      "loss": 0.0175,
+      "step": 2674
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.360691670725767e-07,
+      "loss": 0.0056,
+      "step": 2675
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.360448124695567e-07,
+      "loss": 0.0046,
+      "step": 2676
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.3602045786653675e-07,
+      "loss": 0.0052,
+      "step": 2677
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.3599610326351675e-07,
+      "loss": 0.0063,
+      "step": 2678
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.359717486604968e-07,
+      "loss": 0.0045,
+      "step": 2679
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 4.359473940574768e-07,
+      "loss": 0.0047,
+      "step": 2680
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.3592303945445685e-07,
+      "loss": 0.0037,
+      "step": 2681
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.358986848514369e-07,
+      "loss": 0.012,
+      "step": 2682
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.358743302484169e-07,
+      "loss": 0.0033,
+      "step": 2683
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.3584997564539696e-07,
+      "loss": 0.0105,
+      "step": 2684
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.3582562104237696e-07,
+      "loss": 0.0045,
+      "step": 2685
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.35801266439357e-07,
+      "loss": 0.0053,
+      "step": 2686
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.35776911836337e-07,
+      "loss": 0.0057,
+      "step": 2687
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.3575255723331706e-07,
+      "loss": 0.0041,
+      "step": 2688
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.3572820263029706e-07,
+      "loss": 0.0041,
+      "step": 2689
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 4.357038480272771e-07,
+      "loss": 0.0043,
+      "step": 2690
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.3567949342425716e-07,
+      "loss": 0.0048,
+      "step": 2691
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.3565513882123716e-07,
+      "loss": 0.0054,
+      "step": 2692
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.3563078421821726e-07,
+      "loss": 0.0089,
+      "step": 2693
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.3560642961519726e-07,
+      "loss": 0.0048,
+      "step": 2694
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.355820750121773e-07,
+      "loss": 0.0045,
+      "step": 2695
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.355577204091573e-07,
+      "loss": 0.0055,
+      "step": 2696
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.3553336580613736e-07,
+      "loss": 0.0046,
+      "step": 2697
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.355090112031174e-07,
+      "loss": 0.0053,
+      "step": 2698
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.354846566000974e-07,
+      "loss": 0.0046,
+      "step": 2699
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.3546030199707746e-07,
+      "loss": 0.0044,
+      "step": 2700
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 4.3543594739405746e-07,
+      "loss": 0.0057,
+      "step": 2701
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.354115927910375e-07,
+      "loss": 0.0041,
+      "step": 2702
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.353872381880175e-07,
+      "loss": 0.0056,
+      "step": 2703
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.3536288358499757e-07,
+      "loss": 0.0234,
+      "step": 2704
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.353385289819776e-07,
+      "loss": 0.0049,
+      "step": 2705
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.353141743789576e-07,
+      "loss": 0.0053,
+      "step": 2706
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.3528981977593767e-07,
+      "loss": 0.0101,
+      "step": 2707
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.3526546517291767e-07,
+      "loss": 0.0052,
+      "step": 2708
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.352411105698977e-07,
+      "loss": 0.0059,
+      "step": 2709
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.352167559668777e-07,
+      "loss": 0.0047,
+      "step": 2710
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 4.3519240136385777e-07,
+      "loss": 0.0055,
+      "step": 2711
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.3516804676083777e-07,
+      "loss": 0.0043,
+      "step": 2712
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.351436921578178e-07,
+      "loss": 0.0045,
+      "step": 2713
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.3511933755479787e-07,
+      "loss": 0.0042,
+      "step": 2714
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.3509498295177787e-07,
+      "loss": 0.2102,
+      "step": 2715
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.350706283487579e-07,
+      "loss": 0.0048,
+      "step": 2716
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.350462737457379e-07,
+      "loss": 0.0122,
+      "step": 2717
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.3502191914271797e-07,
+      "loss": 0.0042,
+      "step": 2718
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.3499756453969797e-07,
+      "loss": 0.0068,
+      "step": 2719
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.34973209936678e-07,
+      "loss": 0.0047,
+      "step": 2720
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 4.3494885533365807e-07,
+      "loss": 0.0057,
+      "step": 2721
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.3492450073063807e-07,
+      "loss": 0.0036,
+      "step": 2722
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.349001461276181e-07,
+      "loss": 0.0361,
+      "step": 2723
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.348757915245981e-07,
+      "loss": 0.0044,
+      "step": 2724
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.3485143692157817e-07,
+      "loss": 0.0046,
+      "step": 2725
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.3482708231855817e-07,
+      "loss": 0.0048,
+      "step": 2726
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.348027277155382e-07,
+      "loss": 0.0049,
+      "step": 2727
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.347783731125182e-07,
+      "loss": 0.0044,
+      "step": 2728
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.347540185094983e-07,
+      "loss": 0.0042,
+      "step": 2729
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.347296639064783e-07,
+      "loss": 0.004,
+      "step": 2730
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 4.347053093034583e-07,
+      "loss": 0.0046,
+      "step": 2731
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.346809547004384e-07,
+      "loss": 0.0056,
+      "step": 2732
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.346566000974184e-07,
+      "loss": 0.0048,
+      "step": 2733
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.3463224549439843e-07,
+      "loss": 0.0047,
+      "step": 2734
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.346078908913784e-07,
+      "loss": 0.0056,
+      "step": 2735
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.345835362883585e-07,
+      "loss": 0.004,
+      "step": 2736
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.3455918168533853e-07,
+      "loss": 0.0043,
+      "step": 2737
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.3453482708231853e-07,
+      "loss": 0.0051,
+      "step": 2738
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.345104724792986e-07,
+      "loss": 0.0048,
+      "step": 2739
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.344861178762786e-07,
+      "loss": 0.0042,
+      "step": 2740
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.3446176327325863e-07,
+      "loss": 0.0049,
+      "step": 2741
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 4.3443740867023863e-07,
+      "loss": 0.0041,
+      "step": 2742
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.344130540672187e-07,
+      "loss": 0.0041,
+      "step": 2743
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.343886994641987e-07,
+      "loss": 0.0041,
+      "step": 2744
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.3436434486117873e-07,
+      "loss": 0.0053,
+      "step": 2745
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.343399902581588e-07,
+      "loss": 0.0051,
+      "step": 2746
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.343156356551388e-07,
+      "loss": 0.0045,
+      "step": 2747
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.3429128105211883e-07,
+      "loss": 0.0054,
+      "step": 2748
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.3426692644909883e-07,
+      "loss": 0.0043,
+      "step": 2749
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.342425718460789e-07,
+      "loss": 0.0063,
+      "step": 2750
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.342182172430589e-07,
+      "loss": 0.0038,
+      "step": 2751
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 4.3419386264003893e-07,
+      "loss": 0.0041,
+      "step": 2752
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.34169508037019e-07,
+      "loss": 0.0051,
+      "step": 2753
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.34145153433999e-07,
+      "loss": 0.0041,
+      "step": 2754
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.3412079883097903e-07,
+      "loss": 0.0052,
+      "step": 2755
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.3409644422795903e-07,
+      "loss": 0.0158,
+      "step": 2756
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.340720896249391e-07,
+      "loss": 0.0037,
+      "step": 2757
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.340477350219191e-07,
+      "loss": 0.0046,
+      "step": 2758
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.3402338041889914e-07,
+      "loss": 0.0044,
+      "step": 2759
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.339990258158792e-07,
+      "loss": 0.0046,
+      "step": 2760
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.339746712128592e-07,
+      "loss": 0.0171,
+      "step": 2761
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 4.3395031660983924e-07,
+      "loss": 0.0035,
+      "step": 2762
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.3392596200681924e-07,
+      "loss": 0.004,
+      "step": 2763
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.339016074037993e-07,
+      "loss": 0.0048,
+      "step": 2764
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.338772528007793e-07,
+      "loss": 0.0044,
+      "step": 2765
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.3385289819775934e-07,
+      "loss": 0.1945,
+      "step": 2766
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.3382854359473934e-07,
+      "loss": 0.0052,
+      "step": 2767
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.338041889917194e-07,
+      "loss": 0.0045,
+      "step": 2768
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.3377983438869944e-07,
+      "loss": 0.0049,
+      "step": 2769
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.337554797856795e-07,
+      "loss": 0.0039,
+      "step": 2770
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.3373112518265954e-07,
+      "loss": 0.0037,
+      "step": 2771
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.3370677057963954e-07,
+      "loss": 0.0041,
+      "step": 2772
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 4.336824159766196e-07,
+      "loss": 0.0046,
+      "step": 2773
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.336580613735996e-07,
+      "loss": 0.0041,
+      "step": 2774
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.3363370677057964e-07,
+      "loss": 0.0039,
+      "step": 2775
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.336093521675597e-07,
+      "loss": 0.123,
+      "step": 2776
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.335849975645397e-07,
+      "loss": 0.0049,
+      "step": 2777
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.3356064296151975e-07,
+      "loss": 0.0039,
+      "step": 2778
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.3353628835849974e-07,
+      "loss": 0.0046,
+      "step": 2779
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.335119337554798e-07,
+      "loss": 0.0042,
+      "step": 2780
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.334875791524598e-07,
+      "loss": 0.4321,
+      "step": 2781
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.3346322454943985e-07,
+      "loss": 0.0055,
+      "step": 2782
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 4.334388699464199e-07,
+      "loss": 0.0035,
+      "step": 2783
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.334145153433999e-07,
+      "loss": 0.0043,
+      "step": 2784
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3339016074037995e-07,
+      "loss": 0.0069,
+      "step": 2785
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3336580613735995e-07,
+      "loss": 0.0045,
+      "step": 2786
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3334145153434e-07,
+      "loss": 0.005,
+      "step": 2787
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3331709693132e-07,
+      "loss": 0.0033,
+      "step": 2788
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3329274232830005e-07,
+      "loss": 0.0049,
+      "step": 2789
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3326838772528005e-07,
+      "loss": 0.0036,
+      "step": 2790
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.332440331222601e-07,
+      "loss": 0.0043,
+      "step": 2791
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3321967851924015e-07,
+      "loss": 0.0039,
+      "step": 2792
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 4.3319532391622015e-07,
+      "loss": 0.0047,
+      "step": 2793
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.331709693132002e-07,
+      "loss": 0.005,
+      "step": 2794
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.331466147101802e-07,
+      "loss": 0.0042,
+      "step": 2795
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3312226010716025e-07,
+      "loss": 0.0049,
+      "step": 2796
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3309790550414025e-07,
+      "loss": 0.0049,
+      "step": 2797
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.330735509011203e-07,
+      "loss": 0.0037,
+      "step": 2798
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3304919629810035e-07,
+      "loss": 0.0043,
+      "step": 2799
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3302484169508035e-07,
+      "loss": 0.0055,
+      "step": 2800
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.330004870920604e-07,
+      "loss": 0.0045,
+      "step": 2801
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.329761324890404e-07,
+      "loss": 0.0049,
+      "step": 2802
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3295177788602045e-07,
+      "loss": 0.0045,
+      "step": 2803
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 4.3292742328300045e-07,
+      "loss": 0.0051,
+      "step": 2804
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.329030686799805e-07,
+      "loss": 0.0056,
+      "step": 2805
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.328787140769605e-07,
+      "loss": 0.0039,
+      "step": 2806
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.3285435947394055e-07,
+      "loss": 0.0033,
+      "step": 2807
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.328300048709206e-07,
+      "loss": 0.0037,
+      "step": 2808
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.328056502679006e-07,
+      "loss": 0.0139,
+      "step": 2809
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.3278129566488066e-07,
+      "loss": 0.004,
+      "step": 2810
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.3275694106186066e-07,
+      "loss": 0.0037,
+      "step": 2811
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.327325864588407e-07,
+      "loss": 0.0051,
+      "step": 2812
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.327082318558207e-07,
+      "loss": 1.5049,
+      "step": 2813
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 4.3268387725280076e-07,
+      "loss": 0.0037,
+      "step": 2814
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.326595226497808e-07,
+      "loss": 0.0032,
+      "step": 2815
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.326351680467608e-07,
+      "loss": 0.004,
+      "step": 2816
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.3261081344374086e-07,
+      "loss": 0.004,
+      "step": 2817
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.3258645884072086e-07,
+      "loss": 0.0047,
+      "step": 2818
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.325621042377009e-07,
+      "loss": 0.0046,
+      "step": 2819
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.325377496346809e-07,
+      "loss": 0.0047,
+      "step": 2820
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.3251339503166096e-07,
+      "loss": 0.0047,
+      "step": 2821
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.3248904042864096e-07,
+      "loss": 1.1417,
+      "step": 2822
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.32464685825621e-07,
+      "loss": 0.0079,
+      "step": 2823
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 4.3244033122260106e-07,
+      "loss": 0.0032,
+      "step": 2824
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.3241597661958106e-07,
+      "loss": 0.0045,
+      "step": 2825
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.323916220165611e-07,
+      "loss": 0.0048,
+      "step": 2826
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.323672674135411e-07,
+      "loss": 0.0039,
+      "step": 2827
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.3234291281052116e-07,
+      "loss": 0.0062,
+      "step": 2828
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.3231855820750116e-07,
+      "loss": 0.0047,
+      "step": 2829
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.322942036044812e-07,
+      "loss": 0.0044,
+      "step": 2830
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.3226984900146126e-07,
+      "loss": 0.0039,
+      "step": 2831
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.3224549439844126e-07,
+      "loss": 0.0044,
+      "step": 2832
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.322211397954213e-07,
+      "loss": 0.0042,
+      "step": 2833
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 4.321967851924013e-07,
+      "loss": 0.0039,
+      "step": 2834
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.3217243058938137e-07,
+      "loss": 0.0053,
+      "step": 2835
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.3214807598636136e-07,
+      "loss": 0.004,
+      "step": 2836
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.321237213833414e-07,
+      "loss": 0.0041,
+      "step": 2837
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.320993667803214e-07,
+      "loss": 0.0042,
+      "step": 2838
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.3207501217730147e-07,
+      "loss": 0.0033,
+      "step": 2839
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.320506575742815e-07,
+      "loss": 0.0054,
+      "step": 2840
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.320263029712615e-07,
+      "loss": 0.0041,
+      "step": 2841
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.3200194836824157e-07,
+      "loss": 0.0056,
+      "step": 2842
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.3197759376522157e-07,
+      "loss": 0.0051,
+      "step": 2843
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.319532391622016e-07,
+      "loss": 0.0052,
+      "step": 2844
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 4.319288845591816e-07,
+      "loss": 0.0037,
+      "step": 2845
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.3190452995616167e-07,
+      "loss": 0.3419,
+      "step": 2846
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.3188017535314177e-07,
+      "loss": 0.0048,
+      "step": 2847
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.3185582075012177e-07,
+      "loss": 0.0042,
+      "step": 2848
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.318314661471018e-07,
+      "loss": 0.9761,
+      "step": 2849
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.318071115440818e-07,
+      "loss": 0.0037,
+      "step": 2850
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.317827569410619e-07,
+      "loss": 0.0051,
+      "step": 2851
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.3175840233804187e-07,
+      "loss": 0.0031,
+      "step": 2852
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.317340477350219e-07,
+      "loss": 0.0194,
+      "step": 2853
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.31709693132002e-07,
+      "loss": 0.0032,
+      "step": 2854
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 4.31685338528982e-07,
+      "loss": 0.0043,
+      "step": 2855
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.31660983925962e-07,
+      "loss": 0.0041,
+      "step": 2856
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.31636629322942e-07,
+      "loss": 0.0425,
+      "step": 2857
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.316122747199221e-07,
+      "loss": 0.0036,
+      "step": 2858
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.315879201169021e-07,
+      "loss": 0.0047,
+      "step": 2859
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.3156356551388213e-07,
+      "loss": 0.0041,
+      "step": 2860
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.315392109108621e-07,
+      "loss": 0.0056,
+      "step": 2861
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.315148563078422e-07,
+      "loss": 0.0042,
+      "step": 2862
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.3149050170482223e-07,
+      "loss": 0.0043,
+      "step": 2863
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.3146614710180223e-07,
+      "loss": 0.0037,
+      "step": 2864
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 4.314417924987823e-07,
+      "loss": 0.004,
+      "step": 2865
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.314174378957623e-07,
+      "loss": 0.0037,
+      "step": 2866
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.3139308329274233e-07,
+      "loss": 0.0038,
+      "step": 2867
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.3136872868972233e-07,
+      "loss": 0.0035,
+      "step": 2868
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.313443740867024e-07,
+      "loss": 0.0066,
+      "step": 2869
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.3132001948368243e-07,
+      "loss": 0.0042,
+      "step": 2870
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.3129566488066243e-07,
+      "loss": 0.0042,
+      "step": 2871
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.312713102776425e-07,
+      "loss": 0.0058,
+      "step": 2872
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.312469556746225e-07,
+      "loss": 0.004,
+      "step": 2873
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.3122260107160253e-07,
+      "loss": 0.0043,
+      "step": 2874
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.3119824646858253e-07,
+      "loss": 0.0036,
+      "step": 2875
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 4.311738918655626e-07,
+      "loss": 0.0039,
+      "step": 2876
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.311495372625426e-07,
+      "loss": 0.0049,
+      "step": 2877
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.3112518265952263e-07,
+      "loss": 0.0037,
+      "step": 2878
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.311008280565027e-07,
+      "loss": 0.0046,
+      "step": 2879
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.310764734534827e-07,
+      "loss": 0.0036,
+      "step": 2880
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.3105211885046273e-07,
+      "loss": 0.0039,
+      "step": 2881
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.3102776424744273e-07,
+      "loss": 0.0039,
+      "step": 2882
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.310034096444228e-07,
+      "loss": 0.0042,
+      "step": 2883
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.309790550414028e-07,
+      "loss": 0.0059,
+      "step": 2884
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.3095470043838284e-07,
+      "loss": 0.004,
+      "step": 2885
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 4.309303458353629e-07,
+      "loss": 0.0039,
+      "step": 2886
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.309059912323429e-07,
+      "loss": 0.0044,
+      "step": 2887
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.3088163662932294e-07,
+      "loss": 0.0042,
+      "step": 2888
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.3085728202630294e-07,
+      "loss": 0.0037,
+      "step": 2889
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.30832927423283e-07,
+      "loss": 0.0047,
+      "step": 2890
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.30808572820263e-07,
+      "loss": 0.0041,
+      "step": 2891
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.3078421821724304e-07,
+      "loss": 0.0038,
+      "step": 2892
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.3075986361422304e-07,
+      "loss": 0.0034,
+      "step": 2893
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.307355090112031e-07,
+      "loss": 0.0033,
+      "step": 2894
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.3071115440818314e-07,
+      "loss": 0.2704,
+      "step": 2895
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 4.3068679980516314e-07,
+      "loss": 0.0036,
+      "step": 2896
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.306624452021432e-07,
+      "loss": 0.0045,
+      "step": 2897
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.306380905991232e-07,
+      "loss": 0.0041,
+      "step": 2898
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.3061373599610324e-07,
+      "loss": 0.0062,
+      "step": 2899
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.3058938139308324e-07,
+      "loss": 0.0036,
+      "step": 2900
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.305650267900633e-07,
+      "loss": 0.6922,
+      "step": 2901
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.3054067218704334e-07,
+      "loss": 0.0042,
+      "step": 2902
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.3051631758402334e-07,
+      "loss": 0.0058,
+      "step": 2903
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.304919629810034e-07,
+      "loss": 0.0038,
+      "step": 2904
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.304676083779834e-07,
+      "loss": 0.0045,
+      "step": 2905
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 4.3044325377496344e-07,
+      "loss": 0.004,
+      "step": 2906
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.3041889917194344e-07,
+      "loss": 0.0044,
+      "step": 2907
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.303945445689235e-07,
+      "loss": 0.0052,
+      "step": 2908
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.303701899659035e-07,
+      "loss": 0.0044,
+      "step": 2909
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.3034583536288354e-07,
+      "loss": 0.0065,
+      "step": 2910
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.303214807598636e-07,
+      "loss": 0.0041,
+      "step": 2911
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.302971261568436e-07,
+      "loss": 0.0034,
+      "step": 2912
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.3027277155382365e-07,
+      "loss": 0.768,
+      "step": 2913
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.3024841695080364e-07,
+      "loss": 0.0728,
+      "step": 2914
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.302240623477837e-07,
+      "loss": 0.0034,
+      "step": 2915
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.301997077447637e-07,
+      "loss": 0.007,
+      "step": 2916
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 4.3017535314174375e-07,
+      "loss": 0.616,
+      "step": 2917
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.301509985387238e-07,
+      "loss": 0.0041,
+      "step": 2918
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.301266439357038e-07,
+      "loss": 0.0041,
+      "step": 2919
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.3010228933268385e-07,
+      "loss": 0.0032,
+      "step": 2920
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.3007793472966385e-07,
+      "loss": 1.9184,
+      "step": 2921
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.300535801266439e-07,
+      "loss": 0.0031,
+      "step": 2922
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.3002922552362395e-07,
+      "loss": 0.0044,
+      "step": 2923
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.30004870920604e-07,
+      "loss": 0.0033,
+      "step": 2924
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.2998051631758405e-07,
+      "loss": 0.0069,
+      "step": 2925
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.2995616171456405e-07,
+      "loss": 0.0044,
+      "step": 2926
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 4.299318071115441e-07,
+      "loss": 0.0061,
+      "step": 2927
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.299074525085241e-07,
+      "loss": 0.0029,
+      "step": 2928
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.2988309790550415e-07,
+      "loss": 0.0052,
+      "step": 2929
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.2985874330248415e-07,
+      "loss": 0.004,
+      "step": 2930
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.298343886994642e-07,
+      "loss": 1.2554,
+      "step": 2931
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.298100340964442e-07,
+      "loss": 0.0035,
+      "step": 2932
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.2978567949342425e-07,
+      "loss": 0.0039,
+      "step": 2933
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.297613248904043e-07,
+      "loss": 0.0041,
+      "step": 2934
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.297369702873843e-07,
+      "loss": 0.0036,
+      "step": 2935
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.2971261568436436e-07,
+      "loss": 0.0032,
+      "step": 2936
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 4.2968826108134436e-07,
+      "loss": 0.0102,
+      "step": 2937
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.296639064783244e-07,
+      "loss": 0.005,
+      "step": 2938
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.296395518753044e-07,
+      "loss": 0.0038,
+      "step": 2939
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.2961519727228446e-07,
+      "loss": 0.0055,
+      "step": 2940
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.295908426692645e-07,
+      "loss": 0.0044,
+      "step": 2941
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.295664880662445e-07,
+      "loss": 0.0039,
+      "step": 2942
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.2954213346322456e-07,
+      "loss": 0.0046,
+      "step": 2943
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.2951777886020456e-07,
+      "loss": 0.0045,
+      "step": 2944
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.294934242571846e-07,
+      "loss": 0.0035,
+      "step": 2945
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.294690696541646e-07,
+      "loss": 0.0036,
+      "step": 2946
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.2944471505114466e-07,
+      "loss": 0.0834,
+      "step": 2947
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 4.294203604481247e-07,
+      "loss": 0.0042,
+      "step": 2948
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.293960058451047e-07,
+      "loss": 0.1026,
+      "step": 2949
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.2937165124208476e-07,
+      "loss": 0.0036,
+      "step": 2950
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.2934729663906476e-07,
+      "loss": 0.0045,
+      "step": 2951
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.293229420360448e-07,
+      "loss": 0.0051,
+      "step": 2952
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.292985874330248e-07,
+      "loss": 0.0057,
+      "step": 2953
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.2927423283000486e-07,
+      "loss": 0.0044,
+      "step": 2954
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.2924987822698486e-07,
+      "loss": 0.0144,
+      "step": 2955
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.292255236239649e-07,
+      "loss": 0.0041,
+      "step": 2956
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.2920116902094496e-07,
+      "loss": 0.0038,
+      "step": 2957
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 4.2917681441792496e-07,
+      "loss": 0.0033,
+      "step": 2958
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.29152459814905e-07,
+      "loss": 0.0193,
+      "step": 2959
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.29128105211885e-07,
+      "loss": 0.0038,
+      "step": 2960
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.2910375060886507e-07,
+      "loss": 0.0035,
+      "step": 2961
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.2907939600584506e-07,
+      "loss": 0.0041,
+      "step": 2962
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.290550414028251e-07,
+      "loss": 0.0037,
+      "step": 2963
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.2903068679980517e-07,
+      "loss": 0.0042,
+      "step": 2964
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.2900633219678517e-07,
+      "loss": 0.0049,
+      "step": 2965
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.289819775937652e-07,
+      "loss": 0.0037,
+      "step": 2966
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.289576229907452e-07,
+      "loss": 0.0043,
+      "step": 2967
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 4.2893326838772527e-07,
+      "loss": 0.0043,
+      "step": 2968
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.2890891378470527e-07,
+      "loss": 0.0035,
+      "step": 2969
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.288845591816853e-07,
+      "loss": 0.0035,
+      "step": 2970
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.288602045786653e-07,
+      "loss": 0.0042,
+      "step": 2971
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.2883584997564537e-07,
+      "loss": 0.0048,
+      "step": 2972
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.288114953726254e-07,
+      "loss": 0.0043,
+      "step": 2973
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.287871407696054e-07,
+      "loss": 0.0039,
+      "step": 2974
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.2876278616658547e-07,
+      "loss": 0.0029,
+      "step": 2975
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.2873843156356547e-07,
+      "loss": 0.0028,
+      "step": 2976
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.287140769605455e-07,
+      "loss": 0.0042,
+      "step": 2977
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 4.286897223575255e-07,
+      "loss": 0.0036,
+      "step": 2978
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.2866536775450557e-07,
+      "loss": 0.0033,
+      "step": 2979
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.286410131514856e-07,
+      "loss": 0.0037,
+      "step": 2980
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.286166585484656e-07,
+      "loss": 0.0031,
+      "step": 2981
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.2859230394544567e-07,
+      "loss": 0.0033,
+      "step": 2982
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.2856794934242567e-07,
+      "loss": 0.0034,
+      "step": 2983
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.285435947394057e-07,
+      "loss": 0.0029,
+      "step": 2984
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.285192401363857e-07,
+      "loss": 0.0034,
+      "step": 2985
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.284948855333658e-07,
+      "loss": 0.004,
+      "step": 2986
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.2847053093034577e-07,
+      "loss": 0.0029,
+      "step": 2987
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.284461763273258e-07,
+      "loss": 0.0041,
+      "step": 2988
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 4.284218217243059e-07,
+      "loss": 0.0029,
+      "step": 2989
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.283974671212859e-07,
+      "loss": 0.0038,
+      "step": 2990
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.283731125182659e-07,
+      "loss": 0.004,
+      "step": 2991
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.283487579152459e-07,
+      "loss": 0.0046,
+      "step": 2992
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.28324403312226e-07,
+      "loss": 0.0038,
+      "step": 2993
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.28300048709206e-07,
+      "loss": 0.0045,
+      "step": 2994
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2827569410618603e-07,
+      "loss": 0.005,
+      "step": 2995
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.282513395031661e-07,
+      "loss": 0.0037,
+      "step": 2996
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.282269849001461e-07,
+      "loss": 0.0038,
+      "step": 2997
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2820263029712613e-07,
+      "loss": 0.0038,
+      "step": 2998
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 4.2817827569410613e-07,
+      "loss": 0.0044,
+      "step": 2999
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.2815392109108623e-07,
+      "loss": 0.0047,
+      "step": 3000
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.2812956648806623e-07,
+      "loss": 0.0046,
+      "step": 3001
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.281052118850463e-07,
+      "loss": 0.018,
+      "step": 3002
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.2808085728202633e-07,
+      "loss": 0.0042,
+      "step": 3003
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.2805650267900633e-07,
+      "loss": 0.0043,
+      "step": 3004
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.280321480759864e-07,
+      "loss": 0.0035,
+      "step": 3005
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.280077934729664e-07,
+      "loss": 0.0043,
+      "step": 3006
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.2798343886994643e-07,
+      "loss": 0.0036,
+      "step": 3007
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.2795908426692643e-07,
+      "loss": 0.004,
+      "step": 3008
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 4.279347296639065e-07,
+      "loss": 0.003,
+      "step": 3009
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.279103750608865e-07,
+      "loss": 0.0038,
+      "step": 3010
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.2788602045786654e-07,
+      "loss": 0.0034,
+      "step": 3011
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.278616658548466e-07,
+      "loss": 0.0056,
+      "step": 3012
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.278373112518266e-07,
+      "loss": 0.004,
+      "step": 3013
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.2781295664880664e-07,
+      "loss": 0.0401,
+      "step": 3014
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.2778860204578664e-07,
+      "loss": 0.0043,
+      "step": 3015
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.277642474427667e-07,
+      "loss": 0.0031,
+      "step": 3016
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.277398928397467e-07,
+      "loss": 0.0034,
+      "step": 3017
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.2771553823672674e-07,
+      "loss": 0.003,
+      "step": 3018
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.276911836337068e-07,
+      "loss": 0.0031,
+      "step": 3019
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.276668290306868e-07,
+      "loss": 0.0036,
+      "step": 3020
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.2764247442766684e-07,
+      "loss": 0.0054,
+      "step": 3021
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.2761811982464684e-07,
+      "loss": 0.0034,
+      "step": 3022
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.275937652216269e-07,
+      "loss": 0.0042,
+      "step": 3023
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.275694106186069e-07,
+      "loss": 0.0044,
+      "step": 3024
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.2754505601558694e-07,
+      "loss": 0.0045,
+      "step": 3025
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.2752070141256694e-07,
+      "loss": 0.0057,
+      "step": 3026
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.27496346809547e-07,
+      "loss": 0.0038,
+      "step": 3027
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.2747199220652704e-07,
+      "loss": 0.0043,
+      "step": 3028
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.2744763760350704e-07,
+      "loss": 0.0034,
+      "step": 3029
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 4.274232830004871e-07,
+      "loss": 0.0045,
+      "step": 3030
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.273989283974671e-07,
+      "loss": 0.0071,
+      "step": 3031
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.2737457379444714e-07,
+      "loss": 0.0095,
+      "step": 3032
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.2735021919142714e-07,
+      "loss": 0.0047,
+      "step": 3033
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.273258645884072e-07,
+      "loss": 0.0034,
+      "step": 3034
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.2730150998538725e-07,
+      "loss": 0.0036,
+      "step": 3035
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.2727715538236724e-07,
+      "loss": 0.005,
+      "step": 3036
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.272528007793473e-07,
+      "loss": 0.0036,
+      "step": 3037
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.272284461763273e-07,
+      "loss": 0.0033,
+      "step": 3038
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.2720409157330735e-07,
+      "loss": 0.0028,
+      "step": 3039
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 4.2717973697028734e-07,
+      "loss": 0.003,
+      "step": 3040
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.271553823672674e-07,
+      "loss": 0.004,
+      "step": 3041
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.271310277642474e-07,
+      "loss": 0.005,
+      "step": 3042
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.2710667316122745e-07,
+      "loss": 0.0042,
+      "step": 3043
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.270823185582075e-07,
+      "loss": 0.0317,
+      "step": 3044
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.270579639551875e-07,
+      "loss": 0.0045,
+      "step": 3045
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.2703360935216755e-07,
+      "loss": 0.0034,
+      "step": 3046
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.2700925474914755e-07,
+      "loss": 0.0036,
+      "step": 3047
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.269849001461276e-07,
+      "loss": 0.0047,
+      "step": 3048
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.269605455431076e-07,
+      "loss": 0.0033,
+      "step": 3049
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 4.2693619094008765e-07,
+      "loss": 0.0036,
+      "step": 3050
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.269118363370677e-07,
+      "loss": 0.0034,
+      "step": 3051
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.268874817340477e-07,
+      "loss": 0.1376,
+      "step": 3052
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.2686312713102775e-07,
+      "loss": 0.004,
+      "step": 3053
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.2683877252800775e-07,
+      "loss": 0.0032,
+      "step": 3054
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.268144179249878e-07,
+      "loss": 0.0043,
+      "step": 3055
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.267900633219678e-07,
+      "loss": 0.0039,
+      "step": 3056
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.2676570871894785e-07,
+      "loss": 0.0038,
+      "step": 3057
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.2674135411592785e-07,
+      "loss": 0.0029,
+      "step": 3058
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.267169995129079e-07,
+      "loss": 0.0034,
+      "step": 3059
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.2669264490988795e-07,
+      "loss": 0.0032,
+      "step": 3060
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 4.2666829030686795e-07,
+      "loss": 0.0034,
+      "step": 3061
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.26643935703848e-07,
+      "loss": 0.0036,
+      "step": 3062
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.26619581100828e-07,
+      "loss": 0.0039,
+      "step": 3063
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.2659522649780805e-07,
+      "loss": 0.0041,
+      "step": 3064
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.2657087189478805e-07,
+      "loss": 0.0031,
+      "step": 3065
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.265465172917681e-07,
+      "loss": 0.0043,
+      "step": 3066
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.2652216268874816e-07,
+      "loss": 0.0045,
+      "step": 3067
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.2649780808572816e-07,
+      "loss": 0.0034,
+      "step": 3068
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.264734534827082e-07,
+      "loss": 0.003,
+      "step": 3069
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.264490988796882e-07,
+      "loss": 0.0075,
+      "step": 3070
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 4.2642474427666826e-07,
+      "loss": 0.004,
+      "step": 3071
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2640038967364826e-07,
+      "loss": 0.0039,
+      "step": 3072
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.263760350706283e-07,
+      "loss": 0.0035,
+      "step": 3073
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.263516804676083e-07,
+      "loss": 0.0039,
+      "step": 3074
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2632732586458836e-07,
+      "loss": 0.0032,
+      "step": 3075
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2630297126156846e-07,
+      "loss": 0.0043,
+      "step": 3076
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2627861665854846e-07,
+      "loss": 0.0035,
+      "step": 3077
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.262542620555285e-07,
+      "loss": 0.003,
+      "step": 3078
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.262299074525085e-07,
+      "loss": 0.006,
+      "step": 3079
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2620555284948856e-07,
+      "loss": 0.0034,
+      "step": 3080
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 4.2618119824646856e-07,
+      "loss": 0.0035,
+      "step": 3081
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.261568436434486e-07,
+      "loss": 0.0037,
+      "step": 3082
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.2613248904042866e-07,
+      "loss": 0.0037,
+      "step": 3083
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.2610813443740866e-07,
+      "loss": 0.0033,
+      "step": 3084
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.260837798343887e-07,
+      "loss": 0.0039,
+      "step": 3085
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.260594252313687e-07,
+      "loss": 0.0033,
+      "step": 3086
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.2603507062834877e-07,
+      "loss": 0.0033,
+      "step": 3087
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9965969859017987,
+      "eval_loss": 0.026054147630929947,
+      "eval_runtime": 2.7911,
+      "eval_samples_per_second": 736.992,
+      "eval_steps_per_second": 11.823,
+      "step": 3087
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.2601071602532876e-07,
+      "loss": 0.0029,
+      "step": 3088
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.259863614223088e-07,
+      "loss": 0.0039,
+      "step": 3089
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.2596200681928887e-07,
+      "loss": 0.0032,
+      "step": 3090
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.2593765221626887e-07,
+      "loss": 0.0038,
+      "step": 3091
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 4.259132976132489e-07,
+      "loss": 0.169,
+      "step": 3092
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.258889430102289e-07,
+      "loss": 0.0038,
+      "step": 3093
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.2586458840720897e-07,
+      "loss": 0.0034,
+      "step": 3094
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.2584023380418897e-07,
+      "loss": 0.0031,
+      "step": 3095
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.25815879201169e-07,
+      "loss": 0.0032,
+      "step": 3096
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.25791524598149e-07,
+      "loss": 0.0035,
+      "step": 3097
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.2576716999512907e-07,
+      "loss": 0.0034,
+      "step": 3098
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.257428153921091e-07,
+      "loss": 0.0039,
+      "step": 3099
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.257184607890891e-07,
+      "loss": 0.0054,
+      "step": 3100
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.2569410618606917e-07,
+      "loss": 0.0033,
+      "step": 3101
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 4.2566975158304917e-07,
+      "loss": 0.0032,
+      "step": 3102
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.256453969800292e-07,
+      "loss": 0.0032,
+      "step": 3103
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.256210423770092e-07,
+      "loss": 0.004,
+      "step": 3104
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.2559668777398927e-07,
+      "loss": 0.0038,
+      "step": 3105
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.255723331709693e-07,
+      "loss": 0.0044,
+      "step": 3106
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.255479785679493e-07,
+      "loss": 0.0036,
+      "step": 3107
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.255236239649294e-07,
+      "loss": 0.0047,
+      "step": 3108
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.2549926936190937e-07,
+      "loss": 0.0037,
+      "step": 3109
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.254749147588894e-07,
+      "loss": 0.003,
+      "step": 3110
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.254505601558694e-07,
+      "loss": 0.0038,
+      "step": 3111
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 4.254262055528495e-07,
+      "loss": 0.0032,
+      "step": 3112
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.254018509498295e-07,
+      "loss": 0.0031,
+      "step": 3113
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.253774963468095e-07,
+      "loss": 0.0039,
+      "step": 3114
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.253531417437896e-07,
+      "loss": 0.0044,
+      "step": 3115
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.253287871407696e-07,
+      "loss": 0.9916,
+      "step": 3116
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.2530443253774963e-07,
+      "loss": 0.0033,
+      "step": 3117
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.252800779347296e-07,
+      "loss": 0.0036,
+      "step": 3118
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.252557233317097e-07,
+      "loss": 0.0031,
+      "step": 3119
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.252313687286897e-07,
+      "loss": 0.0038,
+      "step": 3120
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.2520701412566973e-07,
+      "loss": 0.0036,
+      "step": 3121
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.251826595226498e-07,
+      "loss": 0.0027,
+      "step": 3122
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 4.251583049196298e-07,
+      "loss": 0.0031,
+      "step": 3123
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.2513395031660983e-07,
+      "loss": 0.0032,
+      "step": 3124
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.2510959571358983e-07,
+      "loss": 0.0072,
+      "step": 3125
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.250852411105699e-07,
+      "loss": 0.0037,
+      "step": 3126
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.250608865075499e-07,
+      "loss": 0.0037,
+      "step": 3127
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.2503653190452993e-07,
+      "loss": 0.0029,
+      "step": 3128
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.2501217730151e-07,
+      "loss": 0.0035,
+      "step": 3129
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.2498782269849e-07,
+      "loss": 0.0035,
+      "step": 3130
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.2496346809547003e-07,
+      "loss": 0.0034,
+      "step": 3131
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.2493911349245003e-07,
+      "loss": 0.0038,
+      "step": 3132
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 4.249147588894301e-07,
+      "loss": 0.003,
+      "step": 3133
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.248904042864101e-07,
+      "loss": 0.5601,
+      "step": 3134
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.2486604968339013e-07,
+      "loss": 0.0029,
+      "step": 3135
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.2484169508037013e-07,
+      "loss": 0.0037,
+      "step": 3136
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.248173404773502e-07,
+      "loss": 0.0034,
+      "step": 3137
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.2479298587433023e-07,
+      "loss": 0.003,
+      "step": 3138
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.2476863127131023e-07,
+      "loss": 0.0039,
+      "step": 3139
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.247442766682903e-07,
+      "loss": 0.0036,
+      "step": 3140
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.247199220652703e-07,
+      "loss": 0.0037,
+      "step": 3141
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.2469556746225034e-07,
+      "loss": 0.0044,
+      "step": 3142
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 4.2467121285923033e-07,
+      "loss": 0.0039,
+      "step": 3143
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.246468582562104e-07,
+      "loss": 0.0037,
+      "step": 3144
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.2462250365319044e-07,
+      "loss": 0.0036,
+      "step": 3145
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.2459814905017044e-07,
+      "loss": 0.0029,
+      "step": 3146
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.245737944471505e-07,
+      "loss": 0.0032,
+      "step": 3147
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.245494398441305e-07,
+      "loss": 0.0039,
+      "step": 3148
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.2452508524111054e-07,
+      "loss": 0.0033,
+      "step": 3149
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.2450073063809054e-07,
+      "loss": 0.0038,
+      "step": 3150
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.244763760350706e-07,
+      "loss": 0.0028,
+      "step": 3151
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.244520214320506e-07,
+      "loss": 0.0031,
+      "step": 3152
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 4.244276668290307e-07,
+      "loss": 0.0037,
+      "step": 3153
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.2440331222601074e-07,
+      "loss": 0.0032,
+      "step": 3154
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.2437895762299074e-07,
+      "loss": 0.0037,
+      "step": 3155
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.243546030199708e-07,
+      "loss": 0.0039,
+      "step": 3156
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.243302484169508e-07,
+      "loss": 0.0033,
+      "step": 3157
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.2430589381393084e-07,
+      "loss": 0.0028,
+      "step": 3158
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.2428153921091084e-07,
+      "loss": 0.1282,
+      "step": 3159
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.242571846078909e-07,
+      "loss": 0.0029,
+      "step": 3160
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.2423283000487095e-07,
+      "loss": 0.0031,
+      "step": 3161
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.2420847540185094e-07,
+      "loss": 0.0039,
+      "step": 3162
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.24184120798831e-07,
+      "loss": 0.0029,
+      "step": 3163
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 4.24159766195811e-07,
+      "loss": 0.003,
+      "step": 3164
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.2413541159279105e-07,
+      "loss": 0.0039,
+      "step": 3165
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.2411105698977104e-07,
+      "loss": 0.0031,
+      "step": 3166
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.240867023867511e-07,
+      "loss": 0.0041,
+      "step": 3167
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.2406234778373115e-07,
+      "loss": 0.0038,
+      "step": 3168
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.2403799318071115e-07,
+      "loss": 0.0033,
+      "step": 3169
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.240136385776912e-07,
+      "loss": 0.0032,
+      "step": 3170
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.239892839746712e-07,
+      "loss": 0.0041,
+      "step": 3171
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.2396492937165125e-07,
+      "loss": 0.0032,
+      "step": 3172
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.2394057476863125e-07,
+      "loss": 0.004,
+      "step": 3173
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 4.239162201656113e-07,
+      "loss": 0.0029,
+      "step": 3174
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.238918655625913e-07,
+      "loss": 0.0036,
+      "step": 3175
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.2386751095957135e-07,
+      "loss": 0.0045,
+      "step": 3176
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.238431563565514e-07,
+      "loss": 0.0033,
+      "step": 3177
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.238188017535314e-07,
+      "loss": 0.0038,
+      "step": 3178
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.2379444715051145e-07,
+      "loss": 0.0031,
+      "step": 3179
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.2377009254749145e-07,
+      "loss": 0.0032,
+      "step": 3180
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.237457379444715e-07,
+      "loss": 0.0028,
+      "step": 3181
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.237213833414515e-07,
+      "loss": 0.0282,
+      "step": 3182
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.2369702873843155e-07,
+      "loss": 0.0031,
+      "step": 3183
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 4.236726741354116e-07,
+      "loss": 0.0031,
+      "step": 3184
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.236483195323916e-07,
+      "loss": 0.0035,
+      "step": 3185
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.2362396492937165e-07,
+      "loss": 0.0029,
+      "step": 3186
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.2359961032635165e-07,
+      "loss": 0.0029,
+      "step": 3187
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.235752557233317e-07,
+      "loss": 0.0031,
+      "step": 3188
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.235509011203117e-07,
+      "loss": 0.0037,
+      "step": 3189
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.2352654651729175e-07,
+      "loss": 0.0036,
+      "step": 3190
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.2350219191427175e-07,
+      "loss": 0.152,
+      "step": 3191
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.234778373112518e-07,
+      "loss": 0.0026,
+      "step": 3192
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.2345348270823186e-07,
+      "loss": 0.0035,
+      "step": 3193
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.2342912810521186e-07,
+      "loss": 0.0046,
+      "step": 3194
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 4.234047735021919e-07,
+      "loss": 0.0036,
+      "step": 3195
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.233804188991719e-07,
+      "loss": 0.004,
+      "step": 3196
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.2335606429615196e-07,
+      "loss": 0.0037,
+      "step": 3197
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.2333170969313196e-07,
+      "loss": 0.0027,
+      "step": 3198
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.23307355090112e-07,
+      "loss": 0.0037,
+      "step": 3199
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.2328300048709206e-07,
+      "loss": 0.0036,
+      "step": 3200
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.2325864588407206e-07,
+      "loss": 0.0065,
+      "step": 3201
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.232342912810521e-07,
+      "loss": 0.0031,
+      "step": 3202
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.232099366780321e-07,
+      "loss": 0.0034,
+      "step": 3203
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.2318558207501216e-07,
+      "loss": 0.0027,
+      "step": 3204
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 4.2316122747199216e-07,
+      "loss": 0.0342,
+      "step": 3205
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.231368728689722e-07,
+      "loss": 0.0033,
+      "step": 3206
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.231125182659522e-07,
+      "loss": 0.0032,
+      "step": 3207
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.2308816366293226e-07,
+      "loss": 0.004,
+      "step": 3208
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.230638090599123e-07,
+      "loss": 0.0033,
+      "step": 3209
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.230394544568923e-07,
+      "loss": 0.0032,
+      "step": 3210
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.2301509985387236e-07,
+      "loss": 0.0087,
+      "step": 3211
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.2299074525085236e-07,
+      "loss": 0.0032,
+      "step": 3212
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.229663906478324e-07,
+      "loss": 0.0045,
+      "step": 3213
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.229420360448124e-07,
+      "loss": 0.0031,
+      "step": 3214
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 4.2291768144179246e-07,
+      "loss": 0.0037,
+      "step": 3215
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.228933268387725e-07,
+      "loss": 0.0034,
+      "step": 3216
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.228689722357525e-07,
+      "loss": 0.0025,
+      "step": 3217
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.2284461763273257e-07,
+      "loss": 0.0031,
+      "step": 3218
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.2282026302971256e-07,
+      "loss": 0.0034,
+      "step": 3219
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.227959084266926e-07,
+      "loss": 0.0028,
+      "step": 3220
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.227715538236726e-07,
+      "loss": 0.5526,
+      "step": 3221
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.2274719922065267e-07,
+      "loss": 0.0057,
+      "step": 3222
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.2272284461763266e-07,
+      "loss": 0.0033,
+      "step": 3223
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.226984900146127e-07,
+      "loss": 0.0032,
+      "step": 3224
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 4.2267413541159277e-07,
+      "loss": 0.0031,
+      "step": 3225
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.2264978080857277e-07,
+      "loss": 0.0038,
+      "step": 3226
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.226254262055528e-07,
+      "loss": 1.3561,
+      "step": 3227
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.226010716025328e-07,
+      "loss": 0.0031,
+      "step": 3228
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.2257671699951287e-07,
+      "loss": 0.0038,
+      "step": 3229
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.225523623964929e-07,
+      "loss": 0.003,
+      "step": 3230
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.2252800779347297e-07,
+      "loss": 0.0041,
+      "step": 3231
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.22503653190453e-07,
+      "loss": 0.0043,
+      "step": 3232
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.22479298587433e-07,
+      "loss": 0.0029,
+      "step": 3233
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.224549439844131e-07,
+      "loss": 0.003,
+      "step": 3234
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.2243058938139307e-07,
+      "loss": 0.0039,
+      "step": 3235
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 4.224062347783731e-07,
+      "loss": 0.0031,
+      "step": 3236
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.223818801753531e-07,
+      "loss": 0.0034,
+      "step": 3237
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.223575255723332e-07,
+      "loss": 0.0029,
+      "step": 3238
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.223331709693132e-07,
+      "loss": 0.004,
+      "step": 3239
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.223088163662932e-07,
+      "loss": 0.0034,
+      "step": 3240
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.222844617632733e-07,
+      "loss": 0.0032,
+      "step": 3241
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.222601071602533e-07,
+      "loss": 0.0038,
+      "step": 3242
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.2223575255723333e-07,
+      "loss": 0.0026,
+      "step": 3243
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.222113979542133e-07,
+      "loss": 0.0027,
+      "step": 3244
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.221870433511934e-07,
+      "loss": 0.0034,
+      "step": 3245
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 4.221626887481734e-07,
+      "loss": 0.0028,
+      "step": 3246
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.2213833414515343e-07,
+      "loss": 0.0513,
+      "step": 3247
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.221139795421335e-07,
+      "loss": 0.0031,
+      "step": 3248
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.220896249391135e-07,
+      "loss": 0.0029,
+      "step": 3249
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.2206527033609353e-07,
+      "loss": 0.0031,
+      "step": 3250
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.2204091573307353e-07,
+      "loss": 0.0029,
+      "step": 3251
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.220165611300536e-07,
+      "loss": 0.0027,
+      "step": 3252
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.219922065270336e-07,
+      "loss": 0.0037,
+      "step": 3253
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.2196785192401363e-07,
+      "loss": 0.0025,
+      "step": 3254
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.219434973209937e-07,
+      "loss": 0.0377,
+      "step": 3255
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 4.219191427179737e-07,
+      "loss": 0.0031,
+      "step": 3256
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.2189478811495373e-07,
+      "loss": 0.0026,
+      "step": 3257
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.2187043351193373e-07,
+      "loss": 0.0031,
+      "step": 3258
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.218460789089138e-07,
+      "loss": 0.0028,
+      "step": 3259
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.218217243058938e-07,
+      "loss": 0.0031,
+      "step": 3260
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.2179736970287383e-07,
+      "loss": 0.0029,
+      "step": 3261
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.2177301509985383e-07,
+      "loss": 0.0035,
+      "step": 3262
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.217486604968339e-07,
+      "loss": 0.0032,
+      "step": 3263
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.2172430589381393e-07,
+      "loss": 0.0045,
+      "step": 3264
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.2169995129079393e-07,
+      "loss": 0.0032,
+      "step": 3265
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.21675596687774e-07,
+      "loss": 0.0031,
+      "step": 3266
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 4.21651242084754e-07,
+      "loss": 0.003,
+      "step": 3267
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.2162688748173404e-07,
+      "loss": 0.0034,
+      "step": 3268
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.2160253287871403e-07,
+      "loss": 0.0028,
+      "step": 3269
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.215781782756941e-07,
+      "loss": 0.0036,
+      "step": 3270
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.2155382367267414e-07,
+      "loss": 0.0031,
+      "step": 3271
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.2152946906965414e-07,
+      "loss": 0.0025,
+      "step": 3272
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.215051144666342e-07,
+      "loss": 0.0032,
+      "step": 3273
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.214807598636142e-07,
+      "loss": 0.0602,
+      "step": 3274
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.2145640526059424e-07,
+      "loss": 0.0041,
+      "step": 3275
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.2143205065757424e-07,
+      "loss": 0.0026,
+      "step": 3276
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 4.214076960545543e-07,
+      "loss": 0.0026,
+      "step": 3277
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.213833414515343e-07,
+      "loss": 0.0031,
+      "step": 3278
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.2135898684851434e-07,
+      "loss": 0.0029,
+      "step": 3279
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.213346322454944e-07,
+      "loss": 0.0034,
+      "step": 3280
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.213102776424744e-07,
+      "loss": 0.0036,
+      "step": 3281
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.2128592303945444e-07,
+      "loss": 0.003,
+      "step": 3282
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.2126156843643444e-07,
+      "loss": 0.0029,
+      "step": 3283
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.212372138334145e-07,
+      "loss": 0.0026,
+      "step": 3284
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.212128592303945e-07,
+      "loss": 0.003,
+      "step": 3285
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.2118850462737454e-07,
+      "loss": 0.0027,
+      "step": 3286
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 4.211641500243546e-07,
+      "loss": 0.0041,
+      "step": 3287
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.211397954213346e-07,
+      "loss": 0.0024,
+      "step": 3288
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.2111544081831464e-07,
+      "loss": 0.0034,
+      "step": 3289
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.2109108621529464e-07,
+      "loss": 0.0039,
+      "step": 3290
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.210667316122747e-07,
+      "loss": 0.0026,
+      "step": 3291
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.210423770092547e-07,
+      "loss": 0.0033,
+      "step": 3292
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.2101802240623474e-07,
+      "loss": 0.0023,
+      "step": 3293
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.209936678032148e-07,
+      "loss": 0.0034,
+      "step": 3294
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.209693132001948e-07,
+      "loss": 0.0024,
+      "step": 3295
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.2094495859717485e-07,
+      "loss": 0.0036,
+      "step": 3296
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 4.2092060399415484e-07,
+      "loss": 0.01,
+      "step": 3297
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.208962493911349e-07,
+      "loss": 0.0028,
+      "step": 3298
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.208718947881149e-07,
+      "loss": 0.0029,
+      "step": 3299
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.2084754018509495e-07,
+      "loss": 0.0031,
+      "step": 3300
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.2082318558207495e-07,
+      "loss": 0.0027,
+      "step": 3301
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.20798830979055e-07,
+      "loss": 0.0025,
+      "step": 3302
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.2077447637603505e-07,
+      "loss": 0.0031,
+      "step": 3303
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.2075012177301505e-07,
+      "loss": 0.0041,
+      "step": 3304
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.207257671699951e-07,
+      "loss": 0.0033,
+      "step": 3305
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.2070141256697515e-07,
+      "loss": 0.007,
+      "step": 3306
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.206770579639552e-07,
+      "loss": 0.0036,
+      "step": 3307
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 4.206527033609352e-07,
+      "loss": 0.0029,
+      "step": 3308
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.2062834875791525e-07,
+      "loss": 0.0035,
+      "step": 3309
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.206039941548953e-07,
+      "loss": 0.0026,
+      "step": 3310
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.205796395518753e-07,
+      "loss": 0.003,
+      "step": 3311
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.2055528494885535e-07,
+      "loss": 0.0039,
+      "step": 3312
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.2053093034583535e-07,
+      "loss": 0.0032,
+      "step": 3313
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.205065757428154e-07,
+      "loss": 0.0034,
+      "step": 3314
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.204822211397954e-07,
+      "loss": 0.0025,
+      "step": 3315
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.2045786653677545e-07,
+      "loss": 0.0038,
+      "step": 3316
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.204335119337555e-07,
+      "loss": 0.0035,
+      "step": 3317
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 4.204091573307355e-07,
+      "loss": 0.0028,
+      "step": 3318
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.2038480272771556e-07,
+      "loss": 0.0033,
+      "step": 3319
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.2036044812469556e-07,
+      "loss": 0.0035,
+      "step": 3320
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.203360935216756e-07,
+      "loss": 0.0028,
+      "step": 3321
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.203117389186556e-07,
+      "loss": 0.0026,
+      "step": 3322
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.2028738431563566e-07,
+      "loss": 0.003,
+      "step": 3323
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.2026302971261566e-07,
+      "loss": 0.0029,
+      "step": 3324
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.202386751095957e-07,
+      "loss": 0.003,
+      "step": 3325
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.2021432050657576e-07,
+      "loss": 0.0027,
+      "step": 3326
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.2018996590355576e-07,
+      "loss": 0.0031,
+      "step": 3327
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 4.201656113005358e-07,
+      "loss": 0.0023,
+      "step": 3328
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.201412566975158e-07,
+      "loss": 0.0031,
+      "step": 3329
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.2011690209449586e-07,
+      "loss": 0.0029,
+      "step": 3330
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.2009254749147586e-07,
+      "loss": 0.0314,
+      "step": 3331
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.200681928884559e-07,
+      "loss": 0.0209,
+      "step": 3332
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.2004383828543596e-07,
+      "loss": 0.003,
+      "step": 3333
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.2001948368241596e-07,
+      "loss": 0.0031,
+      "step": 3334
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.19995129079396e-07,
+      "loss": 0.0036,
+      "step": 3335
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.19970774476376e-07,
+      "loss": 0.0038,
+      "step": 3336
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.1994641987335606e-07,
+      "loss": 0.0028,
+      "step": 3337
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.1992206527033606e-07,
+      "loss": 0.0029,
+      "step": 3338
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 4.198977106673161e-07,
+      "loss": 0.0034,
+      "step": 3339
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.198733560642961e-07,
+      "loss": 0.0032,
+      "step": 3340
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.1984900146127616e-07,
+      "loss": 0.0029,
+      "step": 3341
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.198246468582562e-07,
+      "loss": 0.0032,
+      "step": 3342
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.198002922552362e-07,
+      "loss": 1.0652,
+      "step": 3343
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.1977593765221627e-07,
+      "loss": 0.0032,
+      "step": 3344
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.1975158304919626e-07,
+      "loss": 0.0028,
+      "step": 3345
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.197272284461763e-07,
+      "loss": 0.0026,
+      "step": 3346
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.197028738431563e-07,
+      "loss": 0.0032,
+      "step": 3347
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.1967851924013637e-07,
+      "loss": 0.0036,
+      "step": 3348
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 4.196541646371164e-07,
+      "loss": 0.0064,
+      "step": 3349
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.196298100340964e-07,
+      "loss": 0.0029,
+      "step": 3350
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.1960545543107647e-07,
+      "loss": 0.003,
+      "step": 3351
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.1958110082805647e-07,
+      "loss": 0.0049,
+      "step": 3352
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.195567462250365e-07,
+      "loss": 0.0037,
+      "step": 3353
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.195323916220165e-07,
+      "loss": 0.0024,
+      "step": 3354
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.1950803701899657e-07,
+      "loss": 0.0054,
+      "step": 3355
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.1948368241597657e-07,
+      "loss": 0.0033,
+      "step": 3356
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.194593278129566e-07,
+      "loss": 0.0028,
+      "step": 3357
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.1943497320993667e-07,
+      "loss": 0.0036,
+      "step": 3358
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 4.1941061860691667e-07,
+      "loss": 0.0029,
+      "step": 3359
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.193862640038967e-07,
+      "loss": 0.0271,
+      "step": 3360
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.193619094008767e-07,
+      "loss": 0.0042,
+      "step": 3361
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.1933755479785677e-07,
+      "loss": 0.0022,
+      "step": 3362
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.1931320019483677e-07,
+      "loss": 0.0026,
+      "step": 3363
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.192888455918168e-07,
+      "loss": 0.0029,
+      "step": 3364
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.1926449098879687e-07,
+      "loss": 0.0028,
+      "step": 3365
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.1924013638577687e-07,
+      "loss": 0.003,
+      "step": 3366
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.192157817827569e-07,
+      "loss": 0.0031,
+      "step": 3367
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.191914271797369e-07,
+      "loss": 0.0031,
+      "step": 3368
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 4.19167072576717e-07,
+      "loss": 0.0032,
+      "step": 3369
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.1914271797369697e-07,
+      "loss": 0.1868,
+      "step": 3370
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.19118363370677e-07,
+      "loss": 0.0029,
+      "step": 3371
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.19094008767657e-07,
+      "loss": 0.0024,
+      "step": 3372
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.190696541646371e-07,
+      "loss": 0.003,
+      "step": 3373
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.190452995616171e-07,
+      "loss": 0.0029,
+      "step": 3374
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.190209449585971e-07,
+      "loss": 0.0025,
+      "step": 3375
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.189965903555772e-07,
+      "loss": 0.0028,
+      "step": 3376
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.189722357525572e-07,
+      "loss": 0.0028,
+      "step": 3377
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.1894788114953723e-07,
+      "loss": 0.003,
+      "step": 3378
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.189235265465172e-07,
+      "loss": 0.0036,
+      "step": 3379
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 4.188991719434973e-07,
+      "loss": 0.0035,
+      "step": 3380
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.1887481734047733e-07,
+      "loss": 0.0023,
+      "step": 3381
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.1885046273745733e-07,
+      "loss": 0.0027,
+      "step": 3382
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.1882610813443743e-07,
+      "loss": 0.0034,
+      "step": 3383
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.1880175353141743e-07,
+      "loss": 0.003,
+      "step": 3384
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.187773989283975e-07,
+      "loss": 0.0026,
+      "step": 3385
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.187530443253775e-07,
+      "loss": 0.0026,
+      "step": 3386
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.1872868972235753e-07,
+      "loss": 0.003,
+      "step": 3387
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.187043351193376e-07,
+      "loss": 0.0032,
+      "step": 3388
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.186799805163176e-07,
+      "loss": 0.0033,
+      "step": 3389
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 4.1865562591329763e-07,
+      "loss": 0.0024,
+      "step": 3390
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.1863127131027763e-07,
+      "loss": 0.0033,
+      "step": 3391
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.186069167072577e-07,
+      "loss": 0.003,
+      "step": 3392
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.185825621042377e-07,
+      "loss": 0.0038,
+      "step": 3393
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.1855820750121774e-07,
+      "loss": 0.003,
+      "step": 3394
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.1853385289819773e-07,
+      "loss": 0.0034,
+      "step": 3395
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.185094982951778e-07,
+      "loss": 0.0031,
+      "step": 3396
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.1848514369215784e-07,
+      "loss": 0.003,
+      "step": 3397
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.1846078908913784e-07,
+      "loss": 0.0029,
+      "step": 3398
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.184364344861179e-07,
+      "loss": 0.0037,
+      "step": 3399
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 4.184120798830979e-07,
+      "loss": 0.0047,
+      "step": 3400
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.1838772528007794e-07,
+      "loss": 0.003,
+      "step": 3401
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.1836337067705794e-07,
+      "loss": 0.0025,
+      "step": 3402
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.18339016074038e-07,
+      "loss": 0.0032,
+      "step": 3403
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.1831466147101804e-07,
+      "loss": 0.003,
+      "step": 3404
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.1829030686799804e-07,
+      "loss": 0.003,
+      "step": 3405
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.182659522649781e-07,
+      "loss": 0.0027,
+      "step": 3406
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.182415976619581e-07,
+      "loss": 0.0022,
+      "step": 3407
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.1821724305893814e-07,
+      "loss": 0.0048,
+      "step": 3408
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.1819288845591814e-07,
+      "loss": 0.0033,
+      "step": 3409
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.181685338528982e-07,
+      "loss": 0.0031,
+      "step": 3410
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 4.181441792498782e-07,
+      "loss": 0.0024,
+      "step": 3411
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.1811982464685824e-07,
+      "loss": 0.0025,
+      "step": 3412
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.180954700438383e-07,
+      "loss": 0.0021,
+      "step": 3413
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.180711154408183e-07,
+      "loss": 0.0024,
+      "step": 3414
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.1804676083779834e-07,
+      "loss": 0.0028,
+      "step": 3415
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.1802240623477834e-07,
+      "loss": 0.0031,
+      "step": 3416
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.179980516317584e-07,
+      "loss": 0.0032,
+      "step": 3417
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.179736970287384e-07,
+      "loss": 0.0027,
+      "step": 3418
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.1794934242571844e-07,
+      "loss": 0.0027,
+      "step": 3419
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.179249878226985e-07,
+      "loss": 0.0027,
+      "step": 3420
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 4.179006332196785e-07,
+      "loss": 0.0031,
+      "step": 3421
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.1787627861665855e-07,
+      "loss": 0.0038,
+      "step": 3422
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.1785192401363854e-07,
+      "loss": 0.0076,
+      "step": 3423
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.178275694106186e-07,
+      "loss": 0.004,
+      "step": 3424
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.178032148075986e-07,
+      "loss": 0.0027,
+      "step": 3425
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.1777886020457865e-07,
+      "loss": 1.891,
+      "step": 3426
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.1775450560155865e-07,
+      "loss": 0.0025,
+      "step": 3427
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.177301509985387e-07,
+      "loss": 0.0027,
+      "step": 3428
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.1770579639551875e-07,
+      "loss": 0.319,
+      "step": 3429
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.1768144179249875e-07,
+      "loss": 0.003,
+      "step": 3430
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 4.176570871894788e-07,
+      "loss": 0.0024,
+      "step": 3431
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.176327325864588e-07,
+      "loss": 0.0039,
+      "step": 3432
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.1760837798343885e-07,
+      "loss": 0.0026,
+      "step": 3433
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.1758402338041885e-07,
+      "loss": 0.0036,
+      "step": 3434
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.175596687773989e-07,
+      "loss": 0.003,
+      "step": 3435
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.1753531417437895e-07,
+      "loss": 0.0026,
+      "step": 3436
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.1751095957135895e-07,
+      "loss": 0.0021,
+      "step": 3437
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.17486604968339e-07,
+      "loss": 0.0036,
+      "step": 3438
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.17462250365319e-07,
+      "loss": 0.0033,
+      "step": 3439
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.1743789576229905e-07,
+      "loss": 0.0024,
+      "step": 3440
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.1741354115927905e-07,
+      "loss": 0.0023,
+      "step": 3441
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 4.173891865562591e-07,
+      "loss": 0.0031,
+      "step": 3442
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.173648319532391e-07,
+      "loss": 0.0029,
+      "step": 3443
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.1734047735021915e-07,
+      "loss": 0.0025,
+      "step": 3444
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.173161227471992e-07,
+      "loss": 0.0026,
+      "step": 3445
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.172917681441792e-07,
+      "loss": 0.0023,
+      "step": 3446
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.1726741354115925e-07,
+      "loss": 0.0058,
+      "step": 3447
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.1724305893813925e-07,
+      "loss": 0.5858,
+      "step": 3448
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.172187043351193e-07,
+      "loss": 0.0026,
+      "step": 3449
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.171943497320993e-07,
+      "loss": 0.0025,
+      "step": 3450
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.1716999512907936e-07,
+      "loss": 0.0025,
+      "step": 3451
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 4.171456405260594e-07,
+      "loss": 0.0025,
+      "step": 3452
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.171212859230394e-07,
+      "loss": 0.0029,
+      "step": 3453
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.1709693132001946e-07,
+      "loss": 0.0025,
+      "step": 3454
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.1707257671699946e-07,
+      "loss": 0.0033,
+      "step": 3455
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.170482221139795e-07,
+      "loss": 0.0026,
+      "step": 3456
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.170238675109595e-07,
+      "loss": 0.0028,
+      "step": 3457
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.1699951290793956e-07,
+      "loss": 0.0026,
+      "step": 3458
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.1697515830491966e-07,
+      "loss": 0.0027,
+      "step": 3459
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.1695080370189966e-07,
+      "loss": 0.003,
+      "step": 3460
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.169264490988797e-07,
+      "loss": 0.0034,
+      "step": 3461
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 4.169020944958597e-07,
+      "loss": 0.0028,
+      "step": 3462
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.1687773989283976e-07,
+      "loss": 0.0477,
+      "step": 3463
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.1685338528981976e-07,
+      "loss": 0.0024,
+      "step": 3464
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.168290306867998e-07,
+      "loss": 0.0029,
+      "step": 3465
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.168046760837798e-07,
+      "loss": 0.003,
+      "step": 3466
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.1678032148075986e-07,
+      "loss": 0.0031,
+      "step": 3467
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.167559668777399e-07,
+      "loss": 0.0022,
+      "step": 3468
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.167316122747199e-07,
+      "loss": 0.0047,
+      "step": 3469
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.1670725767169997e-07,
+      "loss": 0.0026,
+      "step": 3470
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.1668290306867996e-07,
+      "loss": 0.0024,
+      "step": 3471
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 4.1665854846566e-07,
+      "loss": 0.0678,
+      "step": 3472
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.1663419386264e-07,
+      "loss": 0.0036,
+      "step": 3473
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.1660983925962007e-07,
+      "loss": 0.0032,
+      "step": 3474
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.165854846566001e-07,
+      "loss": 0.0023,
+      "step": 3475
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.165611300535801e-07,
+      "loss": 0.0025,
+      "step": 3476
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.1653677545056017e-07,
+      "loss": 0.0029,
+      "step": 3477
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.1651242084754017e-07,
+      "loss": 0.0028,
+      "step": 3478
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.164880662445202e-07,
+      "loss": 0.0032,
+      "step": 3479
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.164637116415002e-07,
+      "loss": 0.0023,
+      "step": 3480
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.1643935703848027e-07,
+      "loss": 0.0029,
+      "step": 3481
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.164150024354603e-07,
+      "loss": 0.0028,
+      "step": 3482
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 4.163906478324403e-07,
+      "loss": 0.0029,
+      "step": 3483
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.1636629322942037e-07,
+      "loss": 0.0028,
+      "step": 3484
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.1634193862640037e-07,
+      "loss": 0.0028,
+      "step": 3485
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.163175840233804e-07,
+      "loss": 0.0058,
+      "step": 3486
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.162932294203604e-07,
+      "loss": 0.0028,
+      "step": 3487
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.1626887481734047e-07,
+      "loss": 0.6188,
+      "step": 3488
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.1624452021432047e-07,
+      "loss": 0.0022,
+      "step": 3489
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.162201656113005e-07,
+      "loss": 0.0395,
+      "step": 3490
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.161958110082806e-07,
+      "loss": 0.0025,
+      "step": 3491
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.1617145640526057e-07,
+      "loss": 0.0027,
+      "step": 3492
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 4.161471018022406e-07,
+      "loss": 0.0037,
+      "step": 3493
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.161227471992206e-07,
+      "loss": 0.0024,
+      "step": 3494
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.160983925962007e-07,
+      "loss": 0.0371,
+      "step": 3495
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.1607403799318067e-07,
+      "loss": 0.0028,
+      "step": 3496
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.160496833901607e-07,
+      "loss": 0.0359,
+      "step": 3497
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.160253287871408e-07,
+      "loss": 0.0024,
+      "step": 3498
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.160009741841208e-07,
+      "loss": 0.0029,
+      "step": 3499
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.1597661958110083e-07,
+      "loss": 0.0031,
+      "step": 3500
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.159522649780808e-07,
+      "loss": 0.0029,
+      "step": 3501
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.159279103750609e-07,
+      "loss": 0.0028,
+      "step": 3502
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 4.159035557720409e-07,
+      "loss": 0.0031,
+      "step": 3503
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.1587920116902093e-07,
+      "loss": 0.0021,
+      "step": 3504
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.158548465660009e-07,
+      "loss": 0.0026,
+      "step": 3505
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.15830491962981e-07,
+      "loss": 0.0026,
+      "step": 3506
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.1580613735996103e-07,
+      "loss": 0.0026,
+      "step": 3507
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.1578178275694103e-07,
+      "loss": 0.0027,
+      "step": 3508
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.157574281539211e-07,
+      "loss": 0.0028,
+      "step": 3509
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.157330735509011e-07,
+      "loss": 0.0024,
+      "step": 3510
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.1570871894788113e-07,
+      "loss": 0.0128,
+      "step": 3511
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.1568436434486113e-07,
+      "loss": 0.002,
+      "step": 3512
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.156600097418412e-07,
+      "loss": 0.0029,
+      "step": 3513
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 4.1563565513882123e-07,
+      "loss": 0.0028,
+      "step": 3514
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.1561130053580123e-07,
+      "loss": 0.016,
+      "step": 3515
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.155869459327813e-07,
+      "loss": 0.9659,
+      "step": 3516
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.155625913297613e-07,
+      "loss": 0.0024,
+      "step": 3517
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.1553823672674133e-07,
+      "loss": 0.0031,
+      "step": 3518
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.1551388212372133e-07,
+      "loss": 0.0038,
+      "step": 3519
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.154895275207014e-07,
+      "loss": 0.0028,
+      "step": 3520
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.154651729176814e-07,
+      "loss": 0.0045,
+      "step": 3521
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.1544081831466143e-07,
+      "loss": 0.0026,
+      "step": 3522
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.154164637116415e-07,
+      "loss": 0.0027,
+      "step": 3523
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 4.153921091086215e-07,
+      "loss": 0.0033,
+      "step": 3524
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.1536775450560154e-07,
+      "loss": 0.0024,
+      "step": 3525
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.1534339990258153e-07,
+      "loss": 0.0025,
+      "step": 3526
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.153190452995616e-07,
+      "loss": 0.0031,
+      "step": 3527
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.152946906965416e-07,
+      "loss": 0.0043,
+      "step": 3528
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.1527033609352164e-07,
+      "loss": 0.0026,
+      "step": 3529
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.152459814905017e-07,
+      "loss": 0.0034,
+      "step": 3530
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.152216268874817e-07,
+      "loss": 0.003,
+      "step": 3531
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.1519727228446174e-07,
+      "loss": 0.0027,
+      "step": 3532
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.1517291768144174e-07,
+      "loss": 0.0024,
+      "step": 3533
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 4.151485630784218e-07,
+      "loss": 0.0076,
+      "step": 3534
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.151242084754018e-07,
+      "loss": 0.0031,
+      "step": 3535
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.150998538723819e-07,
+      "loss": 0.003,
+      "step": 3536
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.1507549926936194e-07,
+      "loss": 0.0025,
+      "step": 3537
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.1505114466634194e-07,
+      "loss": 0.9911,
+      "step": 3538
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.15026790063322e-07,
+      "loss": 0.0036,
+      "step": 3539
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.15002435460302e-07,
+      "loss": 0.0024,
+      "step": 3540
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.1497808085728204e-07,
+      "loss": 0.0025,
+      "step": 3541
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.1495372625426204e-07,
+      "loss": 0.0045,
+      "step": 3542
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.149293716512421e-07,
+      "loss": 0.0026,
+      "step": 3543
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 4.149050170482221e-07,
+      "loss": 0.0031,
+      "step": 3544
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.1488066244520214e-07,
+      "loss": 0.0022,
+      "step": 3545
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.148563078421822e-07,
+      "loss": 0.0027,
+      "step": 3546
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.148319532391622e-07,
+      "loss": 0.0022,
+      "step": 3547
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.1480759863614225e-07,
+      "loss": 0.0029,
+      "step": 3548
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.1478324403312224e-07,
+      "loss": 0.0033,
+      "step": 3549
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.147588894301023e-07,
+      "loss": 0.0031,
+      "step": 3550
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.147345348270823e-07,
+      "loss": 0.003,
+      "step": 3551
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.1471018022406235e-07,
+      "loss": 0.0127,
+      "step": 3552
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.146858256210424e-07,
+      "loss": 0.0034,
+      "step": 3553
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.146614710180224e-07,
+      "loss": 0.0028,
+      "step": 3554
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 4.1463711641500245e-07,
+      "loss": 0.0033,
+      "step": 3555
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.1461276181198245e-07,
+      "loss": 0.0027,
+      "step": 3556
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.145884072089625e-07,
+      "loss": 0.0023,
+      "step": 3557
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.145640526059425e-07,
+      "loss": 0.0031,
+      "step": 3558
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.1453969800292255e-07,
+      "loss": 0.0029,
+      "step": 3559
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.1451534339990255e-07,
+      "loss": 0.0032,
+      "step": 3560
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.144909887968826e-07,
+      "loss": 0.0024,
+      "step": 3561
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.1446663419386265e-07,
+      "loss": 0.0034,
+      "step": 3562
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.1444227959084265e-07,
+      "loss": 0.013,
+      "step": 3563
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.144179249878227e-07,
+      "loss": 0.003,
+      "step": 3564
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 4.143935703848027e-07,
+      "loss": 0.0028,
+      "step": 3565
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.1436921578178275e-07,
+      "loss": 0.003,
+      "step": 3566
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.1434486117876275e-07,
+      "loss": 0.0028,
+      "step": 3567
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.143205065757428e-07,
+      "loss": 0.0023,
+      "step": 3568
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.1429615197272285e-07,
+      "loss": 0.0024,
+      "step": 3569
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.1427179736970285e-07,
+      "loss": 0.0023,
+      "step": 3570
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.142474427666829e-07,
+      "loss": 0.0045,
+      "step": 3571
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.142230881636629e-07,
+      "loss": 0.0024,
+      "step": 3572
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.1419873356064295e-07,
+      "loss": 0.0024,
+      "step": 3573
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.1417437895762295e-07,
+      "loss": 0.0022,
+      "step": 3574
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 4.14150024354603e-07,
+      "loss": 0.0023,
+      "step": 3575
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.14125669751583e-07,
+      "loss": 0.0021,
+      "step": 3576
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.1410131514856306e-07,
+      "loss": 0.0023,
+      "step": 3577
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.140769605455431e-07,
+      "loss": 0.0023,
+      "step": 3578
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.140526059425231e-07,
+      "loss": 0.0032,
+      "step": 3579
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.1402825133950316e-07,
+      "loss": 0.0247,
+      "step": 3580
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.1400389673648316e-07,
+      "loss": 0.0024,
+      "step": 3581
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.139795421334632e-07,
+      "loss": 0.0025,
+      "step": 3582
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.139551875304432e-07,
+      "loss": 0.0026,
+      "step": 3583
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.1393083292742326e-07,
+      "loss": 0.0023,
+      "step": 3584
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.139064783244033e-07,
+      "loss": 0.0021,
+      "step": 3585
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 4.138821237213833e-07,
+      "loss": 0.0031,
+      "step": 3586
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.1385776911836336e-07,
+      "loss": 0.5346,
+      "step": 3587
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.1383341451534336e-07,
+      "loss": 0.0027,
+      "step": 3588
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.138090599123234e-07,
+      "loss": 0.0029,
+      "step": 3589
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.137847053093034e-07,
+      "loss": 0.0025,
+      "step": 3590
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.1376035070628346e-07,
+      "loss": 0.3208,
+      "step": 3591
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.1373599610326346e-07,
+      "loss": 0.0023,
+      "step": 3592
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.137116415002435e-07,
+      "loss": 0.0027,
+      "step": 3593
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.1368728689722356e-07,
+      "loss": 0.0024,
+      "step": 3594
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.1366293229420356e-07,
+      "loss": 0.0025,
+      "step": 3595
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 4.136385776911836e-07,
+      "loss": 0.0027,
+      "step": 3596
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.136142230881636e-07,
+      "loss": 0.0025,
+      "step": 3597
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.1358986848514366e-07,
+      "loss": 0.5817,
+      "step": 3598
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.1356551388212366e-07,
+      "loss": 0.0025,
+      "step": 3599
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.135411592791037e-07,
+      "loss": 0.0021,
+      "step": 3600
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.1351680467608377e-07,
+      "loss": 0.0022,
+      "step": 3601
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.1349245007306376e-07,
+      "loss": 0.0023,
+      "step": 3602
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.134680954700438e-07,
+      "loss": 0.0019,
+      "step": 3603
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.134437408670238e-07,
+      "loss": 0.0021,
+      "step": 3604
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.1341938626400387e-07,
+      "loss": 0.0032,
+      "step": 3605
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 4.1339503166098386e-07,
+      "loss": 0.0023,
+      "step": 3606
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.133706770579639e-07,
+      "loss": 0.003,
+      "step": 3607
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.133463224549439e-07,
+      "loss": 0.0254,
+      "step": 3608
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.1332196785192397e-07,
+      "loss": 0.0029,
+      "step": 3609
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.13297613248904e-07,
+      "loss": 0.0022,
+      "step": 3610
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.13273258645884e-07,
+      "loss": 0.0025,
+      "step": 3611
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.1324890404286407e-07,
+      "loss": 0.0022,
+      "step": 3612
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.132245494398441e-07,
+      "loss": 0.002,
+      "step": 3613
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.1320019483682417e-07,
+      "loss": 0.0033,
+      "step": 3614
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.1317584023380417e-07,
+      "loss": 0.0028,
+      "step": 3615
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 4.131514856307842e-07,
+      "loss": 0.0025,
+      "step": 3616
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.131271310277643e-07,
+      "loss": 0.0022,
+      "step": 3617
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.1310277642474427e-07,
+      "loss": 0.0812,
+      "step": 3618
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.130784218217243e-07,
+      "loss": 0.0025,
+      "step": 3619
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.130540672187043e-07,
+      "loss": 0.0044,
+      "step": 3620
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.130297126156844e-07,
+      "loss": 0.0023,
+      "step": 3621
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.1300535801266437e-07,
+      "loss": 0.0028,
+      "step": 3622
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.129810034096444e-07,
+      "loss": 0.6051,
+      "step": 3623
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.129566488066245e-07,
+      "loss": 0.0027,
+      "step": 3624
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.129322942036045e-07,
+      "loss": 0.0022,
+      "step": 3625
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.1290793960058453e-07,
+      "loss": 0.0019,
+      "step": 3626
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 4.128835849975645e-07,
+      "loss": 0.0027,
+      "step": 3627
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.128592303945446e-07,
+      "loss": 0.0026,
+      "step": 3628
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.128348757915246e-07,
+      "loss": 0.0026,
+      "step": 3629
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.1281052118850463e-07,
+      "loss": 0.0021,
+      "step": 3630
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.127861665854846e-07,
+      "loss": 0.0028,
+      "step": 3631
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.127618119824647e-07,
+      "loss": 0.0027,
+      "step": 3632
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.1273745737944473e-07,
+      "loss": 0.0025,
+      "step": 3633
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.1271310277642473e-07,
+      "loss": 0.0053,
+      "step": 3634
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.126887481734048e-07,
+      "loss": 0.0027,
+      "step": 3635
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.126643935703848e-07,
+      "loss": 0.003,
+      "step": 3636
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 4.1264003896736483e-07,
+      "loss": 0.002,
+      "step": 3637
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.1261568436434483e-07,
+      "loss": 0.0022,
+      "step": 3638
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.125913297613249e-07,
+      "loss": 0.0031,
+      "step": 3639
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.1256697515830493e-07,
+      "loss": 0.0031,
+      "step": 3640
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.1254262055528493e-07,
+      "loss": 0.0033,
+      "step": 3641
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.12518265952265e-07,
+      "loss": 0.9192,
+      "step": 3642
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.12493911349245e-07,
+      "loss": 0.0033,
+      "step": 3643
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.1246955674622503e-07,
+      "loss": 0.0021,
+      "step": 3644
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.1244520214320503e-07,
+      "loss": 0.0025,
+      "step": 3645
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.124208475401851e-07,
+      "loss": 0.0028,
+      "step": 3646
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 4.123964929371651e-07,
+      "loss": 0.0027,
+      "step": 3647
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.1237213833414513e-07,
+      "loss": 0.0032,
+      "step": 3648
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.123477837311252e-07,
+      "loss": 0.0026,
+      "step": 3649
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.123234291281052e-07,
+      "loss": 1.2173,
+      "step": 3650
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.1229907452508524e-07,
+      "loss": 0.0025,
+      "step": 3651
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.1227471992206523e-07,
+      "loss": 0.0027,
+      "step": 3652
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.122503653190453e-07,
+      "loss": 0.003,
+      "step": 3653
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.122260107160253e-07,
+      "loss": 0.0111,
+      "step": 3654
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.1220165611300534e-07,
+      "loss": 0.0034,
+      "step": 3655
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.121773015099854e-07,
+      "loss": 0.0027,
+      "step": 3656
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.121529469069654e-07,
+      "loss": 0.0028,
+      "step": 3657
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 4.1212859230394544e-07,
+      "loss": 0.0026,
+      "step": 3658
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.1210423770092544e-07,
+      "loss": 0.0024,
+      "step": 3659
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.120798830979055e-07,
+      "loss": 0.003,
+      "step": 3660
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.120555284948855e-07,
+      "loss": 0.0024,
+      "step": 3661
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.1203117389186554e-07,
+      "loss": 0.0023,
+      "step": 3662
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.120068192888456e-07,
+      "loss": 0.0068,
+      "step": 3663
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.119824646858256e-07,
+      "loss": 0.0026,
+      "step": 3664
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.1195811008280564e-07,
+      "loss": 0.0021,
+      "step": 3665
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.1193375547978564e-07,
+      "loss": 0.0023,
+      "step": 3666
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.119094008767657e-07,
+      "loss": 0.0026,
+      "step": 3667
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 4.118850462737457e-07,
+      "loss": 0.0022,
+      "step": 3668
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.1186069167072574e-07,
+      "loss": 0.0021,
+      "step": 3669
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.1183633706770574e-07,
+      "loss": 0.0023,
+      "step": 3670
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.118119824646858e-07,
+      "loss": 0.0031,
+      "step": 3671
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.1178762786166584e-07,
+      "loss": 0.0021,
+      "step": 3672
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.1176327325864584e-07,
+      "loss": 0.0026,
+      "step": 3673
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.117389186556259e-07,
+      "loss": 0.004,
+      "step": 3674
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.117145640526059e-07,
+      "loss": 0.0019,
+      "step": 3675
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.1169020944958594e-07,
+      "loss": 0.0021,
+      "step": 3676
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.1166585484656594e-07,
+      "loss": 0.0024,
+      "step": 3677
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 4.11641500243546e-07,
+      "loss": 0.0021,
+      "step": 3678
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.1161714564052605e-07,
+      "loss": 0.2537,
+      "step": 3679
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.1159279103750604e-07,
+      "loss": 0.002,
+      "step": 3680
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.115684364344861e-07,
+      "loss": 0.003,
+      "step": 3681
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.115440818314661e-07,
+      "loss": 0.0025,
+      "step": 3682
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.1151972722844615e-07,
+      "loss": 0.0024,
+      "step": 3683
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.1149537262542615e-07,
+      "loss": 0.0024,
+      "step": 3684
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.114710180224062e-07,
+      "loss": 0.0017,
+      "step": 3685
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.114466634193862e-07,
+      "loss": 0.0044,
+      "step": 3686
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.1142230881636625e-07,
+      "loss": 0.0029,
+      "step": 3687
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 4.113979542133463e-07,
+      "loss": 0.0026,
+      "step": 3688
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.1137359961032635e-07,
+      "loss": 0.0032,
+      "step": 3689
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.113492450073064e-07,
+      "loss": 0.0375,
+      "step": 3690
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.113248904042864e-07,
+      "loss": 0.0027,
+      "step": 3691
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.1130053580126645e-07,
+      "loss": 0.0023,
+      "step": 3692
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.1127618119824645e-07,
+      "loss": 0.0021,
+      "step": 3693
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.112518265952265e-07,
+      "loss": 0.0021,
+      "step": 3694
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.1122747199220655e-07,
+      "loss": 0.0025,
+      "step": 3695
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.1120311738918655e-07,
+      "loss": 0.0033,
+      "step": 3696
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.111787627861666e-07,
+      "loss": 0.0021,
+      "step": 3697
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.111544081831466e-07,
+      "loss": 0.0023,
+      "step": 3698
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 4.1113005358012665e-07,
+      "loss": 0.0022,
+      "step": 3699
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.1110569897710665e-07,
+      "loss": 0.0021,
+      "step": 3700
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.110813443740867e-07,
+      "loss": 0.0024,
+      "step": 3701
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.1105698977106676e-07,
+      "loss": 0.0025,
+      "step": 3702
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.1103263516804676e-07,
+      "loss": 0.0025,
+      "step": 3703
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.110082805650268e-07,
+      "loss": 0.0022,
+      "step": 3704
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.109839259620068e-07,
+      "loss": 0.0021,
+      "step": 3705
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.1095957135898686e-07,
+      "loss": 0.0603,
+      "step": 3706
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.1093521675596686e-07,
+      "loss": 0.0022,
+      "step": 3707
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.109108621529469e-07,
+      "loss": 0.0021,
+      "step": 3708
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 4.108865075499269e-07,
+      "loss": 0.0026,
+      "step": 3709
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.1086215294690696e-07,
+      "loss": 0.0023,
+      "step": 3710
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.10837798343887e-07,
+      "loss": 0.0022,
+      "step": 3711
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.10813443740867e-07,
+      "loss": 0.0039,
+      "step": 3712
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.1078908913784706e-07,
+      "loss": 0.0026,
+      "step": 3713
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.1076473453482706e-07,
+      "loss": 0.0019,
+      "step": 3714
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.107403799318071e-07,
+      "loss": 0.0023,
+      "step": 3715
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.107160253287871e-07,
+      "loss": 0.0023,
+      "step": 3716
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.1069167072576716e-07,
+      "loss": 0.0024,
+      "step": 3717
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.106673161227472e-07,
+      "loss": 0.0019,
+      "step": 3718
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 4.106429615197272e-07,
+      "loss": 0.0023,
+      "step": 3719
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.1061860691670726e-07,
+      "loss": 0.0018,
+      "step": 3720
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.1059425231368726e-07,
+      "loss": 0.0022,
+      "step": 3721
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.105698977106673e-07,
+      "loss": 0.0031,
+      "step": 3722
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.105455431076473e-07,
+      "loss": 0.0022,
+      "step": 3723
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.1052118850462736e-07,
+      "loss": 0.0023,
+      "step": 3724
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.1049683390160736e-07,
+      "loss": 0.003,
+      "step": 3725
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.104724792985874e-07,
+      "loss": 0.0056,
+      "step": 3726
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.1044812469556747e-07,
+      "loss": 0.0037,
+      "step": 3727
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.1042377009254746e-07,
+      "loss": 0.0022,
+      "step": 3728
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.103994154895275e-07,
+      "loss": 0.0019,
+      "step": 3729
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 4.103750608865075e-07,
+      "loss": 0.002,
+      "step": 3730
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.1035070628348757e-07,
+      "loss": 0.0023,
+      "step": 3731
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.1032635168046756e-07,
+      "loss": 0.0025,
+      "step": 3732
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.103019970774476e-07,
+      "loss": 0.0024,
+      "step": 3733
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.1027764247442767e-07,
+      "loss": 0.0047,
+      "step": 3734
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.1025328787140767e-07,
+      "loss": 0.0028,
+      "step": 3735
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.102289332683877e-07,
+      "loss": 0.002,
+      "step": 3736
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.102045786653677e-07,
+      "loss": 0.0024,
+      "step": 3737
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.1018022406234777e-07,
+      "loss": 0.0022,
+      "step": 3738
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.1015586945932777e-07,
+      "loss": 0.0024,
+      "step": 3739
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 4.101315148563078e-07,
+      "loss": 0.002,
+      "step": 3740
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.101071602532878e-07,
+      "loss": 0.0021,
+      "step": 3741
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.1008280565026787e-07,
+      "loss": 0.0018,
+      "step": 3742
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.100584510472479e-07,
+      "loss": 0.0021,
+      "step": 3743
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.100340964442279e-07,
+      "loss": 0.0022,
+      "step": 3744
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.1000974184120797e-07,
+      "loss": 0.0024,
+      "step": 3745
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.0998538723818797e-07,
+      "loss": 0.0027,
+      "step": 3746
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.09961032635168e-07,
+      "loss": 0.0025,
+      "step": 3747
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.09936678032148e-07,
+      "loss": 0.0023,
+      "step": 3748
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.0991232342912807e-07,
+      "loss": 0.0028,
+      "step": 3749
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 4.098879688261081e-07,
+      "loss": 0.0023,
+      "step": 3750
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.098636142230881e-07,
+      "loss": 0.0024,
+      "step": 3751
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.098392596200682e-07,
+      "loss": 0.0027,
+      "step": 3752
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.0981490501704817e-07,
+      "loss": 0.0024,
+      "step": 3753
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.097905504140282e-07,
+      "loss": 0.002,
+      "step": 3754
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.097661958110082e-07,
+      "loss": 0.0039,
+      "step": 3755
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.097418412079883e-07,
+      "loss": 0.002,
+      "step": 3756
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.0971748660496827e-07,
+      "loss": 0.0026,
+      "step": 3757
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.096931320019483e-07,
+      "loss": 0.0025,
+      "step": 3758
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.096687773989284e-07,
+      "loss": 0.0065,
+      "step": 3759
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 4.096444227959084e-07,
+      "loss": 0.0034,
+      "step": 3760
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.0962006819288843e-07,
+      "loss": 0.0022,
+      "step": 3761
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.095957135898684e-07,
+      "loss": 0.0021,
+      "step": 3762
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.095713589868485e-07,
+      "loss": 0.002,
+      "step": 3763
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.095470043838285e-07,
+      "loss": 0.0027,
+      "step": 3764
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.0952264978080853e-07,
+      "loss": 0.003,
+      "step": 3765
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.0949829517778863e-07,
+      "loss": 0.0022,
+      "step": 3766
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.0947394057476863e-07,
+      "loss": 0.0025,
+      "step": 3767
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.094495859717487e-07,
+      "loss": 0.0023,
+      "step": 3768
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.094252313687287e-07,
+      "loss": 0.0018,
+      "step": 3769
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.0940087676570873e-07,
+      "loss": 0.0023,
+      "step": 3770
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 4.0937652216268873e-07,
+      "loss": 0.0021,
+      "step": 3771
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.093521675596688e-07,
+      "loss": 0.0022,
+      "step": 3772
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.0932781295664883e-07,
+      "loss": 0.0022,
+      "step": 3773
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.0930345835362883e-07,
+      "loss": 0.0024,
+      "step": 3774
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.092791037506089e-07,
+      "loss": 0.0021,
+      "step": 3775
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.092547491475889e-07,
+      "loss": 0.0022,
+      "step": 3776
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.0923039454456894e-07,
+      "loss": 0.002,
+      "step": 3777
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.0920603994154893e-07,
+      "loss": 0.0024,
+      "step": 3778
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.09181685338529e-07,
+      "loss": 0.003,
+      "step": 3779
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.09157330735509e-07,
+      "loss": 0.0022,
+      "step": 3780
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 4.0913297613248904e-07,
+      "loss": 0.0023,
+      "step": 3781
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.091086215294691e-07,
+      "loss": 0.0025,
+      "step": 3782
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.090842669264491e-07,
+      "loss": 0.0024,
+      "step": 3783
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.0905991232342914e-07,
+      "loss": 0.0024,
+      "step": 3784
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.0903555772040914e-07,
+      "loss": 0.0021,
+      "step": 3785
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.090112031173892e-07,
+      "loss": 0.002,
+      "step": 3786
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.089868485143692e-07,
+      "loss": 0.0023,
+      "step": 3787
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.0896249391134924e-07,
+      "loss": 0.0023,
+      "step": 3788
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.089381393083293e-07,
+      "loss": 0.0024,
+      "step": 3789
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.089137847053093e-07,
+      "loss": 0.0025,
+      "step": 3790
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 4.0888943010228934e-07,
+      "loss": 0.0027,
+      "step": 3791
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.0886507549926934e-07,
+      "loss": 0.0026,
+      "step": 3792
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.088407208962494e-07,
+      "loss": 0.0028,
+      "step": 3793
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.088163662932294e-07,
+      "loss": 0.0026,
+      "step": 3794
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.0879201169020944e-07,
+      "loss": 0.003,
+      "step": 3795
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.0876765708718944e-07,
+      "loss": 1.002,
+      "step": 3796
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.087433024841695e-07,
+      "loss": 0.0023,
+      "step": 3797
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.0871894788114954e-07,
+      "loss": 0.0019,
+      "step": 3798
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.0869459327812954e-07,
+      "loss": 0.0023,
+      "step": 3799
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.086702386751096e-07,
+      "loss": 0.0028,
+      "step": 3800
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.086458840720896e-07,
+      "loss": 0.0028,
+      "step": 3801
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 4.0862152946906964e-07,
+      "loss": 0.0022,
+      "step": 3802
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.0859717486604964e-07,
+      "loss": 0.0029,
+      "step": 3803
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.085728202630297e-07,
+      "loss": 0.0026,
+      "step": 3804
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.0854846566000975e-07,
+      "loss": 0.0019,
+      "step": 3805
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.0852411105698974e-07,
+      "loss": 0.0019,
+      "step": 3806
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.084997564539698e-07,
+      "loss": 0.0022,
+      "step": 3807
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.084754018509498e-07,
+      "loss": 0.0022,
+      "step": 3808
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.0845104724792985e-07,
+      "loss": 0.0028,
+      "step": 3809
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.0842669264490985e-07,
+      "loss": 0.6675,
+      "step": 3810
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.084023380418899e-07,
+      "loss": 0.003,
+      "step": 3811
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 4.083779834388699e-07,
+      "loss": 0.0019,
+      "step": 3812
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.0835362883584995e-07,
+      "loss": 0.0028,
+      "step": 3813
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.0832927423283e-07,
+      "loss": 0.0021,
+      "step": 3814
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.0830491962981e-07,
+      "loss": 0.0021,
+      "step": 3815
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.0828056502679005e-07,
+      "loss": 0.0022,
+      "step": 3816
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.0825621042377005e-07,
+      "loss": 0.0022,
+      "step": 3817
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.082318558207501e-07,
+      "loss": 0.0021,
+      "step": 3818
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.082075012177301e-07,
+      "loss": 0.0018,
+      "step": 3819
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.0818314661471015e-07,
+      "loss": 0.0026,
+      "step": 3820
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.081587920116902e-07,
+      "loss": 0.0025,
+      "step": 3821
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.081344374086702e-07,
+      "loss": 0.0023,
+      "step": 3822
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.0811008280565025e-07,
+      "loss": 0.0034,
+      "step": 3823
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.0808572820263025e-07,
+      "loss": 0.0035,
+      "step": 3824
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.080613735996103e-07,
+      "loss": 0.0021,
+      "step": 3825
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.080370189965903e-07,
+      "loss": 0.003,
+      "step": 3826
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.0801266439357035e-07,
+      "loss": 0.0022,
+      "step": 3827
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.079883097905504e-07,
+      "loss": 0.0035,
+      "step": 3828
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.079639551875304e-07,
+      "loss": 0.0019,
+      "step": 3829
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.0793960058451045e-07,
+      "loss": 0.002,
+      "step": 3830
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.0791524598149045e-07,
+      "loss": 0.0021,
+      "step": 3831
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.078908913784705e-07,
+      "loss": 0.0024,
+      "step": 3832
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 4.078665367754505e-07,
+      "loss": 0.0021,
+      "step": 3833
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.0784218217243056e-07,
+      "loss": 0.002,
+      "step": 3834
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.0781782756941055e-07,
+      "loss": 0.0022,
+      "step": 3835
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.077934729663906e-07,
+      "loss": 0.0021,
+      "step": 3836
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.0776911836337066e-07,
+      "loss": 0.0018,
+      "step": 3837
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.0774476376035066e-07,
+      "loss": 0.0024,
+      "step": 3838
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.077204091573307e-07,
+      "loss": 0.0028,
+      "step": 3839
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.076960545543107e-07,
+      "loss": 1.3881,
+      "step": 3840
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.0767169995129076e-07,
+      "loss": 0.0022,
+      "step": 3841
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.076473453482708e-07,
+      "loss": 0.0022,
+      "step": 3842
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.0762299074525086e-07,
+      "loss": 0.0166,
+      "step": 3843
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.075986361422309e-07,
+      "loss": 0.0019,
+      "step": 3844
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.075742815392109e-07,
+      "loss": 0.0019,
+      "step": 3845
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.0754992693619096e-07,
+      "loss": 0.0024,
+      "step": 3846
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.0752557233317096e-07,
+      "loss": 0.0021,
+      "step": 3847
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.07501217730151e-07,
+      "loss": 0.002,
+      "step": 3848
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.07476863127131e-07,
+      "loss": 0.0025,
+      "step": 3849
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.0745250852411106e-07,
+      "loss": 0.0025,
+      "step": 3850
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.074281539210911e-07,
+      "loss": 0.002,
+      "step": 3851
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.074037993180711e-07,
+      "loss": 0.014,
+      "step": 3852
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 4.0737944471505117e-07,
+      "loss": 0.0016,
+      "step": 3853
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.0735509011203116e-07,
+      "loss": 0.0019,
+      "step": 3854
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.073307355090112e-07,
+      "loss": 0.0032,
+      "step": 3855
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.073063809059912e-07,
+      "loss": 0.0022,
+      "step": 3856
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.0728202630297127e-07,
+      "loss": 0.0023,
+      "step": 3857
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.0725767169995127e-07,
+      "loss": 0.0022,
+      "step": 3858
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.072333170969313e-07,
+      "loss": 0.0018,
+      "step": 3859
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.0720896249391137e-07,
+      "loss": 0.0024,
+      "step": 3860
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.0718460789089137e-07,
+      "loss": 1.5796,
+      "step": 3861
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.071602532878714e-07,
+      "loss": 0.0025,
+      "step": 3862
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.071358986848514e-07,
+      "loss": 0.0021,
+      "step": 3863
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.0711154408183147e-07,
+      "loss": 0.0018,
+      "step": 3864
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.0708718947881147e-07,
+      "loss": 0.0023,
+      "step": 3865
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.070628348757915e-07,
+      "loss": 0.0024,
+      "step": 3866
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.0703848027277157e-07,
+      "loss": 0.0037,
+      "step": 3867
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.0701412566975157e-07,
+      "loss": 0.0019,
+      "step": 3868
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.069897710667316e-07,
+      "loss": 0.0023,
+      "step": 3869
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.069654164637116e-07,
+      "loss": 0.0024,
+      "step": 3870
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.0694106186069167e-07,
+      "loss": 0.2009,
+      "step": 3871
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.0691670725767167e-07,
+      "loss": 0.0019,
+      "step": 3872
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.068923526546517e-07,
+      "loss": 0.0022,
+      "step": 3873
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 4.068679980516317e-07,
+      "loss": 0.0018,
+      "step": 3874
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.0684364344861177e-07,
+      "loss": 0.0026,
+      "step": 3875
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.068192888455918e-07,
+      "loss": 0.003,
+      "step": 3876
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.067949342425718e-07,
+      "loss": 0.0021,
+      "step": 3877
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.067705796395519e-07,
+      "loss": 0.002,
+      "step": 3878
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.0674622503653187e-07,
+      "loss": 0.0022,
+      "step": 3879
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.067218704335119e-07,
+      "loss": 0.0018,
+      "step": 3880
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.066975158304919e-07,
+      "loss": 0.0024,
+      "step": 3881
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.06673161227472e-07,
+      "loss": 0.0027,
+      "step": 3882
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.0664880662445203e-07,
+      "loss": 0.0023,
+      "step": 3883
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 4.06624452021432e-07,
+      "loss": 0.0019,
+      "step": 3884
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.066000974184121e-07,
+      "loss": 0.0021,
+      "step": 3885
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.065757428153921e-07,
+      "loss": 0.0021,
+      "step": 3886
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.0655138821237213e-07,
+      "loss": 0.0024,
+      "step": 3887
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.065270336093521e-07,
+      "loss": 0.0022,
+      "step": 3888
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.065026790063322e-07,
+      "loss": 0.0026,
+      "step": 3889
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.064783244033122e-07,
+      "loss": 0.0024,
+      "step": 3890
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.0645396980029223e-07,
+      "loss": 0.0019,
+      "step": 3891
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.064296151972723e-07,
+      "loss": 0.0246,
+      "step": 3892
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.064052605942523e-07,
+      "loss": 0.0025,
+      "step": 3893
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 4.0638090599123233e-07,
+      "loss": 0.002,
+      "step": 3894
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.0635655138821233e-07,
+      "loss": 0.0025,
+      "step": 3895
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.063321967851924e-07,
+      "loss": 0.002,
+      "step": 3896
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.063078421821724e-07,
+      "loss": 0.0029,
+      "step": 3897
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.0628348757915243e-07,
+      "loss": 0.002,
+      "step": 3898
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.062591329761325e-07,
+      "loss": 0.0024,
+      "step": 3899
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.062347783731125e-07,
+      "loss": 0.0022,
+      "step": 3900
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.0621042377009253e-07,
+      "loss": 0.0022,
+      "step": 3901
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.0618606916707253e-07,
+      "loss": 0.003,
+      "step": 3902
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.061617145640526e-07,
+      "loss": 0.0018,
+      "step": 3903
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.061373599610326e-07,
+      "loss": 0.0022,
+      "step": 3904
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 4.0611300535801263e-07,
+      "loss": 0.002,
+      "step": 3905
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.0608865075499263e-07,
+      "loss": 0.002,
+      "step": 3906
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.060642961519727e-07,
+      "loss": 0.0017,
+      "step": 3907
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.0603994154895274e-07,
+      "loss": 0.0023,
+      "step": 3908
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.0601558694593273e-07,
+      "loss": 0.0019,
+      "step": 3909
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.059912323429128e-07,
+      "loss": 0.0024,
+      "step": 3910
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.059668777398928e-07,
+      "loss": 0.0021,
+      "step": 3911
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.0594252313687284e-07,
+      "loss": 0.0019,
+      "step": 3912
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.0591816853385283e-07,
+      "loss": 0.0019,
+      "step": 3913
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.058938139308329e-07,
+      "loss": 0.0023,
+      "step": 3914
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 4.0586945932781294e-07,
+      "loss": 0.002,
+      "step": 3915
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.0584510472479294e-07,
+      "loss": 0.0019,
+      "step": 3916
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.05820750121773e-07,
+      "loss": 0.0026,
+      "step": 3917
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.05796395518753e-07,
+      "loss": 0.0026,
+      "step": 3918
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.057720409157331e-07,
+      "loss": 0.0032,
+      "step": 3919
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.057476863127131e-07,
+      "loss": 0.0019,
+      "step": 3920
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.0572333170969314e-07,
+      "loss": 0.0044,
+      "step": 3921
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.056989771066732e-07,
+      "loss": 0.0025,
+      "step": 3922
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.056746225036532e-07,
+      "loss": 0.0079,
+      "step": 3923
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.0565026790063324e-07,
+      "loss": 0.0021,
+      "step": 3924
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 4.0562591329761324e-07,
+      "loss": 0.002,
+      "step": 3925
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.056015586945933e-07,
+      "loss": 0.0026,
+      "step": 3926
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.055772040915733e-07,
+      "loss": 0.004,
+      "step": 3927
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.0555284948855334e-07,
+      "loss": 0.0032,
+      "step": 3928
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.0552849488553334e-07,
+      "loss": 0.0023,
+      "step": 3929
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.055041402825134e-07,
+      "loss": 0.0027,
+      "step": 3930
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.0547978567949345e-07,
+      "loss": 0.0023,
+      "step": 3931
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.0545543107647344e-07,
+      "loss": 0.0025,
+      "step": 3932
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.054310764734535e-07,
+      "loss": 0.0024,
+      "step": 3933
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.054067218704335e-07,
+      "loss": 0.0022,
+      "step": 3934
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 4.0538236726741355e-07,
+      "loss": 0.0022,
+      "step": 3935
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.0535801266439355e-07,
+      "loss": 0.0022,
+      "step": 3936
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.053336580613736e-07,
+      "loss": 0.0024,
+      "step": 3937
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.0530930345835365e-07,
+      "loss": 0.0019,
+      "step": 3938
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.0528494885533365e-07,
+      "loss": 0.002,
+      "step": 3939
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.052605942523137e-07,
+      "loss": 0.0028,
+      "step": 3940
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.052362396492937e-07,
+      "loss": 0.0019,
+      "step": 3941
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.0521188504627375e-07,
+      "loss": 0.0021,
+      "step": 3942
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.0518753044325375e-07,
+      "loss": 0.0019,
+      "step": 3943
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.051631758402338e-07,
+      "loss": 0.0019,
+      "step": 3944
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.051388212372138e-07,
+      "loss": 0.0023,
+      "step": 3945
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 4.0511446663419385e-07,
+      "loss": 0.0021,
+      "step": 3946
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.050901120311739e-07,
+      "loss": 0.885,
+      "step": 3947
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.050657574281539e-07,
+      "loss": 0.0023,
+      "step": 3948
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.0504140282513395e-07,
+      "loss": 0.0072,
+      "step": 3949
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.0501704822211395e-07,
+      "loss": 0.0022,
+      "step": 3950
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.04992693619094e-07,
+      "loss": 0.0024,
+      "step": 3951
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.04968339016074e-07,
+      "loss": 0.0021,
+      "step": 3952
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.0494398441305405e-07,
+      "loss": 0.0022,
+      "step": 3953
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.049196298100341e-07,
+      "loss": 0.0022,
+      "step": 3954
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.048952752070141e-07,
+      "loss": 0.0026,
+      "step": 3955
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 4.0487092060399415e-07,
+      "loss": 0.0023,
+      "step": 3956
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.0484656600097415e-07,
+      "loss": 0.0019,
+      "step": 3957
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.048222113979542e-07,
+      "loss": 0.0021,
+      "step": 3958
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.047978567949342e-07,
+      "loss": 0.0015,
+      "step": 3959
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.0477350219191426e-07,
+      "loss": 0.0036,
+      "step": 3960
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.0474914758889425e-07,
+      "loss": 0.0016,
+      "step": 3961
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.047247929858743e-07,
+      "loss": 0.0018,
+      "step": 3962
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.0470043838285436e-07,
+      "loss": 0.0023,
+      "step": 3963
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.0467608377983436e-07,
+      "loss": 0.0021,
+      "step": 3964
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.046517291768144e-07,
+      "loss": 0.0016,
+      "step": 3965
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 4.046273745737944e-07,
+      "loss": 0.0023,
+      "step": 3966
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.0460301997077446e-07,
+      "loss": 0.0039,
+      "step": 3967
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.0457866536775446e-07,
+      "loss": 0.0017,
+      "step": 3968
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.045543107647345e-07,
+      "loss": 0.0022,
+      "step": 3969
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.0452995616171456e-07,
+      "loss": 0.0019,
+      "step": 3970
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.0450560155869456e-07,
+      "loss": 0.0325,
+      "step": 3971
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.044812469556746e-07,
+      "loss": 0.0019,
+      "step": 3972
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.044568923526546e-07,
+      "loss": 0.002,
+      "step": 3973
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.0443253774963466e-07,
+      "loss": 0.0022,
+      "step": 3974
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.0440818314661466e-07,
+      "loss": 1.6834,
+      "step": 3975
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.043838285435947e-07,
+      "loss": 0.0018,
+      "step": 3976
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 4.043594739405747e-07,
+      "loss": 0.0024,
+      "step": 3977
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.0433511933755476e-07,
+      "loss": 0.0025,
+      "step": 3978
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.043107647345348e-07,
+      "loss": 0.0025,
+      "step": 3979
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.042864101315148e-07,
+      "loss": 0.0022,
+      "step": 3980
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.0426205552849486e-07,
+      "loss": 0.002,
+      "step": 3981
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.0423770092547486e-07,
+      "loss": 0.0023,
+      "step": 3982
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.042133463224549e-07,
+      "loss": 0.0033,
+      "step": 3983
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.041889917194349e-07,
+      "loss": 0.002,
+      "step": 3984
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.0416463711641496e-07,
+      "loss": 0.0019,
+      "step": 3985
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.04140282513395e-07,
+      "loss": 0.0022,
+      "step": 3986
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 4.04115927910375e-07,
+      "loss": 0.0021,
+      "step": 3987
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.0409157330735507e-07,
+      "loss": 0.0022,
+      "step": 3988
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.0406721870433506e-07,
+      "loss": 0.0021,
+      "step": 3989
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.040428641013151e-07,
+      "loss": 0.0024,
+      "step": 3990
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.040185094982951e-07,
+      "loss": 0.0019,
+      "step": 3991
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.0399415489527517e-07,
+      "loss": 0.0028,
+      "step": 3992
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.039698002922552e-07,
+      "loss": 0.0023,
+      "step": 3993
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.039454456892352e-07,
+      "loss": 0.0018,
+      "step": 3994
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.0392109108621527e-07,
+      "loss": 0.0019,
+      "step": 3995
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.038967364831953e-07,
+      "loss": 0.0022,
+      "step": 3996
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 4.0387238188017537e-07,
+      "loss": 0.0017,
+      "step": 3997
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.0384802727715537e-07,
+      "loss": 0.0018,
+      "step": 3998
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.038236726741354e-07,
+      "loss": 1.6688,
+      "step": 3999
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.037993180711154e-07,
+      "loss": 0.0024,
+      "step": 4000
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.0377496346809547e-07,
+      "loss": 0.0019,
+      "step": 4001
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.037506088650755e-07,
+      "loss": 0.002,
+      "step": 4002
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.037262542620555e-07,
+      "loss": 0.0016,
+      "step": 4003
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.037018996590356e-07,
+      "loss": 0.0024,
+      "step": 4004
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.0367754505601557e-07,
+      "loss": 0.0021,
+      "step": 4005
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.036531904529956e-07,
+      "loss": 0.0021,
+      "step": 4006
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 4.036288358499756e-07,
+      "loss": 0.0019,
+      "step": 4007
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.036044812469557e-07,
+      "loss": 0.9915,
+      "step": 4008
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.0358012664393573e-07,
+      "loss": 0.0022,
+      "step": 4009
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.035557720409157e-07,
+      "loss": 0.0026,
+      "step": 4010
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.035314174378958e-07,
+      "loss": 0.0018,
+      "step": 4011
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.035070628348758e-07,
+      "loss": 0.008,
+      "step": 4012
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.0348270823185583e-07,
+      "loss": 0.0022,
+      "step": 4013
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.034583536288358e-07,
+      "loss": 0.0021,
+      "step": 4014
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.034339990258159e-07,
+      "loss": 0.0021,
+      "step": 4015
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.034096444227959e-07,
+      "loss": 0.0063,
+      "step": 4016
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.0338528981977593e-07,
+      "loss": 0.2094,
+      "step": 4017
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 4.03360935216756e-07,
+      "loss": 0.0024,
+      "step": 4018
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.03336580613736e-07,
+      "loss": 0.0021,
+      "step": 4019
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.0331222601071603e-07,
+      "loss": 0.002,
+      "step": 4020
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.0328787140769603e-07,
+      "loss": 0.0019,
+      "step": 4021
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.032635168046761e-07,
+      "loss": 0.002,
+      "step": 4022
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.032391622016561e-07,
+      "loss": 0.0028,
+      "step": 4023
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.0321480759863613e-07,
+      "loss": 0.002,
+      "step": 4024
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.031904529956162e-07,
+      "loss": 0.0021,
+      "step": 4025
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.031660983925962e-07,
+      "loss": 0.006,
+      "step": 4026
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.0314174378957623e-07,
+      "loss": 0.0155,
+      "step": 4027
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 4.0311738918655623e-07,
+      "loss": 0.0019,
+      "step": 4028
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.030930345835363e-07,
+      "loss": 0.002,
+      "step": 4029
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.030686799805163e-07,
+      "loss": 0.0021,
+      "step": 4030
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.0304432537749633e-07,
+      "loss": 0.0022,
+      "step": 4031
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.030199707744764e-07,
+      "loss": 0.0019,
+      "step": 4032
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.029956161714564e-07,
+      "loss": 0.0025,
+      "step": 4033
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.0297126156843644e-07,
+      "loss": 0.0021,
+      "step": 4034
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.0294690696541643e-07,
+      "loss": 0.0016,
+      "step": 4035
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.029225523623965e-07,
+      "loss": 0.0098,
+      "step": 4036
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.028981977593765e-07,
+      "loss": 0.002,
+      "step": 4037
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 4.0287384315635654e-07,
+      "loss": 0.0019,
+      "step": 4038
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.0284948855333653e-07,
+      "loss": 0.002,
+      "step": 4039
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.028251339503166e-07,
+      "loss": 0.0019,
+      "step": 4040
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.0280077934729664e-07,
+      "loss": 0.0022,
+      "step": 4041
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.0277642474427664e-07,
+      "loss": 0.002,
+      "step": 4042
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.027520701412567e-07,
+      "loss": 0.0016,
+      "step": 4043
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.027277155382367e-07,
+      "loss": 0.0018,
+      "step": 4044
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.0270336093521674e-07,
+      "loss": 0.0023,
+      "step": 4045
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.0267900633219674e-07,
+      "loss": 0.0102,
+      "step": 4046
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.026546517291768e-07,
+      "loss": 0.0018,
+      "step": 4047
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.0263029712615684e-07,
+      "loss": 0.0019,
+      "step": 4048
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 4.0260594252313684e-07,
+      "loss": 0.0018,
+      "step": 4049
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.025815879201169e-07,
+      "loss": 0.0019,
+      "step": 4050
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.025572333170969e-07,
+      "loss": 0.0025,
+      "step": 4051
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.0253287871407694e-07,
+      "loss": 0.0018,
+      "step": 4052
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.0250852411105694e-07,
+      "loss": 0.0019,
+      "step": 4053
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.02484169508037e-07,
+      "loss": 0.0017,
+      "step": 4054
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.02459814905017e-07,
+      "loss": 0.0021,
+      "step": 4055
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.0243546030199704e-07,
+      "loss": 0.0018,
+      "step": 4056
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.024111056989771e-07,
+      "loss": 0.0024,
+      "step": 4057
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.023867510959571e-07,
+      "loss": 0.0023,
+      "step": 4058
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 4.0236239649293714e-07,
+      "loss": 0.0019,
+      "step": 4059
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.0233804188991714e-07,
+      "loss": 0.0019,
+      "step": 4060
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.023136872868972e-07,
+      "loss": 0.002,
+      "step": 4061
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.022893326838772e-07,
+      "loss": 1.0794,
+      "step": 4062
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.0226497808085724e-07,
+      "loss": 0.002,
+      "step": 4063
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.022406234778373e-07,
+      "loss": 0.0024,
+      "step": 4064
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.022162688748173e-07,
+      "loss": 0.0024,
+      "step": 4065
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.0219191427179735e-07,
+      "loss": 0.005,
+      "step": 4066
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.0216755966877735e-07,
+      "loss": 0.0025,
+      "step": 4067
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.021432050657574e-07,
+      "loss": 0.0019,
+      "step": 4068
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 4.021188504627374e-07,
+      "loss": 0.0021,
+      "step": 4069
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.0209449585971745e-07,
+      "loss": 0.0016,
+      "step": 4070
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.0207014125669745e-07,
+      "loss": 0.002,
+      "step": 4071
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.0204578665367755e-07,
+      "loss": 0.0018,
+      "step": 4072
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.020214320506576e-07,
+      "loss": 0.0023,
+      "step": 4073
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.019970774476376e-07,
+      "loss": 1.2354,
+      "step": 4074
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.0197272284461765e-07,
+      "loss": 0.0021,
+      "step": 4075
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.0194836824159765e-07,
+      "loss": 0.0018,
+      "step": 4076
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.019240136385777e-07,
+      "loss": 0.0023,
+      "step": 4077
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.018996590355577e-07,
+      "loss": 1.0537,
+      "step": 4078
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 4.0187530443253775e-07,
+      "loss": 0.0024,
+      "step": 4079
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.018509498295178e-07,
+      "loss": 0.781,
+      "step": 4080
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.018265952264978e-07,
+      "loss": 0.002,
+      "step": 4081
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.0180224062347785e-07,
+      "loss": 0.0019,
+      "step": 4082
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.0177788602045785e-07,
+      "loss": 0.0019,
+      "step": 4083
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.017535314174379e-07,
+      "loss": 0.002,
+      "step": 4084
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.017291768144179e-07,
+      "loss": 0.002,
+      "step": 4085
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.0170482221139796e-07,
+      "loss": 0.0019,
+      "step": 4086
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.01680467608378e-07,
+      "loss": 0.0023,
+      "step": 4087
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.01656113005358e-07,
+      "loss": 0.0019,
+      "step": 4088
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.0163175840233806e-07,
+      "loss": 0.0015,
+      "step": 4089
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 4.0160740379931806e-07,
+      "loss": 0.0017,
+      "step": 4090
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.015830491962981e-07,
+      "loss": 0.0024,
+      "step": 4091
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.015586945932781e-07,
+      "loss": 0.0021,
+      "step": 4092
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.0153433999025816e-07,
+      "loss": 0.0021,
+      "step": 4093
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.0150998538723816e-07,
+      "loss": 0.0022,
+      "step": 4094
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.014856307842182e-07,
+      "loss": 0.0018,
+      "step": 4095
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.0146127618119826e-07,
+      "loss": 0.0023,
+      "step": 4096
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.0143692157817826e-07,
+      "loss": 0.0029,
+      "step": 4097
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.014125669751583e-07,
+      "loss": 0.0045,
+      "step": 4098
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.013882123721383e-07,
+      "loss": 0.0016,
+      "step": 4099
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 4.0136385776911836e-07,
+      "loss": 0.0023,
+      "step": 4100
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.0133950316609836e-07,
+      "loss": 0.005,
+      "step": 4101
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.013151485630784e-07,
+      "loss": 0.0037,
+      "step": 4102
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.0129079396005846e-07,
+      "loss": 0.0018,
+      "step": 4103
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.0126643935703846e-07,
+      "loss": 0.0024,
+      "step": 4104
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.012420847540185e-07,
+      "loss": 0.0018,
+      "step": 4105
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.012177301509985e-07,
+      "loss": 0.002,
+      "step": 4106
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.0119337554797856e-07,
+      "loss": 0.0022,
+      "step": 4107
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.0116902094495856e-07,
+      "loss": 0.0024,
+      "step": 4108
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.011446663419386e-07,
+      "loss": 0.002,
+      "step": 4109
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 4.011203117389186e-07,
+      "loss": 0.0021,
+      "step": 4110
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.0109595713589866e-07,
+      "loss": 0.0025,
+      "step": 4111
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.010716025328787e-07,
+      "loss": 0.0018,
+      "step": 4112
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.010472479298587e-07,
+      "loss": 0.0458,
+      "step": 4113
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.0102289332683877e-07,
+      "loss": 0.0019,
+      "step": 4114
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.0099853872381876e-07,
+      "loss": 0.0019,
+      "step": 4115
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.009741841207988e-07,
+      "loss": 0.0017,
+      "step": 4116
+    },
+    {
+      "epoch": 4.0,
+      "eval_accuracy": 0.997569275644142,
+      "eval_loss": 0.022041579708456993,
+      "eval_runtime": 2.804,
+      "eval_samples_per_second": 733.582,
+      "eval_steps_per_second": 11.769,
+      "step": 4116
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.009498295177788e-07,
+      "loss": 0.0018,
+      "step": 4117
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.0092547491475887e-07,
+      "loss": 0.0016,
+      "step": 4118
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.009011203117389e-07,
+      "loss": 0.0022,
+      "step": 4119
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.008767657087189e-07,
+      "loss": 0.002,
+      "step": 4120
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 4.0085241110569897e-07,
+      "loss": 0.0023,
+      "step": 4121
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.0082805650267897e-07,
+      "loss": 0.0016,
+      "step": 4122
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.00803701899659e-07,
+      "loss": 0.0024,
+      "step": 4123
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.00779347296639e-07,
+      "loss": 0.0017,
+      "step": 4124
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.0075499269361907e-07,
+      "loss": 0.0022,
+      "step": 4125
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.0073063809059907e-07,
+      "loss": 0.0017,
+      "step": 4126
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.007062834875791e-07,
+      "loss": 0.0015,
+      "step": 4127
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.0068192888455917e-07,
+      "loss": 0.0021,
+      "step": 4128
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.0065757428153917e-07,
+      "loss": 0.0022,
+      "step": 4129
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.006332196785192e-07,
+      "loss": 0.0032,
+      "step": 4130
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 4.006088650754992e-07,
+      "loss": 0.0018,
+      "step": 4131
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.0058451047247927e-07,
+      "loss": 0.0021,
+      "step": 4132
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.0056015586945927e-07,
+      "loss": 0.0019,
+      "step": 4133
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.005358012664393e-07,
+      "loss": 1.6847,
+      "step": 4134
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.005114466634194e-07,
+      "loss": 0.0025,
+      "step": 4135
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.0048709206039937e-07,
+      "loss": 0.0019,
+      "step": 4136
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.004627374573794e-07,
+      "loss": 0.0017,
+      "step": 4137
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.004383828543594e-07,
+      "loss": 0.0024,
+      "step": 4138
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.004140282513395e-07,
+      "loss": 0.0021,
+      "step": 4139
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.0038967364831947e-07,
+      "loss": 0.0019,
+      "step": 4140
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 4.003653190452995e-07,
+      "loss": 0.0018,
+      "step": 4141
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.003409644422795e-07,
+      "loss": 0.0019,
+      "step": 4142
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.003166098392596e-07,
+      "loss": 0.0024,
+      "step": 4143
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.0029225523623963e-07,
+      "loss": 0.0016,
+      "step": 4144
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.002679006332196e-07,
+      "loss": 0.0019,
+      "step": 4145
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.002435460301997e-07,
+      "loss": 0.0015,
+      "step": 4146
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.002191914271797e-07,
+      "loss": 0.0017,
+      "step": 4147
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.0019483682415973e-07,
+      "loss": 0.0018,
+      "step": 4148
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.001704822211398e-07,
+      "loss": 0.0023,
+      "step": 4149
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.0014612761811983e-07,
+      "loss": 0.0019,
+      "step": 4150
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.001217730150999e-07,
+      "loss": 0.0023,
+      "step": 4151
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 4.000974184120799e-07,
+      "loss": 0.0018,
+      "step": 4152
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.0007306380905993e-07,
+      "loss": 0.0022,
+      "step": 4153
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.0004870920603993e-07,
+      "loss": 0.0016,
+      "step": 4154
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4.0002435460302e-07,
+      "loss": 0.002,
+      "step": 4155
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 4e-07,
+      "loss": 0.0021,
+      "step": 4156
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.9997564539698003e-07,
+      "loss": 0.0021,
+      "step": 4157
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.999512907939601e-07,
+      "loss": 0.0019,
+      "step": 4158
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.999269361909401e-07,
+      "loss": 0.0024,
+      "step": 4159
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.9990258158792014e-07,
+      "loss": 0.0021,
+      "step": 4160
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.9987822698490013e-07,
+      "loss": 0.0017,
+      "step": 4161
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 3.998538723818802e-07,
+      "loss": 0.0019,
+      "step": 4162
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.998295177788602e-07,
+      "loss": 0.003,
+      "step": 4163
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.9980516317584024e-07,
+      "loss": 0.0394,
+      "step": 4164
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.9978080857282023e-07,
+      "loss": 0.0019,
+      "step": 4165
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.997564539698003e-07,
+      "loss": 0.002,
+      "step": 4166
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.9973209936678034e-07,
+      "loss": 0.2239,
+      "step": 4167
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.9970774476376034e-07,
+      "loss": 0.0066,
+      "step": 4168
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.996833901607404e-07,
+      "loss": 0.0019,
+      "step": 4169
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.996590355577204e-07,
+      "loss": 0.002,
+      "step": 4170
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.9963468095470044e-07,
+      "loss": 0.0016,
+      "step": 4171
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 3.9961032635168044e-07,
+      "loss": 0.0018,
+      "step": 4172
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.995859717486605e-07,
+      "loss": 0.0019,
+      "step": 4173
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.9956161714564054e-07,
+      "loss": 0.0022,
+      "step": 4174
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.9953726254262054e-07,
+      "loss": 0.0023,
+      "step": 4175
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.995129079396006e-07,
+      "loss": 0.0021,
+      "step": 4176
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.994885533365806e-07,
+      "loss": 1.6858,
+      "step": 4177
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.9946419873356064e-07,
+      "loss": 0.002,
+      "step": 4178
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.9943984413054064e-07,
+      "loss": 0.0023,
+      "step": 4179
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.994154895275207e-07,
+      "loss": 0.0016,
+      "step": 4180
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.993911349245007e-07,
+      "loss": 0.0021,
+      "step": 4181
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 3.9936678032148074e-07,
+      "loss": 0.0018,
+      "step": 4182
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.993424257184608e-07,
+      "loss": 0.0021,
+      "step": 4183
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.993180711154408e-07,
+      "loss": 0.0016,
+      "step": 4184
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.9929371651242084e-07,
+      "loss": 0.0017,
+      "step": 4185
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.9926936190940084e-07,
+      "loss": 0.002,
+      "step": 4186
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.992450073063809e-07,
+      "loss": 0.0018,
+      "step": 4187
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.992206527033609e-07,
+      "loss": 0.0017,
+      "step": 4188
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.9919629810034094e-07,
+      "loss": 0.0018,
+      "step": 4189
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.99171943497321e-07,
+      "loss": 0.0022,
+      "step": 4190
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.99147588894301e-07,
+      "loss": 0.0039,
+      "step": 4191
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.9912323429128105e-07,
+      "loss": 0.0021,
+      "step": 4192
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 3.9909887968826105e-07,
+      "loss": 0.0019,
+      "step": 4193
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.990745250852411e-07,
+      "loss": 0.0019,
+      "step": 4194
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.990501704822211e-07,
+      "loss": 0.0017,
+      "step": 4195
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.9902581587920115e-07,
+      "loss": 0.276,
+      "step": 4196
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.990014612761812e-07,
+      "loss": 0.0019,
+      "step": 4197
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.989771066731612e-07,
+      "loss": 0.0022,
+      "step": 4198
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.9895275207014125e-07,
+      "loss": 0.0018,
+      "step": 4199
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.9892839746712125e-07,
+      "loss": 0.0016,
+      "step": 4200
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.989040428641013e-07,
+      "loss": 0.0021,
+      "step": 4201
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.988796882610813e-07,
+      "loss": 0.0019,
+      "step": 4202
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 3.9885533365806135e-07,
+      "loss": 0.0021,
+      "step": 4203
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.9883097905504135e-07,
+      "loss": 0.0298,
+      "step": 4204
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.988066244520214e-07,
+      "loss": 0.0196,
+      "step": 4205
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.9878226984900145e-07,
+      "loss": 0.0015,
+      "step": 4206
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.9875791524598145e-07,
+      "loss": 0.0019,
+      "step": 4207
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.987335606429615e-07,
+      "loss": 0.0029,
+      "step": 4208
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.987092060399415e-07,
+      "loss": 0.0017,
+      "step": 4209
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.9868485143692155e-07,
+      "loss": 0.002,
+      "step": 4210
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.9866049683390155e-07,
+      "loss": 0.0015,
+      "step": 4211
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.986361422308816e-07,
+      "loss": 0.0017,
+      "step": 4212
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 3.9861178762786165e-07,
+      "loss": 0.0023,
+      "step": 4213
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.9858743302484165e-07,
+      "loss": 0.002,
+      "step": 4214
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.985630784218217e-07,
+      "loss": 0.0024,
+      "step": 4215
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.985387238188017e-07,
+      "loss": 0.0019,
+      "step": 4216
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.9851436921578176e-07,
+      "loss": 0.0018,
+      "step": 4217
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.9849001461276175e-07,
+      "loss": 0.0021,
+      "step": 4218
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.984656600097418e-07,
+      "loss": 0.002,
+      "step": 4219
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.984413054067218e-07,
+      "loss": 0.0021,
+      "step": 4220
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.9841695080370186e-07,
+      "loss": 0.0019,
+      "step": 4221
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.983925962006819e-07,
+      "loss": 0.0021,
+      "step": 4222
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.983682415976619e-07,
+      "loss": 0.0019,
+      "step": 4223
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 3.9834388699464196e-07,
+      "loss": 0.0017,
+      "step": 4224
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.98319532391622e-07,
+      "loss": 0.0019,
+      "step": 4225
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.9829517778860206e-07,
+      "loss": 0.0024,
+      "step": 4226
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.9827082318558206e-07,
+      "loss": 0.0019,
+      "step": 4227
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.982464685825621e-07,
+      "loss": 0.0021,
+      "step": 4228
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.9822211397954216e-07,
+      "loss": 0.0019,
+      "step": 4229
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.9819775937652216e-07,
+      "loss": 0.0018,
+      "step": 4230
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.981734047735022e-07,
+      "loss": 0.0022,
+      "step": 4231
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.981490501704822e-07,
+      "loss": 0.0016,
+      "step": 4232
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.9812469556746226e-07,
+      "loss": 0.002,
+      "step": 4233
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 3.9810034096444226e-07,
+      "loss": 0.0019,
+      "step": 4234
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.980759863614223e-07,
+      "loss": 0.0019,
+      "step": 4235
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.9805163175840237e-07,
+      "loss": 0.002,
+      "step": 4236
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.9802727715538236e-07,
+      "loss": 0.0023,
+      "step": 4237
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.980029225523624e-07,
+      "loss": 0.0018,
+      "step": 4238
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.979785679493424e-07,
+      "loss": 0.0027,
+      "step": 4239
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.9795421334632247e-07,
+      "loss": 0.0019,
+      "step": 4240
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.9792985874330247e-07,
+      "loss": 0.0016,
+      "step": 4241
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.979055041402825e-07,
+      "loss": 0.0433,
+      "step": 4242
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.978811495372625e-07,
+      "loss": 0.0018,
+      "step": 4243
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 3.9785679493424257e-07,
+      "loss": 0.0018,
+      "step": 4244
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.978324403312226e-07,
+      "loss": 0.002,
+      "step": 4245
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.978080857282026e-07,
+      "loss": 0.0017,
+      "step": 4246
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.9778373112518267e-07,
+      "loss": 0.0021,
+      "step": 4247
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.9775937652216267e-07,
+      "loss": 0.0023,
+      "step": 4248
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.977350219191427e-07,
+      "loss": 0.0017,
+      "step": 4249
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.977106673161227e-07,
+      "loss": 0.0021,
+      "step": 4250
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.9768631271310277e-07,
+      "loss": 0.0017,
+      "step": 4251
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.976619581100828e-07,
+      "loss": 0.0018,
+      "step": 4252
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.976376035070628e-07,
+      "loss": 0.002,
+      "step": 4253
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 3.9761324890404287e-07,
+      "loss": 0.0017,
+      "step": 4254
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.9758889430102287e-07,
+      "loss": 0.0014,
+      "step": 4255
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.975645396980029e-07,
+      "loss": 0.0019,
+      "step": 4256
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.975401850949829e-07,
+      "loss": 0.0027,
+      "step": 4257
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.9751583049196297e-07,
+      "loss": 0.0014,
+      "step": 4258
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.9749147588894297e-07,
+      "loss": 0.0019,
+      "step": 4259
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.97467121285923e-07,
+      "loss": 0.002,
+      "step": 4260
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.974427666829031e-07,
+      "loss": 0.0015,
+      "step": 4261
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.9741841207988307e-07,
+      "loss": 0.0022,
+      "step": 4262
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.973940574768631e-07,
+      "loss": 0.0017,
+      "step": 4263
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.973697028738431e-07,
+      "loss": 0.0023,
+      "step": 4264
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 3.973453482708232e-07,
+      "loss": 0.0021,
+      "step": 4265
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.973209936678032e-07,
+      "loss": 0.0022,
+      "step": 4266
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.972966390647832e-07,
+      "loss": 0.002,
+      "step": 4267
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.972722844617633e-07,
+      "loss": 0.002,
+      "step": 4268
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.972479298587433e-07,
+      "loss": 0.0021,
+      "step": 4269
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.9722357525572333e-07,
+      "loss": 0.0015,
+      "step": 4270
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.971992206527033e-07,
+      "loss": 0.0028,
+      "step": 4271
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.971748660496834e-07,
+      "loss": 0.0017,
+      "step": 4272
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.971505114466634e-07,
+      "loss": 0.5097,
+      "step": 4273
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.9712615684364343e-07,
+      "loss": 0.0019,
+      "step": 4274
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 3.9710180224062343e-07,
+      "loss": 0.0029,
+      "step": 4275
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.970774476376035e-07,
+      "loss": 0.0018,
+      "step": 4276
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.9705309303458353e-07,
+      "loss": 0.002,
+      "step": 4277
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.9702873843156353e-07,
+      "loss": 0.0044,
+      "step": 4278
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.970043838285436e-07,
+      "loss": 0.0014,
+      "step": 4279
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.969800292255236e-07,
+      "loss": 0.0025,
+      "step": 4280
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.9695567462250363e-07,
+      "loss": 0.002,
+      "step": 4281
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.9693132001948363e-07,
+      "loss": 0.0038,
+      "step": 4282
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.969069654164637e-07,
+      "loss": 0.0018,
+      "step": 4283
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.9688261081344373e-07,
+      "loss": 0.0018,
+      "step": 4284
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 3.9685825621042373e-07,
+      "loss": 0.002,
+      "step": 4285
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.968339016074038e-07,
+      "loss": 0.0014,
+      "step": 4286
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.968095470043838e-07,
+      "loss": 0.0018,
+      "step": 4287
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.9678519240136383e-07,
+      "loss": 0.1848,
+      "step": 4288
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.9676083779834383e-07,
+      "loss": 0.0017,
+      "step": 4289
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.967364831953239e-07,
+      "loss": 0.0027,
+      "step": 4290
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.967121285923039e-07,
+      "loss": 0.0022,
+      "step": 4291
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.9668777398928393e-07,
+      "loss": 0.0018,
+      "step": 4292
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.96663419386264e-07,
+      "loss": 0.0017,
+      "step": 4293
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.96639064783244e-07,
+      "loss": 0.0016,
+      "step": 4294
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.9661471018022404e-07,
+      "loss": 0.7211,
+      "step": 4295
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 3.9659035557720403e-07,
+      "loss": 0.0361,
+      "step": 4296
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.965660009741841e-07,
+      "loss": 0.0018,
+      "step": 4297
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.965416463711641e-07,
+      "loss": 0.0071,
+      "step": 4298
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.9651729176814414e-07,
+      "loss": 0.0015,
+      "step": 4299
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.964929371651242e-07,
+      "loss": 0.0017,
+      "step": 4300
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.964685825621042e-07,
+      "loss": 0.0018,
+      "step": 4301
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.964442279590843e-07,
+      "loss": 0.0021,
+      "step": 4302
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.964198733560643e-07,
+      "loss": 0.0021,
+      "step": 4303
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.9639551875304434e-07,
+      "loss": 0.0022,
+      "step": 4304
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.9637116415002434e-07,
+      "loss": 0.0025,
+      "step": 4305
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 3.963468095470044e-07,
+      "loss": 0.0016,
+      "step": 4306
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.9632245494398444e-07,
+      "loss": 0.0022,
+      "step": 4307
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.9629810034096444e-07,
+      "loss": 0.0017,
+      "step": 4308
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.962737457379445e-07,
+      "loss": 0.0023,
+      "step": 4309
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.962493911349245e-07,
+      "loss": 0.0027,
+      "step": 4310
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.9622503653190454e-07,
+      "loss": 0.0017,
+      "step": 4311
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.9620068192888454e-07,
+      "loss": 0.0015,
+      "step": 4312
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.961763273258646e-07,
+      "loss": 0.0019,
+      "step": 4313
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.961519727228446e-07,
+      "loss": 0.002,
+      "step": 4314
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.9612761811982464e-07,
+      "loss": 0.0018,
+      "step": 4315
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 3.961032635168047e-07,
+      "loss": 1.7551,
+      "step": 4316
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.960789089137847e-07,
+      "loss": 0.0017,
+      "step": 4317
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.9605455431076475e-07,
+      "loss": 0.002,
+      "step": 4318
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.9603019970774475e-07,
+      "loss": 0.0018,
+      "step": 4319
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.960058451047248e-07,
+      "loss": 0.0016,
+      "step": 4320
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.959814905017048e-07,
+      "loss": 0.0025,
+      "step": 4321
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.9595713589868485e-07,
+      "loss": 0.0016,
+      "step": 4322
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.959327812956649e-07,
+      "loss": 0.0152,
+      "step": 4323
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.959084266926449e-07,
+      "loss": 0.0015,
+      "step": 4324
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.9588407208962495e-07,
+      "loss": 0.0018,
+      "step": 4325
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 3.9585971748660495e-07,
+      "loss": 0.0018,
+      "step": 4326
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.95835362883585e-07,
+      "loss": 0.0021,
+      "step": 4327
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.95811008280565e-07,
+      "loss": 0.0015,
+      "step": 4328
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.9578665367754505e-07,
+      "loss": 0.0017,
+      "step": 4329
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.9576229907452505e-07,
+      "loss": 0.0018,
+      "step": 4330
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.957379444715051e-07,
+      "loss": 0.0018,
+      "step": 4331
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.9571358986848515e-07,
+      "loss": 0.0018,
+      "step": 4332
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.9568923526546515e-07,
+      "loss": 0.0223,
+      "step": 4333
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.956648806624452e-07,
+      "loss": 0.0017,
+      "step": 4334
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.956405260594252e-07,
+      "loss": 0.0014,
+      "step": 4335
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.9561617145640525e-07,
+      "loss": 0.0027,
+      "step": 4336
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 3.9559181685338525e-07,
+      "loss": 0.0018,
+      "step": 4337
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.955674622503653e-07,
+      "loss": 0.0024,
+      "step": 4338
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.9554310764734535e-07,
+      "loss": 0.0017,
+      "step": 4339
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.9551875304432535e-07,
+      "loss": 0.0024,
+      "step": 4340
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.954943984413054e-07,
+      "loss": 0.0015,
+      "step": 4341
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.954700438382854e-07,
+      "loss": 0.0019,
+      "step": 4342
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.9544568923526546e-07,
+      "loss": 0.0018,
+      "step": 4343
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.9542133463224545e-07,
+      "loss": 0.002,
+      "step": 4344
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.953969800292255e-07,
+      "loss": 0.0051,
+      "step": 4345
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.953726254262055e-07,
+      "loss": 0.0018,
+      "step": 4346
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 3.9534827082318556e-07,
+      "loss": 0.002,
+      "step": 4347
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.953239162201656e-07,
+      "loss": 0.0037,
+      "step": 4348
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.952995616171456e-07,
+      "loss": 0.0018,
+      "step": 4349
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.9527520701412566e-07,
+      "loss": 0.002,
+      "step": 4350
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.9525085241110566e-07,
+      "loss": 0.0019,
+      "step": 4351
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.952264978080857e-07,
+      "loss": 0.0017,
+      "step": 4352
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.952021432050657e-07,
+      "loss": 0.002,
+      "step": 4353
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.9517778860204576e-07,
+      "loss": 0.0017,
+      "step": 4354
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.951534339990258e-07,
+      "loss": 0.002,
+      "step": 4355
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.951290793960058e-07,
+      "loss": 0.0022,
+      "step": 4356
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 3.9510472479298586e-07,
+      "loss": 0.0014,
+      "step": 4357
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.9508037018996586e-07,
+      "loss": 0.0015,
+      "step": 4358
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.950560155869459e-07,
+      "loss": 0.0021,
+      "step": 4359
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.950316609839259e-07,
+      "loss": 0.0016,
+      "step": 4360
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.9500730638090596e-07,
+      "loss": 0.0021,
+      "step": 4361
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.94982951777886e-07,
+      "loss": 0.1146,
+      "step": 4362
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.94958597174866e-07,
+      "loss": 0.0228,
+      "step": 4363
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.9493424257184606e-07,
+      "loss": 0.002,
+      "step": 4364
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.9490988796882606e-07,
+      "loss": 0.002,
+      "step": 4365
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.948855333658061e-07,
+      "loss": 0.0016,
+      "step": 4366
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.948611787627861e-07,
+      "loss": 0.0013,
+      "step": 4367
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 3.9483682415976616e-07,
+      "loss": 0.0015,
+      "step": 4368
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.9481246955674616e-07,
+      "loss": 0.0022,
+      "step": 4369
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.947881149537262e-07,
+      "loss": 0.0047,
+      "step": 4370
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.9476376035070627e-07,
+      "loss": 0.0016,
+      "step": 4371
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.9473940574768626e-07,
+      "loss": 0.002,
+      "step": 4372
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.947150511446663e-07,
+      "loss": 0.0016,
+      "step": 4373
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.946906965416463e-07,
+      "loss": 0.0017,
+      "step": 4374
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.9466634193862637e-07,
+      "loss": 0.0016,
+      "step": 4375
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.9464198733560637e-07,
+      "loss": 0.0018,
+      "step": 4376
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.946176327325864e-07,
+      "loss": 0.0014,
+      "step": 4377
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 3.9459327812956647e-07,
+      "loss": 0.0019,
+      "step": 4378
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.945689235265465e-07,
+      "loss": 0.9041,
+      "step": 4379
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.9454456892352657e-07,
+      "loss": 0.002,
+      "step": 4380
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.9452021432050657e-07,
+      "loss": 0.0016,
+      "step": 4381
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.944958597174866e-07,
+      "loss": 0.0017,
+      "step": 4382
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.944715051144666e-07,
+      "loss": 0.002,
+      "step": 4383
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.9444715051144667e-07,
+      "loss": 0.0014,
+      "step": 4384
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.9442279590842667e-07,
+      "loss": 0.0023,
+      "step": 4385
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.943984413054067e-07,
+      "loss": 0.0021,
+      "step": 4386
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.943740867023868e-07,
+      "loss": 0.0016,
+      "step": 4387
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 3.9434973209936677e-07,
+      "loss": 0.0016,
+      "step": 4388
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.943253774963468e-07,
+      "loss": 0.0034,
+      "step": 4389
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.943010228933268e-07,
+      "loss": 0.0017,
+      "step": 4390
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.942766682903069e-07,
+      "loss": 0.002,
+      "step": 4391
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.942523136872869e-07,
+      "loss": 0.0018,
+      "step": 4392
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.942279590842669e-07,
+      "loss": 2.1923,
+      "step": 4393
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.94203604481247e-07,
+      "loss": 0.0018,
+      "step": 4394
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.94179249878227e-07,
+      "loss": 0.0019,
+      "step": 4395
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.9415489527520703e-07,
+      "loss": 0.0017,
+      "step": 4396
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.94130540672187e-07,
+      "loss": 0.0014,
+      "step": 4397
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 3.941061860691671e-07,
+      "loss": 0.0017,
+      "step": 4398
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.940818314661471e-07,
+      "loss": 0.0019,
+      "step": 4399
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.9405747686312713e-07,
+      "loss": 0.002,
+      "step": 4400
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.940331222601072e-07,
+      "loss": 0.0015,
+      "step": 4401
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.940087676570872e-07,
+      "loss": 0.002,
+      "step": 4402
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.9398441305406723e-07,
+      "loss": 0.0014,
+      "step": 4403
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.9396005845104723e-07,
+      "loss": 0.0018,
+      "step": 4404
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.939357038480273e-07,
+      "loss": 0.0015,
+      "step": 4405
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.939113492450073e-07,
+      "loss": 0.0016,
+      "step": 4406
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.9388699464198733e-07,
+      "loss": 0.002,
+      "step": 4407
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.9386264003896733e-07,
+      "loss": 0.0017,
+      "step": 4408
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 3.938382854359474e-07,
+      "loss": 0.0022,
+      "step": 4409
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.9381393083292743e-07,
+      "loss": 1.3412,
+      "step": 4410
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.9378957622990743e-07,
+      "loss": 0.0017,
+      "step": 4411
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.937652216268875e-07,
+      "loss": 0.0016,
+      "step": 4412
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.937408670238675e-07,
+      "loss": 0.0015,
+      "step": 4413
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.9371651242084753e-07,
+      "loss": 0.0018,
+      "step": 4414
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.9369215781782753e-07,
+      "loss": 0.002,
+      "step": 4415
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.936678032148076e-07,
+      "loss": 0.0021,
+      "step": 4416
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.9364344861178764e-07,
+      "loss": 0.0022,
+      "step": 4417
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.9361909400876763e-07,
+      "loss": 0.0016,
+      "step": 4418
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 3.935947394057477e-07,
+      "loss": 0.0019,
+      "step": 4419
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.935703848027277e-07,
+      "loss": 0.0017,
+      "step": 4420
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.9354603019970774e-07,
+      "loss": 0.0015,
+      "step": 4421
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.9352167559668773e-07,
+      "loss": 0.0015,
+      "step": 4422
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.934973209936678e-07,
+      "loss": 0.0017,
+      "step": 4423
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.934729663906478e-07,
+      "loss": 0.0018,
+      "step": 4424
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.9344861178762784e-07,
+      "loss": 0.0019,
+      "step": 4425
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.934242571846079e-07,
+      "loss": 0.0016,
+      "step": 4426
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.933999025815879e-07,
+      "loss": 0.0016,
+      "step": 4427
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.9337554797856794e-07,
+      "loss": 0.0016,
+      "step": 4428
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 3.9335119337554794e-07,
+      "loss": 0.0017,
+      "step": 4429
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.93326838772528e-07,
+      "loss": 0.0017,
+      "step": 4430
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.93302484169508e-07,
+      "loss": 0.0017,
+      "step": 4431
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.9327812956648804e-07,
+      "loss": 0.0017,
+      "step": 4432
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.932537749634681e-07,
+      "loss": 0.0021,
+      "step": 4433
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.932294203604481e-07,
+      "loss": 0.002,
+      "step": 4434
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.9320506575742814e-07,
+      "loss": 0.0017,
+      "step": 4435
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.9318071115440814e-07,
+      "loss": 0.0018,
+      "step": 4436
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.931563565513882e-07,
+      "loss": 0.0018,
+      "step": 4437
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.931320019483682e-07,
+      "loss": 0.0031,
+      "step": 4438
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.9310764734534824e-07,
+      "loss": 0.0018,
+      "step": 4439
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 3.9308329274232824e-07,
+      "loss": 0.0019,
+      "step": 4440
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.930589381393083e-07,
+      "loss": 0.0017,
+      "step": 4441
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.9303458353628834e-07,
+      "loss": 0.0019,
+      "step": 4442
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.9301022893326834e-07,
+      "loss": 0.002,
+      "step": 4443
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.929858743302484e-07,
+      "loss": 0.0016,
+      "step": 4444
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.929615197272284e-07,
+      "loss": 0.0017,
+      "step": 4445
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.9293716512420844e-07,
+      "loss": 0.0017,
+      "step": 4446
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.9291281052118844e-07,
+      "loss": 0.0013,
+      "step": 4447
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.928884559181685e-07,
+      "loss": 0.003,
+      "step": 4448
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.9286410131514855e-07,
+      "loss": 0.0018,
+      "step": 4449
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 3.9283974671212855e-07,
+      "loss": 0.0017,
+      "step": 4450
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.928153921091086e-07,
+      "loss": 0.0014,
+      "step": 4451
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.927910375060886e-07,
+      "loss": 0.0016,
+      "step": 4452
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.9276668290306865e-07,
+      "loss": 0.0025,
+      "step": 4453
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.9274232830004865e-07,
+      "loss": 0.002,
+      "step": 4454
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.9271797369702875e-07,
+      "loss": 0.002,
+      "step": 4455
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.926936190940088e-07,
+      "loss": 0.0019,
+      "step": 4456
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.926692644909888e-07,
+      "loss": 0.0015,
+      "step": 4457
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.9264490988796885e-07,
+      "loss": 0.002,
+      "step": 4458
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.9262055528494885e-07,
+      "loss": 0.0018,
+      "step": 4459
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 3.925962006819289e-07,
+      "loss": 0.0019,
+      "step": 4460
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.925718460789089e-07,
+      "loss": 0.8531,
+      "step": 4461
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.9254749147588895e-07,
+      "loss": 0.0014,
+      "step": 4462
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.9252313687286895e-07,
+      "loss": 0.0017,
+      "step": 4463
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.92498782269849e-07,
+      "loss": 0.0023,
+      "step": 4464
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.9247442766682905e-07,
+      "loss": 0.0019,
+      "step": 4465
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.9245007306380905e-07,
+      "loss": 0.002,
+      "step": 4466
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.924257184607891e-07,
+      "loss": 0.0019,
+      "step": 4467
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.924013638577691e-07,
+      "loss": 0.0015,
+      "step": 4468
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.9237700925474916e-07,
+      "loss": 0.0876,
+      "step": 4469
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.9235265465172915e-07,
+      "loss": 0.0021,
+      "step": 4470
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 3.923283000487092e-07,
+      "loss": 0.0017,
+      "step": 4471
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.9230394544568926e-07,
+      "loss": 0.0016,
+      "step": 4472
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.9227959084266926e-07,
+      "loss": 0.0018,
+      "step": 4473
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.922552362396493e-07,
+      "loss": 1.8899,
+      "step": 4474
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.922308816366293e-07,
+      "loss": 0.0016,
+      "step": 4475
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.9220652703360936e-07,
+      "loss": 0.0017,
+      "step": 4476
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.9218217243058936e-07,
+      "loss": 0.0019,
+      "step": 4477
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.921578178275694e-07,
+      "loss": 0.0017,
+      "step": 4478
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.921334632245494e-07,
+      "loss": 0.0016,
+      "step": 4479
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.9210910862152946e-07,
+      "loss": 0.0015,
+      "step": 4480
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 3.920847540185095e-07,
+      "loss": 0.0018,
+      "step": 4481
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.920603994154895e-07,
+      "loss": 0.0015,
+      "step": 4482
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.9203604481246956e-07,
+      "loss": 0.0026,
+      "step": 4483
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.9201169020944956e-07,
+      "loss": 0.0016,
+      "step": 4484
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.919873356064296e-07,
+      "loss": 0.0015,
+      "step": 4485
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.919629810034096e-07,
+      "loss": 0.0017,
+      "step": 4486
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.9193862640038966e-07,
+      "loss": 0.0017,
+      "step": 4487
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.919142717973697e-07,
+      "loss": 0.0014,
+      "step": 4488
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.918899171943497e-07,
+      "loss": 0.0015,
+      "step": 4489
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.9186556259132976e-07,
+      "loss": 0.0026,
+      "step": 4490
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 3.9184120798830976e-07,
+      "loss": 0.0016,
+      "step": 4491
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.918168533852898e-07,
+      "loss": 0.0017,
+      "step": 4492
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.917924987822698e-07,
+      "loss": 0.0014,
+      "step": 4493
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.9176814417924986e-07,
+      "loss": 0.002,
+      "step": 4494
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.9174378957622986e-07,
+      "loss": 0.0023,
+      "step": 4495
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.917194349732099e-07,
+      "loss": 0.0019,
+      "step": 4496
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.9169508037018997e-07,
+      "loss": 0.0021,
+      "step": 4497
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.9167072576716996e-07,
+      "loss": 0.0016,
+      "step": 4498
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.9164637116415e-07,
+      "loss": 0.0015,
+      "step": 4499
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.9162201656113e-07,
+      "loss": 0.0024,
+      "step": 4500
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 3.9159766195811007e-07,
+      "loss": 0.0017,
+      "step": 4501
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.9157330735509007e-07,
+      "loss": 0.0128,
+      "step": 4502
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.915489527520701e-07,
+      "loss": 0.0018,
+      "step": 4503
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.9152459814905017e-07,
+      "loss": 0.0016,
+      "step": 4504
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.9150024354603017e-07,
+      "loss": 0.0016,
+      "step": 4505
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.914758889430102e-07,
+      "loss": 1.6304,
+      "step": 4506
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.914515343399902e-07,
+      "loss": 0.0014,
+      "step": 4507
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.9142717973697027e-07,
+      "loss": 0.0016,
+      "step": 4508
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.9140282513395027e-07,
+      "loss": 0.0018,
+      "step": 4509
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.913784705309303e-07,
+      "loss": 0.0015,
+      "step": 4510
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.913541159279103e-07,
+      "loss": 0.0312,
+      "step": 4511
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 3.9132976132489037e-07,
+      "loss": 0.0016,
+      "step": 4512
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.913054067218704e-07,
+      "loss": 0.0019,
+      "step": 4513
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.912810521188504e-07,
+      "loss": 0.0017,
+      "step": 4514
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.9125669751583047e-07,
+      "loss": 0.0016,
+      "step": 4515
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.9123234291281047e-07,
+      "loss": 0.0014,
+      "step": 4516
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.912079883097905e-07,
+      "loss": 0.0017,
+      "step": 4517
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.911836337067705e-07,
+      "loss": 0.0018,
+      "step": 4518
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.9115927910375057e-07,
+      "loss": 0.0016,
+      "step": 4519
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.911349245007306e-07,
+      "loss": 0.0019,
+      "step": 4520
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.911105698977106e-07,
+      "loss": 0.0022,
+      "step": 4521
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 3.910862152946907e-07,
+      "loss": 0.0018,
+      "step": 4522
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.9106186069167067e-07,
+      "loss": 0.0019,
+      "step": 4523
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.910375060886507e-07,
+      "loss": 0.0019,
+      "step": 4524
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.910131514856307e-07,
+      "loss": 0.0016,
+      "step": 4525
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.909887968826108e-07,
+      "loss": 0.0017,
+      "step": 4526
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.909644422795908e-07,
+      "loss": 0.0018,
+      "step": 4527
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.909400876765708e-07,
+      "loss": 0.0014,
+      "step": 4528
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.909157330735509e-07,
+      "loss": 0.0017,
+      "step": 4529
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.908913784705309e-07,
+      "loss": 0.002,
+      "step": 4530
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.9086702386751093e-07,
+      "loss": 0.0022,
+      "step": 4531
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 3.90842669264491e-07,
+      "loss": 0.0018,
+      "step": 4532
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.9081831466147103e-07,
+      "loss": 0.0017,
+      "step": 4533
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.9079396005845103e-07,
+      "loss": 0.0017,
+      "step": 4534
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.907696054554311e-07,
+      "loss": 0.0036,
+      "step": 4535
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.9074525085241113e-07,
+      "loss": 0.0018,
+      "step": 4536
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.9072089624939113e-07,
+      "loss": 0.0014,
+      "step": 4537
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.906965416463712e-07,
+      "loss": 0.002,
+      "step": 4538
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.906721870433512e-07,
+      "loss": 0.0025,
+      "step": 4539
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.9064783244033123e-07,
+      "loss": 0.0021,
+      "step": 4540
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.9062347783731123e-07,
+      "loss": 0.0016,
+      "step": 4541
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.905991232342913e-07,
+      "loss": 0.0015,
+      "step": 4542
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 3.9057476863127134e-07,
+      "loss": 0.0014,
+      "step": 4543
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.9055041402825133e-07,
+      "loss": 0.0021,
+      "step": 4544
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.905260594252314e-07,
+      "loss": 0.0015,
+      "step": 4545
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.905017048222114e-07,
+      "loss": 0.0017,
+      "step": 4546
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.9047735021919144e-07,
+      "loss": 0.0012,
+      "step": 4547
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.9045299561617143e-07,
+      "loss": 0.0018,
+      "step": 4548
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.904286410131515e-07,
+      "loss": 0.0015,
+      "step": 4549
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.904042864101315e-07,
+      "loss": 0.0013,
+      "step": 4550
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.9037993180711154e-07,
+      "loss": 0.0021,
+      "step": 4551
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.903555772040916e-07,
+      "loss": 0.0032,
+      "step": 4552
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 3.903312226010716e-07,
+      "loss": 0.0018,
+      "step": 4553
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.9030686799805164e-07,
+      "loss": 0.0015,
+      "step": 4554
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.9028251339503164e-07,
+      "loss": 0.0016,
+      "step": 4555
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.902581587920117e-07,
+      "loss": 0.0022,
+      "step": 4556
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.902338041889917e-07,
+      "loss": 0.002,
+      "step": 4557
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.9020944958597174e-07,
+      "loss": 0.0026,
+      "step": 4558
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.901850949829518e-07,
+      "loss": 0.0015,
+      "step": 4559
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.901607403799318e-07,
+      "loss": 0.0016,
+      "step": 4560
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.9013638577691184e-07,
+      "loss": 0.0018,
+      "step": 4561
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.9011203117389184e-07,
+      "loss": 0.0019,
+      "step": 4562
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 3.900876765708719e-07,
+      "loss": 0.0016,
+      "step": 4563
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.900633219678519e-07,
+      "loss": 0.0015,
+      "step": 4564
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.9003896736483194e-07,
+      "loss": 0.0016,
+      "step": 4565
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.90014612761812e-07,
+      "loss": 0.0031,
+      "step": 4566
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.89990258158792e-07,
+      "loss": 0.0014,
+      "step": 4567
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.8996590355577204e-07,
+      "loss": 0.0019,
+      "step": 4568
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.8994154895275204e-07,
+      "loss": 0.0019,
+      "step": 4569
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.899171943497321e-07,
+      "loss": 0.0023,
+      "step": 4570
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.898928397467121e-07,
+      "loss": 0.0023,
+      "step": 4571
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.8986848514369214e-07,
+      "loss": 0.0018,
+      "step": 4572
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 3.8984413054067214e-07,
+      "loss": 0.0019,
+      "step": 4573
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.898197759376522e-07,
+      "loss": 0.0016,
+      "step": 4574
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.8979542133463225e-07,
+      "loss": 0.0017,
+      "step": 4575
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.8977106673161225e-07,
+      "loss": 0.0014,
+      "step": 4576
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.897467121285923e-07,
+      "loss": 0.0024,
+      "step": 4577
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.897223575255723e-07,
+      "loss": 0.0016,
+      "step": 4578
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.8969800292255235e-07,
+      "loss": 0.0018,
+      "step": 4579
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.8967364831953235e-07,
+      "loss": 0.0017,
+      "step": 4580
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.896492937165124e-07,
+      "loss": 0.0018,
+      "step": 4581
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.8962493911349245e-07,
+      "loss": 0.0014,
+      "step": 4582
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.8960058451047245e-07,
+      "loss": 0.0016,
+      "step": 4583
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 3.895762299074525e-07,
+      "loss": 0.0021,
+      "step": 4584
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.895518753044325e-07,
+      "loss": 0.0019,
+      "step": 4585
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.8952752070141255e-07,
+      "loss": 0.0015,
+      "step": 4586
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.8950316609839255e-07,
+      "loss": 0.0017,
+      "step": 4587
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.894788114953726e-07,
+      "loss": 0.0013,
+      "step": 4588
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.894544568923526e-07,
+      "loss": 0.0014,
+      "step": 4589
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.8943010228933265e-07,
+      "loss": 0.0016,
+      "step": 4590
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.894057476863127e-07,
+      "loss": 0.0017,
+      "step": 4591
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.893813930832927e-07,
+      "loss": 0.0017,
+      "step": 4592
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.8935703848027275e-07,
+      "loss": 0.0016,
+      "step": 4593
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 3.8933268387725275e-07,
+      "loss": 0.0017,
+      "step": 4594
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.893083292742328e-07,
+      "loss": 0.1133,
+      "step": 4595
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.892839746712128e-07,
+      "loss": 0.0018,
+      "step": 4596
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.8925962006819285e-07,
+      "loss": 0.0013,
+      "step": 4597
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.892352654651729e-07,
+      "loss": 0.0014,
+      "step": 4598
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.892109108621529e-07,
+      "loss": 0.0014,
+      "step": 4599
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.8918655625913296e-07,
+      "loss": 0.0022,
+      "step": 4600
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.8916220165611295e-07,
+      "loss": 0.0013,
+      "step": 4601
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.89137847053093e-07,
+      "loss": 0.0015,
+      "step": 4602
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.89113492450073e-07,
+      "loss": 0.0017,
+      "step": 4603
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 3.8908913784705306e-07,
+      "loss": 0.0017,
+      "step": 4604
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.8906478324403305e-07,
+      "loss": 0.0017,
+      "step": 4605
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.890404286410131e-07,
+      "loss": 0.0013,
+      "step": 4606
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.8901607403799316e-07,
+      "loss": 0.002,
+      "step": 4607
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.889917194349732e-07,
+      "loss": 0.002,
+      "step": 4608
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.8896736483195326e-07,
+      "loss": 0.0014,
+      "step": 4609
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.8894301022893326e-07,
+      "loss": 0.0015,
+      "step": 4610
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.889186556259133e-07,
+      "loss": 0.0032,
+      "step": 4611
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.888943010228933e-07,
+      "loss": 0.0014,
+      "step": 4612
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.8886994641987336e-07,
+      "loss": 0.0117,
+      "step": 4613
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.888455918168534e-07,
+      "loss": 0.0016,
+      "step": 4614
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 3.888212372138334e-07,
+      "loss": 1.591,
+      "step": 4615
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.8879688261081346e-07,
+      "loss": 0.0016,
+      "step": 4616
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.8877252800779346e-07,
+      "loss": 0.0017,
+      "step": 4617
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.887481734047735e-07,
+      "loss": 0.0014,
+      "step": 4618
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.887238188017535e-07,
+      "loss": 0.0014,
+      "step": 4619
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.8869946419873356e-07,
+      "loss": 0.0024,
+      "step": 4620
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.886751095957136e-07,
+      "loss": 0.0016,
+      "step": 4621
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.886507549926936e-07,
+      "loss": 0.0017,
+      "step": 4622
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.8862640038967367e-07,
+      "loss": 0.0017,
+      "step": 4623
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.8860204578665367e-07,
+      "loss": 0.0013,
+      "step": 4624
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 3.885776911836337e-07,
+      "loss": 0.0015,
+      "step": 4625
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.885533365806137e-07,
+      "loss": 0.0025,
+      "step": 4626
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.8852898197759377e-07,
+      "loss": 0.0015,
+      "step": 4627
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.8850462737457377e-07,
+      "loss": 0.0014,
+      "step": 4628
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.884802727715538e-07,
+      "loss": 0.0018,
+      "step": 4629
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.8845591816853387e-07,
+      "loss": 0.0014,
+      "step": 4630
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.8843156356551387e-07,
+      "loss": 0.0013,
+      "step": 4631
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.884072089624939e-07,
+      "loss": 0.0014,
+      "step": 4632
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.883828543594739e-07,
+      "loss": 0.0018,
+      "step": 4633
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.8835849975645397e-07,
+      "loss": 0.0016,
+      "step": 4634
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 3.8833414515343397e-07,
+      "loss": 0.0015,
+      "step": 4635
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.88309790550414e-07,
+      "loss": 0.0018,
+      "step": 4636
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.8828543594739407e-07,
+      "loss": 0.0012,
+      "step": 4637
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.8826108134437407e-07,
+      "loss": 0.0013,
+      "step": 4638
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.882367267413541e-07,
+      "loss": 0.0016,
+      "step": 4639
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.882123721383341e-07,
+      "loss": 0.0016,
+      "step": 4640
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.8818801753531417e-07,
+      "loss": 0.0018,
+      "step": 4641
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.8816366293229417e-07,
+      "loss": 0.0017,
+      "step": 4642
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.881393083292742e-07,
+      "loss": 0.0018,
+      "step": 4643
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.881149537262542e-07,
+      "loss": 0.0018,
+      "step": 4644
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 3.8809059912323427e-07,
+      "loss": 0.002,
+      "step": 4645
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.880662445202143e-07,
+      "loss": 0.0015,
+      "step": 4646
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.880418899171943e-07,
+      "loss": 0.0014,
+      "step": 4647
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.880175353141744e-07,
+      "loss": 0.0017,
+      "step": 4648
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.879931807111544e-07,
+      "loss": 0.2982,
+      "step": 4649
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.879688261081344e-07,
+      "loss": 0.0019,
+      "step": 4650
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.879444715051144e-07,
+      "loss": 0.0017,
+      "step": 4651
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.879201169020945e-07,
+      "loss": 0.0021,
+      "step": 4652
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.8789576229907453e-07,
+      "loss": 0.0015,
+      "step": 4653
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.878714076960545e-07,
+      "loss": 0.0153,
+      "step": 4654
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.878470530930346e-07,
+      "loss": 0.0018,
+      "step": 4655
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 3.878226984900146e-07,
+      "loss": 0.003,
+      "step": 4656
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.8779834388699463e-07,
+      "loss": 0.0017,
+      "step": 4657
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.8777398928397463e-07,
+      "loss": 0.0013,
+      "step": 4658
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.877496346809547e-07,
+      "loss": 0.0017,
+      "step": 4659
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.877252800779347e-07,
+      "loss": 0.0015,
+      "step": 4660
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.8770092547491473e-07,
+      "loss": 0.0017,
+      "step": 4661
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.876765708718948e-07,
+      "loss": 0.0019,
+      "step": 4662
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.876522162688748e-07,
+      "loss": 0.6333,
+      "step": 4663
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.8762786166585483e-07,
+      "loss": 0.0018,
+      "step": 4664
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.8760350706283483e-07,
+      "loss": 0.0015,
+      "step": 4665
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 3.875791524598149e-07,
+      "loss": 0.0014,
+      "step": 4666
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.875547978567949e-07,
+      "loss": 0.0014,
+      "step": 4667
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.8753044325377493e-07,
+      "loss": 0.0016,
+      "step": 4668
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.87506088650755e-07,
+      "loss": 0.0015,
+      "step": 4669
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.87481734047735e-07,
+      "loss": 0.0015,
+      "step": 4670
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.8745737944471503e-07,
+      "loss": 0.0014,
+      "step": 4671
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.8743302484169503e-07,
+      "loss": 0.0015,
+      "step": 4672
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.874086702386751e-07,
+      "loss": 0.0017,
+      "step": 4673
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.873843156356551e-07,
+      "loss": 0.0017,
+      "step": 4674
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.8735996103263513e-07,
+      "loss": 0.0019,
+      "step": 4675
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 3.8733560642961513e-07,
+      "loss": 0.0015,
+      "step": 4676
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.873112518265952e-07,
+      "loss": 0.0023,
+      "step": 4677
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.8728689722357524e-07,
+      "loss": 0.0014,
+      "step": 4678
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.8726254262055523e-07,
+      "loss": 0.0025,
+      "step": 4679
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.872381880175353e-07,
+      "loss": 0.0013,
+      "step": 4680
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.872138334145153e-07,
+      "loss": 0.0019,
+      "step": 4681
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.8718947881149534e-07,
+      "loss": 0.0018,
+      "step": 4682
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.8716512420847534e-07,
+      "loss": 0.0013,
+      "step": 4683
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.871407696054554e-07,
+      "loss": 0.0016,
+      "step": 4684
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.871164150024355e-07,
+      "loss": 0.0014,
+      "step": 4685
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.870920603994155e-07,
+      "loss": 0.0014,
+      "step": 4686
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 3.8706770579639554e-07,
+      "loss": 0.0017,
+      "step": 4687
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.8704335119337554e-07,
+      "loss": 0.0014,
+      "step": 4688
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.870189965903556e-07,
+      "loss": 0.0015,
+      "step": 4689
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.869946419873356e-07,
+      "loss": 0.0017,
+      "step": 4690
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.8697028738431564e-07,
+      "loss": 0.0012,
+      "step": 4691
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.869459327812957e-07,
+      "loss": 0.0013,
+      "step": 4692
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.869215781782757e-07,
+      "loss": 0.0021,
+      "step": 4693
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.8689722357525574e-07,
+      "loss": 0.0015,
+      "step": 4694
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.8687286897223574e-07,
+      "loss": 0.0182,
+      "step": 4695
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.868485143692158e-07,
+      "loss": 0.0172,
+      "step": 4696
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 3.868241597661958e-07,
+      "loss": 0.0013,
+      "step": 4697
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.8679980516317584e-07,
+      "loss": 0.0017,
+      "step": 4698
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.8677545056015584e-07,
+      "loss": 0.0017,
+      "step": 4699
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.867510959571359e-07,
+      "loss": 0.0021,
+      "step": 4700
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.8672674135411595e-07,
+      "loss": 0.0018,
+      "step": 4701
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.8670238675109595e-07,
+      "loss": 0.388,
+      "step": 4702
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.86678032148076e-07,
+      "loss": 0.0018,
+      "step": 4703
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.86653677545056e-07,
+      "loss": 0.0116,
+      "step": 4704
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.8662932294203605e-07,
+      "loss": 0.0016,
+      "step": 4705
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.8660496833901605e-07,
+      "loss": 0.0017,
+      "step": 4706
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 3.865806137359961e-07,
+      "loss": 0.0013,
+      "step": 4707
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.8655625913297615e-07,
+      "loss": 0.0018,
+      "step": 4708
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.8653190452995615e-07,
+      "loss": 0.0015,
+      "step": 4709
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.865075499269362e-07,
+      "loss": 0.0017,
+      "step": 4710
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.864831953239162e-07,
+      "loss": 0.0016,
+      "step": 4711
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.8645884072089625e-07,
+      "loss": 0.0019,
+      "step": 4712
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.8643448611787625e-07,
+      "loss": 0.0013,
+      "step": 4713
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.864101315148563e-07,
+      "loss": 0.0015,
+      "step": 4714
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.863857769118363e-07,
+      "loss": 0.5966,
+      "step": 4715
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.8636142230881635e-07,
+      "loss": 0.0016,
+      "step": 4716
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 3.863370677057964e-07,
+      "loss": 0.0017,
+      "step": 4717
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.863127131027764e-07,
+      "loss": 0.0017,
+      "step": 4718
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.8628835849975645e-07,
+      "loss": 0.0018,
+      "step": 4719
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.8626400389673645e-07,
+      "loss": 0.0025,
+      "step": 4720
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.862396492937165e-07,
+      "loss": 0.0023,
+      "step": 4721
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.862152946906965e-07,
+      "loss": 0.0016,
+      "step": 4722
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.8619094008767655e-07,
+      "loss": 0.0017,
+      "step": 4723
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.861665854846566e-07,
+      "loss": 0.0015,
+      "step": 4724
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.861422308816366e-07,
+      "loss": 0.0013,
+      "step": 4725
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.8611787627861666e-07,
+      "loss": 0.0016,
+      "step": 4726
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.8609352167559665e-07,
+      "loss": 0.0013,
+      "step": 4727
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 3.860691670725767e-07,
+      "loss": 0.0019,
+      "step": 4728
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.860448124695567e-07,
+      "loss": 0.0016,
+      "step": 4729
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.8602045786653676e-07,
+      "loss": 0.0015,
+      "step": 4730
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.859961032635168e-07,
+      "loss": 0.0011,
+      "step": 4731
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.859717486604968e-07,
+      "loss": 0.0089,
+      "step": 4732
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.8594739405747686e-07,
+      "loss": 0.0022,
+      "step": 4733
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.8592303945445686e-07,
+      "loss": 0.0013,
+      "step": 4734
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.858986848514369e-07,
+      "loss": 0.0019,
+      "step": 4735
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.858743302484169e-07,
+      "loss": 0.0011,
+      "step": 4736
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.8584997564539696e-07,
+      "loss": 0.0014,
+      "step": 4737
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 3.8582562104237696e-07,
+      "loss": 1.2249,
+      "step": 4738
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.85801266439357e-07,
+      "loss": 0.0016,
+      "step": 4739
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.8577691183633706e-07,
+      "loss": 0.0018,
+      "step": 4740
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.8575255723331706e-07,
+      "loss": 0.0017,
+      "step": 4741
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.857282026302971e-07,
+      "loss": 0.0021,
+      "step": 4742
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.857038480272771e-07,
+      "loss": 0.0015,
+      "step": 4743
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.8567949342425716e-07,
+      "loss": 0.0015,
+      "step": 4744
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.8565513882123716e-07,
+      "loss": 0.0016,
+      "step": 4745
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.856307842182172e-07,
+      "loss": 0.0013,
+      "step": 4746
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.8560642961519726e-07,
+      "loss": 0.0016,
+      "step": 4747
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 3.8558207501217726e-07,
+      "loss": 0.0015,
+      "step": 4748
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.855577204091573e-07,
+      "loss": 0.0015,
+      "step": 4749
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.855333658061373e-07,
+      "loss": 0.0021,
+      "step": 4750
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.8550901120311736e-07,
+      "loss": 0.0015,
+      "step": 4751
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.8548465660009736e-07,
+      "loss": 0.0016,
+      "step": 4752
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.854603019970774e-07,
+      "loss": 0.0016,
+      "step": 4753
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.854359473940574e-07,
+      "loss": 0.0013,
+      "step": 4754
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.8541159279103746e-07,
+      "loss": 0.0013,
+      "step": 4755
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.853872381880175e-07,
+      "loss": 0.0024,
+      "step": 4756
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.853628835849975e-07,
+      "loss": 0.0019,
+      "step": 4757
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.8533852898197757e-07,
+      "loss": 0.0021,
+      "step": 4758
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 3.8531417437895757e-07,
+      "loss": 0.0018,
+      "step": 4759
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.852898197759376e-07,
+      "loss": 0.0013,
+      "step": 4760
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.852654651729176e-07,
+      "loss": 0.002,
+      "step": 4761
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.852411105698977e-07,
+      "loss": 0.0014,
+      "step": 4762
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.8521675596687777e-07,
+      "loss": 0.0017,
+      "step": 4763
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.8519240136385777e-07,
+      "loss": 0.0013,
+      "step": 4764
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.851680467608378e-07,
+      "loss": 0.0015,
+      "step": 4765
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.851436921578178e-07,
+      "loss": 0.0014,
+      "step": 4766
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.8511933755479787e-07,
+      "loss": 0.0018,
+      "step": 4767
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.8509498295177787e-07,
+      "loss": 0.0018,
+      "step": 4768
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 3.850706283487579e-07,
+      "loss": 0.0017,
+      "step": 4769
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.85046273745738e-07,
+      "loss": 0.0015,
+      "step": 4770
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.8502191914271797e-07,
+      "loss": 0.0026,
+      "step": 4771
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.84997564539698e-07,
+      "loss": 0.0012,
+      "step": 4772
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.84973209936678e-07,
+      "loss": 0.0015,
+      "step": 4773
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.849488553336581e-07,
+      "loss": 0.0017,
+      "step": 4774
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.849245007306381e-07,
+      "loss": 0.0016,
+      "step": 4775
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.849001461276181e-07,
+      "loss": 0.0024,
+      "step": 4776
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.848757915245981e-07,
+      "loss": 0.0014,
+      "step": 4777
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.848514369215782e-07,
+      "loss": 0.0013,
+      "step": 4778
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 3.8482708231855823e-07,
+      "loss": 0.1811,
+      "step": 4779
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.848027277155382e-07,
+      "loss": 0.0018,
+      "step": 4780
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.847783731125183e-07,
+      "loss": 0.0014,
+      "step": 4781
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.847540185094983e-07,
+      "loss": 0.0022,
+      "step": 4782
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.8472966390647833e-07,
+      "loss": 0.0016,
+      "step": 4783
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.8470530930345833e-07,
+      "loss": 0.0015,
+      "step": 4784
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.846809547004384e-07,
+      "loss": 0.0015,
+      "step": 4785
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.8465660009741843e-07,
+      "loss": 0.0014,
+      "step": 4786
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.8463224549439843e-07,
+      "loss": 0.0018,
+      "step": 4787
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.846078908913785e-07,
+      "loss": 0.0013,
+      "step": 4788
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 3.845835362883585e-07,
+      "loss": 0.0019,
+      "step": 4789
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.8455918168533853e-07,
+      "loss": 0.0016,
+      "step": 4790
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.8453482708231853e-07,
+      "loss": 0.0014,
+      "step": 4791
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.845104724792986e-07,
+      "loss": 0.0013,
+      "step": 4792
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.844861178762786e-07,
+      "loss": 0.0014,
+      "step": 4793
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.8446176327325863e-07,
+      "loss": 0.0014,
+      "step": 4794
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.844374086702387e-07,
+      "loss": 0.0017,
+      "step": 4795
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.844130540672187e-07,
+      "loss": 0.0048,
+      "step": 4796
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.8438869946419873e-07,
+      "loss": 0.0014,
+      "step": 4797
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.8436434486117873e-07,
+      "loss": 0.0019,
+      "step": 4798
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.843399902581588e-07,
+      "loss": 0.0014,
+      "step": 4799
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 3.843156356551388e-07,
+      "loss": 0.0015,
+      "step": 4800
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.8429128105211883e-07,
+      "loss": 0.0048,
+      "step": 4801
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.842669264490989e-07,
+      "loss": 0.0015,
+      "step": 4802
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.842425718460789e-07,
+      "loss": 0.0013,
+      "step": 4803
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.8421821724305894e-07,
+      "loss": 0.0015,
+      "step": 4804
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.8419386264003893e-07,
+      "loss": 0.0014,
+      "step": 4805
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.84169508037019e-07,
+      "loss": 0.0015,
+      "step": 4806
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.84145153433999e-07,
+      "loss": 0.0013,
+      "step": 4807
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.8412079883097904e-07,
+      "loss": 0.0016,
+      "step": 4808
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.8409644422795904e-07,
+      "loss": 0.0016,
+      "step": 4809
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 3.840720896249391e-07,
+      "loss": 0.0015,
+      "step": 4810
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.8404773502191914e-07,
+      "loss": 0.0016,
+      "step": 4811
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.8402338041889914e-07,
+      "loss": 0.0016,
+      "step": 4812
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.839990258158792e-07,
+      "loss": 0.002,
+      "step": 4813
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.839746712128592e-07,
+      "loss": 0.0012,
+      "step": 4814
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.8395031660983924e-07,
+      "loss": 0.0013,
+      "step": 4815
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.8392596200681924e-07,
+      "loss": 0.0014,
+      "step": 4816
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.839016074037993e-07,
+      "loss": 0.0018,
+      "step": 4817
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.8387725280077934e-07,
+      "loss": 0.0014,
+      "step": 4818
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.8385289819775934e-07,
+      "loss": 0.001,
+      "step": 4819
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 3.838285435947394e-07,
+      "loss": 0.0023,
+      "step": 4820
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.838041889917194e-07,
+      "loss": 0.0016,
+      "step": 4821
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.8377983438869944e-07,
+      "loss": 0.0013,
+      "step": 4822
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.8375547978567944e-07,
+      "loss": 0.0012,
+      "step": 4823
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.837311251826595e-07,
+      "loss": 0.0016,
+      "step": 4824
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.837067705796395e-07,
+      "loss": 0.0016,
+      "step": 4825
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.8368241597661954e-07,
+      "loss": 0.0016,
+      "step": 4826
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.836580613735996e-07,
+      "loss": 0.0012,
+      "step": 4827
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.836337067705796e-07,
+      "loss": 0.0027,
+      "step": 4828
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.8360935216755964e-07,
+      "loss": 0.0016,
+      "step": 4829
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.8358499756453964e-07,
+      "loss": 0.0012,
+      "step": 4830
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 3.835606429615197e-07,
+      "loss": 0.0016,
+      "step": 4831
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.835362883584997e-07,
+      "loss": 0.0013,
+      "step": 4832
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.8351193375547975e-07,
+      "loss": 0.0015,
+      "step": 4833
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.834875791524598e-07,
+      "loss": 0.0015,
+      "step": 4834
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.834632245494398e-07,
+      "loss": 0.0014,
+      "step": 4835
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.8343886994641985e-07,
+      "loss": 0.0015,
+      "step": 4836
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.8341451534339985e-07,
+      "loss": 0.9191,
+      "step": 4837
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.8339016074037995e-07,
+      "loss": 0.0014,
+      "step": 4838
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.8336580613735995e-07,
+      "loss": 0.0012,
+      "step": 4839
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.8334145153434e-07,
+      "loss": 0.0012,
+      "step": 4840
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 3.8331709693132005e-07,
+      "loss": 0.0013,
+      "step": 4841
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.8329274232830005e-07,
+      "loss": 0.0015,
+      "step": 4842
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.832683877252801e-07,
+      "loss": 0.0016,
+      "step": 4843
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.832440331222601e-07,
+      "loss": 0.0014,
+      "step": 4844
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.8321967851924015e-07,
+      "loss": 0.0017,
+      "step": 4845
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.8319532391622015e-07,
+      "loss": 0.0014,
+      "step": 4846
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.831709693132002e-07,
+      "loss": 0.0014,
+      "step": 4847
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.831466147101802e-07,
+      "loss": 0.002,
+      "step": 4848
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.8312226010716025e-07,
+      "loss": 1.1329,
+      "step": 4849
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.830979055041403e-07,
+      "loss": 0.0013,
+      "step": 4850
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 3.830735509011203e-07,
+      "loss": 0.0012,
+      "step": 4851
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.8304919629810036e-07,
+      "loss": 0.0013,
+      "step": 4852
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.8302484169508035e-07,
+      "loss": 0.0016,
+      "step": 4853
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.830004870920604e-07,
+      "loss": 0.0016,
+      "step": 4854
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.829761324890404e-07,
+      "loss": 0.0012,
+      "step": 4855
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.8295177788602046e-07,
+      "loss": 0.0014,
+      "step": 4856
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.829274232830005e-07,
+      "loss": 0.0015,
+      "step": 4857
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.829030686799805e-07,
+      "loss": 0.0011,
+      "step": 4858
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.8287871407696056e-07,
+      "loss": 0.0017,
+      "step": 4859
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.8285435947394056e-07,
+      "loss": 0.0014,
+      "step": 4860
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.828300048709206e-07,
+      "loss": 0.0014,
+      "step": 4861
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 3.828056502679006e-07,
+      "loss": 0.0014,
+      "step": 4862
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.8278129566488066e-07,
+      "loss": 0.001,
+      "step": 4863
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.8275694106186066e-07,
+      "loss": 0.0015,
+      "step": 4864
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.827325864588407e-07,
+      "loss": 0.0014,
+      "step": 4865
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.8270823185582076e-07,
+      "loss": 0.0016,
+      "step": 4866
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.8268387725280076e-07,
+      "loss": 0.0016,
+      "step": 4867
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.826595226497808e-07,
+      "loss": 0.0013,
+      "step": 4868
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.826351680467608e-07,
+      "loss": 0.0014,
+      "step": 4869
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.8261081344374086e-07,
+      "loss": 0.0015,
+      "step": 4870
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.8258645884072086e-07,
+      "loss": 0.0013,
+      "step": 4871
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 3.825621042377009e-07,
+      "loss": 0.0016,
+      "step": 4872
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.8253774963468096e-07,
+      "loss": 0.0013,
+      "step": 4873
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.8251339503166096e-07,
+      "loss": 0.002,
+      "step": 4874
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.82489040428641e-07,
+      "loss": 0.0021,
+      "step": 4875
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.82464685825621e-07,
+      "loss": 0.0053,
+      "step": 4876
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.8244033122260106e-07,
+      "loss": 0.0015,
+      "step": 4877
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.8241597661958106e-07,
+      "loss": 0.0017,
+      "step": 4878
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.823916220165611e-07,
+      "loss": 0.001,
+      "step": 4879
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.823672674135411e-07,
+      "loss": 0.0014,
+      "step": 4880
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.8234291281052116e-07,
+      "loss": 0.0016,
+      "step": 4881
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.823185582075012e-07,
+      "loss": 0.0015,
+      "step": 4882
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.822942036044812e-07,
+      "loss": 0.0018,
+      "step": 4883
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.8226984900146127e-07,
+      "loss": 0.0015,
+      "step": 4884
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.8224549439844127e-07,
+      "loss": 0.0018,
+      "step": 4885
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.822211397954213e-07,
+      "loss": 0.0019,
+      "step": 4886
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.821967851924013e-07,
+      "loss": 0.0016,
+      "step": 4887
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.8217243058938137e-07,
+      "loss": 0.0012,
+      "step": 4888
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.821480759863614e-07,
+      "loss": 0.0012,
+      "step": 4889
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.821237213833414e-07,
+      "loss": 0.0013,
+      "step": 4890
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.8209936678032147e-07,
+      "loss": 0.0013,
+      "step": 4891
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.8207501217730147e-07,
+      "loss": 0.0013,
+      "step": 4892
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.820506575742815e-07,
+      "loss": 0.0016,
+      "step": 4893
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.820263029712615e-07,
+      "loss": 0.0017,
+      "step": 4894
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.8200194836824157e-07,
+      "loss": 0.0012,
+      "step": 4895
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.8197759376522157e-07,
+      "loss": 0.0012,
+      "step": 4896
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.819532391622016e-07,
+      "loss": 0.0016,
+      "step": 4897
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.8192888455918167e-07,
+      "loss": 0.0011,
+      "step": 4898
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.8190452995616167e-07,
+      "loss": 0.0016,
+      "step": 4899
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.818801753531417e-07,
+      "loss": 0.0018,
+      "step": 4900
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.818558207501217e-07,
+      "loss": 0.0014,
+      "step": 4901
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.8183146614710177e-07,
+      "loss": 0.0014,
+      "step": 4902
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.8180711154408177e-07,
+      "loss": 0.0057,
+      "step": 4903
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.817827569410618e-07,
+      "loss": 0.0016,
+      "step": 4904
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.817584023380419e-07,
+      "loss": 0.0014,
+      "step": 4905
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.8173404773502187e-07,
+      "loss": 0.0012,
+      "step": 4906
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.817096931320019e-07,
+      "loss": 0.002,
+      "step": 4907
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.816853385289819e-07,
+      "loss": 0.0013,
+      "step": 4908
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.81660983925962e-07,
+      "loss": 0.0016,
+      "step": 4909
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.81636629322942e-07,
+      "loss": 0.0012,
+      "step": 4910
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.81612274719922e-07,
+      "loss": 0.0013,
+      "step": 4911
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.815879201169021e-07,
+      "loss": 0.0013,
+      "step": 4912
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.815635655138821e-07,
+      "loss": 0.0014,
+      "step": 4913
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.8153921091086213e-07,
+      "loss": 0.002,
+      "step": 4914
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.815148563078422e-07,
+      "loss": 0.3646,
+      "step": 4915
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.8149050170482223e-07,
+      "loss": 0.0015,
+      "step": 4916
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.8146614710180223e-07,
+      "loss": 0.0015,
+      "step": 4917
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.814417924987823e-07,
+      "loss": 0.0013,
+      "step": 4918
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.814174378957623e-07,
+      "loss": 0.0014,
+      "step": 4919
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.8139308329274233e-07,
+      "loss": 0.0013,
+      "step": 4920
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.813687286897224e-07,
+      "loss": 0.0016,
+      "step": 4921
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.813443740867024e-07,
+      "loss": 0.0015,
+      "step": 4922
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.8132001948368243e-07,
+      "loss": 0.0016,
+      "step": 4923
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.8129566488066243e-07,
+      "loss": 0.0011,
+      "step": 4924
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.812713102776425e-07,
+      "loss": 0.0013,
+      "step": 4925
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.812469556746225e-07,
+      "loss": 0.0012,
+      "step": 4926
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.8122260107160253e-07,
+      "loss": 1.1113,
+      "step": 4927
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.811982464685826e-07,
+      "loss": 0.0013,
+      "step": 4928
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.811738918655626e-07,
+      "loss": 0.0017,
+      "step": 4929
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.8114953726254264e-07,
+      "loss": 0.0015,
+      "step": 4930
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.8112518265952263e-07,
+      "loss": 0.0014,
+      "step": 4931
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.811008280565027e-07,
+      "loss": 0.0356,
+      "step": 4932
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.810764734534827e-07,
+      "loss": 0.0015,
+      "step": 4933
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.8105211885046274e-07,
+      "loss": 0.0014,
+      "step": 4934
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.810277642474428e-07,
+      "loss": 0.0022,
+      "step": 4935
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.810034096444228e-07,
+      "loss": 0.0016,
+      "step": 4936
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.8097905504140284e-07,
+      "loss": 0.0012,
+      "step": 4937
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.8095470043838284e-07,
+      "loss": 0.0014,
+      "step": 4938
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.809303458353629e-07,
+      "loss": 0.0017,
+      "step": 4939
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.809059912323429e-07,
+      "loss": 0.0017,
+      "step": 4940
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.8088163662932294e-07,
+      "loss": 0.001,
+      "step": 4941
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.8085728202630294e-07,
+      "loss": 0.0016,
+      "step": 4942
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.80832927423283e-07,
+      "loss": 0.0013,
+      "step": 4943
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.8080857282026304e-07,
+      "loss": 0.0013,
+      "step": 4944
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.8078421821724304e-07,
+      "loss": 0.0016,
+      "step": 4945
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.807598636142231e-07,
+      "loss": 0.0013,
+      "step": 4946
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.807355090112031e-07,
+      "loss": 0.0015,
+      "step": 4947
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.8071115440818314e-07,
+      "loss": 0.0016,
+      "step": 4948
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.8068679980516314e-07,
+      "loss": 0.0013,
+      "step": 4949
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.806624452021432e-07,
+      "loss": 0.0013,
+      "step": 4950
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.8063809059912324e-07,
+      "loss": 0.0014,
+      "step": 4951
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.8061373599610324e-07,
+      "loss": 0.0014,
+      "step": 4952
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.805893813930833e-07,
+      "loss": 0.0018,
+      "step": 4953
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 3.805650267900633e-07,
+      "loss": 0.0016,
+      "step": 4954
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.8054067218704334e-07,
+      "loss": 0.0012,
+      "step": 4955
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.8051631758402334e-07,
+      "loss": 0.0016,
+      "step": 4956
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.804919629810034e-07,
+      "loss": 0.0012,
+      "step": 4957
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.804676083779834e-07,
+      "loss": 0.0018,
+      "step": 4958
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.8044325377496345e-07,
+      "loss": 0.0013,
+      "step": 4959
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.804188991719435e-07,
+      "loss": 0.0015,
+      "step": 4960
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.803945445689235e-07,
+      "loss": 0.0011,
+      "step": 4961
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.8037018996590355e-07,
+      "loss": 0.0017,
+      "step": 4962
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.8034583536288355e-07,
+      "loss": 0.0013,
+      "step": 4963
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 3.803214807598636e-07,
+      "loss": 0.0012,
+      "step": 4964
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.802971261568436e-07,
+      "loss": 0.0014,
+      "step": 4965
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.8027277155382365e-07,
+      "loss": 0.0012,
+      "step": 4966
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.802484169508037e-07,
+      "loss": 0.0012,
+      "step": 4967
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.802240623477837e-07,
+      "loss": 0.0012,
+      "step": 4968
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.8019970774476375e-07,
+      "loss": 0.0014,
+      "step": 4969
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.8017535314174375e-07,
+      "loss": 0.2431,
+      "step": 4970
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.801509985387238e-07,
+      "loss": 0.0014,
+      "step": 4971
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.801266439357038e-07,
+      "loss": 0.0014,
+      "step": 4972
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.8010228933268385e-07,
+      "loss": 0.0012,
+      "step": 4973
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.8007793472966385e-07,
+      "loss": 0.0017,
+      "step": 4974
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 3.800535801266439e-07,
+      "loss": 0.0016,
+      "step": 4975
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.8002922552362395e-07,
+      "loss": 0.0012,
+      "step": 4976
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.8000487092060395e-07,
+      "loss": 0.0022,
+      "step": 4977
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.79980516317584e-07,
+      "loss": 0.0016,
+      "step": 4978
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.79956161714564e-07,
+      "loss": 0.0016,
+      "step": 4979
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.7993180711154405e-07,
+      "loss": 0.0014,
+      "step": 4980
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.7990745250852405e-07,
+      "loss": 0.0012,
+      "step": 4981
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.798830979055041e-07,
+      "loss": 0.0012,
+      "step": 4982
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.7985874330248416e-07,
+      "loss": 0.0015,
+      "step": 4983
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.7983438869946415e-07,
+      "loss": 0.0017,
+      "step": 4984
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 3.798100340964442e-07,
+      "loss": 1.7118,
+      "step": 4985
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.797856794934242e-07,
+      "loss": 0.9929,
+      "step": 4986
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.7976132489040426e-07,
+      "loss": 0.0017,
+      "step": 4987
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.7973697028738425e-07,
+      "loss": 0.0013,
+      "step": 4988
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.797126156843643e-07,
+      "loss": 0.0015,
+      "step": 4989
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.796882610813443e-07,
+      "loss": 0.0014,
+      "step": 4990
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.796639064783244e-07,
+      "loss": 0.001,
+      "step": 4991
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.7963955187530446e-07,
+      "loss": 0.0017,
+      "step": 4992
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.7961519727228446e-07,
+      "loss": 0.0015,
+      "step": 4993
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.795908426692645e-07,
+      "loss": 0.0015,
+      "step": 4994
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 3.795664880662445e-07,
+      "loss": 0.0012,
+      "step": 4995
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.7954213346322456e-07,
+      "loss": 0.0011,
+      "step": 4996
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.7951777886020456e-07,
+      "loss": 0.0016,
+      "step": 4997
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.794934242571846e-07,
+      "loss": 0.0016,
+      "step": 4998
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.7946906965416466e-07,
+      "loss": 0.0023,
+      "step": 4999
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.7944471505114466e-07,
+      "loss": 0.002,
+      "step": 5000
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.794203604481247e-07,
+      "loss": 0.0012,
+      "step": 5001
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.793960058451047e-07,
+      "loss": 0.9489,
+      "step": 5002
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.7937165124208476e-07,
+      "loss": 0.0014,
+      "step": 5003
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.7934729663906476e-07,
+      "loss": 0.0015,
+      "step": 5004
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.793229420360448e-07,
+      "loss": 0.0017,
+      "step": 5005
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 3.7929858743302487e-07,
+      "loss": 0.0123,
+      "step": 5006
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.7927423283000487e-07,
+      "loss": 0.0013,
+      "step": 5007
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.792498782269849e-07,
+      "loss": 0.0013,
+      "step": 5008
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.792255236239649e-07,
+      "loss": 0.0013,
+      "step": 5009
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.7920116902094497e-07,
+      "loss": 0.0014,
+      "step": 5010
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.7917681441792497e-07,
+      "loss": 0.0015,
+      "step": 5011
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.79152459814905e-07,
+      "loss": 0.001,
+      "step": 5012
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.79128105211885e-07,
+      "loss": 0.0014,
+      "step": 5013
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.7910375060886507e-07,
+      "loss": 0.0015,
+      "step": 5014
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.790793960058451e-07,
+      "loss": 0.0012,
+      "step": 5015
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 3.790550414028251e-07,
+      "loss": 0.0011,
+      "step": 5016
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.7903068679980517e-07,
+      "loss": 0.0015,
+      "step": 5017
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.7900633219678517e-07,
+      "loss": 0.001,
+      "step": 5018
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.789819775937652e-07,
+      "loss": 0.0012,
+      "step": 5019
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.789576229907452e-07,
+      "loss": 0.0014,
+      "step": 5020
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.7893326838772527e-07,
+      "loss": 0.0013,
+      "step": 5021
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.789089137847053e-07,
+      "loss": 0.0015,
+      "step": 5022
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.788845591816853e-07,
+      "loss": 0.0011,
+      "step": 5023
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.7886020457866537e-07,
+      "loss": 0.0016,
+      "step": 5024
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.7883584997564537e-07,
+      "loss": 0.001,
+      "step": 5025
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 3.788114953726254e-07,
+      "loss": 0.0023,
+      "step": 5026
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.787871407696054e-07,
+      "loss": 0.0044,
+      "step": 5027
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.7876278616658547e-07,
+      "loss": 0.0015,
+      "step": 5028
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.7873843156356547e-07,
+      "loss": 0.0013,
+      "step": 5029
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.787140769605455e-07,
+      "loss": 0.0016,
+      "step": 5030
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.786897223575256e-07,
+      "loss": 0.0016,
+      "step": 5031
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.786653677545056e-07,
+      "loss": 0.0013,
+      "step": 5032
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.786410131514856e-07,
+      "loss": 0.0017,
+      "step": 5033
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.786166585484656e-07,
+      "loss": 0.0012,
+      "step": 5034
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.785923039454457e-07,
+      "loss": 0.0017,
+      "step": 5035
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 3.785679493424257e-07,
+      "loss": 0.0042,
+      "step": 5036
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.785435947394057e-07,
+      "loss": 0.0015,
+      "step": 5037
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.785192401363858e-07,
+      "loss": 0.0015,
+      "step": 5038
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.784948855333658e-07,
+      "loss": 0.0016,
+      "step": 5039
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.7847053093034583e-07,
+      "loss": 0.0014,
+      "step": 5040
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.7844617632732583e-07,
+      "loss": 0.001,
+      "step": 5041
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.784218217243059e-07,
+      "loss": 0.0013,
+      "step": 5042
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.783974671212859e-07,
+      "loss": 0.0013,
+      "step": 5043
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.7837311251826593e-07,
+      "loss": 0.0015,
+      "step": 5044
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.7834875791524593e-07,
+      "loss": 0.0011,
+      "step": 5045
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.78324403312226e-07,
+      "loss": 0.0234,
+      "step": 5046
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 3.7830004870920603e-07,
+      "loss": 0.0011,
+      "step": 5047
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.7827569410618603e-07,
+      "loss": 0.0013,
+      "step": 5048
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.782513395031661e-07,
+      "loss": 0.0018,
+      "step": 5049
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.782269849001461e-07,
+      "loss": 0.0022,
+      "step": 5050
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.7820263029712613e-07,
+      "loss": 0.0014,
+      "step": 5051
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.7817827569410613e-07,
+      "loss": 0.1587,
+      "step": 5052
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.781539210910862e-07,
+      "loss": 0.0019,
+      "step": 5053
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.7812956648806623e-07,
+      "loss": 0.0014,
+      "step": 5054
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.7810521188504623e-07,
+      "loss": 0.0014,
+      "step": 5055
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.780808572820263e-07,
+      "loss": 0.0013,
+      "step": 5056
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 3.780565026790063e-07,
+      "loss": 0.0013,
+      "step": 5057
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.7803214807598633e-07,
+      "loss": 0.0014,
+      "step": 5058
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.7800779347296633e-07,
+      "loss": 0.0014,
+      "step": 5059
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.779834388699464e-07,
+      "loss": 0.0014,
+      "step": 5060
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.779590842669264e-07,
+      "loss": 0.1127,
+      "step": 5061
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.7793472966390643e-07,
+      "loss": 0.0013,
+      "step": 5062
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.779103750608865e-07,
+      "loss": 0.0017,
+      "step": 5063
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.778860204578665e-07,
+      "loss": 0.0013,
+      "step": 5064
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.7786166585484654e-07,
+      "loss": 0.0014,
+      "step": 5065
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.7783731125182654e-07,
+      "loss": 0.0014,
+      "step": 5066
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 3.778129566488066e-07,
+      "loss": 0.0012,
+      "step": 5067
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.7778860204578664e-07,
+      "loss": 0.0014,
+      "step": 5068
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.777642474427667e-07,
+      "loss": 0.0013,
+      "step": 5069
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.7773989283974674e-07,
+      "loss": 0.0014,
+      "step": 5070
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.7771553823672674e-07,
+      "loss": 0.0013,
+      "step": 5071
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.776911836337068e-07,
+      "loss": 0.4655,
+      "step": 5072
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.776668290306868e-07,
+      "loss": 0.0013,
+      "step": 5073
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.7764247442766684e-07,
+      "loss": 0.0017,
+      "step": 5074
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.7761811982464684e-07,
+      "loss": 0.0011,
+      "step": 5075
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.775937652216269e-07,
+      "loss": 0.0013,
+      "step": 5076
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.7756941061860694e-07,
+      "loss": 0.0016,
+      "step": 5077
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 3.7754505601558694e-07,
+      "loss": 0.0016,
+      "step": 5078
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.77520701412567e-07,
+      "loss": 0.0012,
+      "step": 5079
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.77496346809547e-07,
+      "loss": 0.0012,
+      "step": 5080
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.7747199220652704e-07,
+      "loss": 0.0015,
+      "step": 5081
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.7744763760350704e-07,
+      "loss": 0.0019,
+      "step": 5082
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.774232830004871e-07,
+      "loss": 0.0011,
+      "step": 5083
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.773989283974671e-07,
+      "loss": 0.0015,
+      "step": 5084
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.7737457379444715e-07,
+      "loss": 0.0014,
+      "step": 5085
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.773502191914272e-07,
+      "loss": 0.0032,
+      "step": 5086
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.773258645884072e-07,
+      "loss": 0.0016,
+      "step": 5087
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 3.7730150998538725e-07,
+      "loss": 0.0012,
+      "step": 5088
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.7727715538236725e-07,
+      "loss": 0.0014,
+      "step": 5089
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.772528007793473e-07,
+      "loss": 0.0015,
+      "step": 5090
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.772284461763273e-07,
+      "loss": 0.0011,
+      "step": 5091
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.7720409157330735e-07,
+      "loss": 0.0015,
+      "step": 5092
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.771797369702874e-07,
+      "loss": 0.0016,
+      "step": 5093
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.771553823672674e-07,
+      "loss": 0.0014,
+      "step": 5094
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.7713102776424745e-07,
+      "loss": 0.1352,
+      "step": 5095
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.7710667316122745e-07,
+      "loss": 0.0013,
+      "step": 5096
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.770823185582075e-07,
+      "loss": 0.0014,
+      "step": 5097
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 3.770579639551875e-07,
+      "loss": 0.0026,
+      "step": 5098
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.7703360935216755e-07,
+      "loss": 0.0017,
+      "step": 5099
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.770092547491476e-07,
+      "loss": 0.0017,
+      "step": 5100
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.769849001461276e-07,
+      "loss": 0.0014,
+      "step": 5101
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.7696054554310765e-07,
+      "loss": 0.0012,
+      "step": 5102
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.7693619094008765e-07,
+      "loss": 0.0011,
+      "step": 5103
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.769118363370677e-07,
+      "loss": 0.0011,
+      "step": 5104
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.768874817340477e-07,
+      "loss": 0.0013,
+      "step": 5105
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.7686312713102775e-07,
+      "loss": 0.0012,
+      "step": 5106
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.7683877252800775e-07,
+      "loss": 0.0012,
+      "step": 5107
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 3.768144179249878e-07,
+      "loss": 0.0013,
+      "step": 5108
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.7679006332196786e-07,
+      "loss": 0.0013,
+      "step": 5109
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.7676570871894785e-07,
+      "loss": 0.0014,
+      "step": 5110
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.767413541159279e-07,
+      "loss": 0.0014,
+      "step": 5111
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.767169995129079e-07,
+      "loss": 0.0015,
+      "step": 5112
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.7669264490988796e-07,
+      "loss": 0.0014,
+      "step": 5113
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.7666829030686795e-07,
+      "loss": 0.0018,
+      "step": 5114
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.76643935703848e-07,
+      "loss": 0.0015,
+      "step": 5115
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.7661958110082806e-07,
+      "loss": 0.0015,
+      "step": 5116
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.7659522649780806e-07,
+      "loss": 0.0013,
+      "step": 5117
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.765708718947881e-07,
+      "loss": 0.0017,
+      "step": 5118
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 3.765465172917681e-07,
+      "loss": 0.0015,
+      "step": 5119
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.7652216268874816e-07,
+      "loss": 0.0014,
+      "step": 5120
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.7649780808572816e-07,
+      "loss": 0.0017,
+      "step": 5121
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.764734534827082e-07,
+      "loss": 0.0013,
+      "step": 5122
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.764490988796882e-07,
+      "loss": 0.0014,
+      "step": 5123
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.7642474427666826e-07,
+      "loss": 0.0014,
+      "step": 5124
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.764003896736483e-07,
+      "loss": 0.0013,
+      "step": 5125
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.763760350706283e-07,
+      "loss": 0.0014,
+      "step": 5126
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.7635168046760836e-07,
+      "loss": 0.0019,
+      "step": 5127
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.7632732586458836e-07,
+      "loss": 0.0014,
+      "step": 5128
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.763029712615684e-07,
+      "loss": 0.0011,
+      "step": 5129
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.762786166585484e-07,
+      "loss": 0.001,
+      "step": 5130
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.7625426205552846e-07,
+      "loss": 0.0012,
+      "step": 5131
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.762299074525085e-07,
+      "loss": 0.0015,
+      "step": 5132
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.762055528494885e-07,
+      "loss": 0.0031,
+      "step": 5133
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.7618119824646856e-07,
+      "loss": 0.0017,
+      "step": 5134
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.7615684364344856e-07,
+      "loss": 0.0014,
+      "step": 5135
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.761324890404286e-07,
+      "loss": 0.0013,
+      "step": 5136
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.761081344374086e-07,
+      "loss": 0.0011,
+      "step": 5137
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.7608377983438866e-07,
+      "loss": 0.0009,
+      "step": 5138
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 3.7605942523136866e-07,
+      "loss": 0.0012,
+      "step": 5139
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.760350706283487e-07,
+      "loss": 0.6754,
+      "step": 5140
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.7601071602532877e-07,
+      "loss": 0.0013,
+      "step": 5141
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.7598636142230877e-07,
+      "loss": 0.0012,
+      "step": 5142
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.759620068192888e-07,
+      "loss": 0.0016,
+      "step": 5143
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.759376522162688e-07,
+      "loss": 0.0014,
+      "step": 5144
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.759132976132489e-07,
+      "loss": 0.001,
+      "step": 5145
+    },
+    {
+      "epoch": 5.0,
+      "eval_accuracy": 0.9980554205153136,
+      "eval_loss": 0.0173475444316864,
+      "eval_runtime": 2.7797,
+      "eval_samples_per_second": 739.996,
+      "eval_steps_per_second": 11.872,
+      "step": 5145
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.758889430102289e-07,
+      "loss": 0.0032,
+      "step": 5146
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.7586458840720897e-07,
+      "loss": 0.0009,
+      "step": 5147
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.75840233804189e-07,
+      "loss": 0.0012,
+      "step": 5148
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.75815879201169e-07,
+      "loss": 0.0012,
+      "step": 5149
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 3.7579152459814907e-07,
+      "loss": 0.0013,
+      "step": 5150
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.7576716999512907e-07,
+      "loss": 0.0013,
+      "step": 5151
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.757428153921091e-07,
+      "loss": 0.0019,
+      "step": 5152
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.757184607890891e-07,
+      "loss": 0.0014,
+      "step": 5153
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.7569410618606917e-07,
+      "loss": 0.0011,
+      "step": 5154
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.756697515830492e-07,
+      "loss": 0.0015,
+      "step": 5155
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.756453969800292e-07,
+      "loss": 0.0014,
+      "step": 5156
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.756210423770093e-07,
+      "loss": 0.0015,
+      "step": 5157
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.755966877739893e-07,
+      "loss": 0.0016,
+      "step": 5158
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.755723331709693e-07,
+      "loss": 0.0014,
+      "step": 5159
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 3.755479785679493e-07,
+      "loss": 0.0011,
+      "step": 5160
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.755236239649294e-07,
+      "loss": 0.0009,
+      "step": 5161
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.754992693619094e-07,
+      "loss": 0.0011,
+      "step": 5162
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.754749147588894e-07,
+      "loss": 0.0011,
+      "step": 5163
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.754505601558695e-07,
+      "loss": 0.0013,
+      "step": 5164
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.754262055528495e-07,
+      "loss": 0.001,
+      "step": 5165
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.7540185094982953e-07,
+      "loss": 0.0011,
+      "step": 5166
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.7537749634680953e-07,
+      "loss": 0.0017,
+      "step": 5167
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.753531417437896e-07,
+      "loss": 0.0016,
+      "step": 5168
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.753287871407696e-07,
+      "loss": 0.0011,
+      "step": 5169
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 3.7530443253774963e-07,
+      "loss": 0.0011,
+      "step": 5170
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.752800779347297e-07,
+      "loss": 0.0012,
+      "step": 5171
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.752557233317097e-07,
+      "loss": 0.0013,
+      "step": 5172
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.7523136872868973e-07,
+      "loss": 0.0009,
+      "step": 5173
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.7520701412566973e-07,
+      "loss": 0.0012,
+      "step": 5174
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.751826595226498e-07,
+      "loss": 0.0012,
+      "step": 5175
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.751583049196298e-07,
+      "loss": 0.0011,
+      "step": 5176
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.7513395031660983e-07,
+      "loss": 0.0017,
+      "step": 5177
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.7510959571358983e-07,
+      "loss": 0.0013,
+      "step": 5178
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.750852411105699e-07,
+      "loss": 0.0012,
+      "step": 5179
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.7506088650754993e-07,
+      "loss": 0.001,
+      "step": 5180
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 3.7503653190452993e-07,
+      "loss": 0.0013,
+      "step": 5181
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.7501217730151e-07,
+      "loss": 0.0011,
+      "step": 5182
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.7498782269849e-07,
+      "loss": 0.001,
+      "step": 5183
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.7496346809547003e-07,
+      "loss": 0.0015,
+      "step": 5184
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.7493911349245003e-07,
+      "loss": 0.0011,
+      "step": 5185
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.749147588894301e-07,
+      "loss": 0.0013,
+      "step": 5186
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.7489040428641014e-07,
+      "loss": 0.0014,
+      "step": 5187
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.7486604968339013e-07,
+      "loss": 0.0013,
+      "step": 5188
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.748416950803702e-07,
+      "loss": 0.0016,
+      "step": 5189
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.748173404773502e-07,
+      "loss": 0.0013,
+      "step": 5190
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 3.7479298587433024e-07,
+      "loss": 0.0013,
+      "step": 5191
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.7476863127131024e-07,
+      "loss": 0.0012,
+      "step": 5192
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.747442766682903e-07,
+      "loss": 0.001,
+      "step": 5193
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.747199220652703e-07,
+      "loss": 0.0014,
+      "step": 5194
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.7469556746225034e-07,
+      "loss": 1.305,
+      "step": 5195
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.746712128592304e-07,
+      "loss": 0.0012,
+      "step": 5196
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.746468582562104e-07,
+      "loss": 0.0015,
+      "step": 5197
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.7462250365319044e-07,
+      "loss": 0.0011,
+      "step": 5198
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.7459814905017044e-07,
+      "loss": 0.0014,
+      "step": 5199
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.745737944471505e-07,
+      "loss": 0.001,
+      "step": 5200
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 3.745494398441305e-07,
+      "loss": 0.0012,
+      "step": 5201
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.7452508524111054e-07,
+      "loss": 0.0014,
+      "step": 5202
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.745007306380906e-07,
+      "loss": 0.0012,
+      "step": 5203
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.744763760350706e-07,
+      "loss": 0.0014,
+      "step": 5204
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.7445202143205064e-07,
+      "loss": 0.001,
+      "step": 5205
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.7442766682903064e-07,
+      "loss": 0.0013,
+      "step": 5206
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.744033122260107e-07,
+      "loss": 0.0015,
+      "step": 5207
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.743789576229907e-07,
+      "loss": 0.0024,
+      "step": 5208
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.7435460301997074e-07,
+      "loss": 0.0013,
+      "step": 5209
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.7433024841695074e-07,
+      "loss": 0.0018,
+      "step": 5210
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 3.743058938139308e-07,
+      "loss": 0.0015,
+      "step": 5211
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.7428153921091084e-07,
+      "loss": 0.0021,
+      "step": 5212
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.7425718460789084e-07,
+      "loss": 0.0009,
+      "step": 5213
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.742328300048709e-07,
+      "loss": 0.0015,
+      "step": 5214
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.742084754018509e-07,
+      "loss": 0.0015,
+      "step": 5215
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.7418412079883095e-07,
+      "loss": 0.0009,
+      "step": 5216
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.7415976619581094e-07,
+      "loss": 0.0011,
+      "step": 5217
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.74135411592791e-07,
+      "loss": 0.0015,
+      "step": 5218
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.7411105698977105e-07,
+      "loss": 0.0015,
+      "step": 5219
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.7408670238675105e-07,
+      "loss": 0.0011,
+      "step": 5220
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.7406234778373115e-07,
+      "loss": 0.0011,
+      "step": 5221
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 3.7403799318071115e-07,
+      "loss": 0.0014,
+      "step": 5222
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.740136385776912e-07,
+      "loss": 0.0014,
+      "step": 5223
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.739892839746712e-07,
+      "loss": 0.0013,
+      "step": 5224
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.7396492937165125e-07,
+      "loss": 0.0015,
+      "step": 5225
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.739405747686313e-07,
+      "loss": 0.0017,
+      "step": 5226
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.739162201656113e-07,
+      "loss": 0.0012,
+      "step": 5227
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.7389186556259135e-07,
+      "loss": 0.0013,
+      "step": 5228
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.7386751095957135e-07,
+      "loss": 0.0014,
+      "step": 5229
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.738431563565514e-07,
+      "loss": 0.0016,
+      "step": 5230
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.738188017535314e-07,
+      "loss": 0.0011,
+      "step": 5231
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 3.7379444715051145e-07,
+      "loss": 0.0011,
+      "step": 5232
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.7377009254749145e-07,
+      "loss": 0.0013,
+      "step": 5233
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.737457379444715e-07,
+      "loss": 0.0015,
+      "step": 5234
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.7372138334145156e-07,
+      "loss": 0.0009,
+      "step": 5235
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.7369702873843155e-07,
+      "loss": 0.0012,
+      "step": 5236
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.736726741354116e-07,
+      "loss": 0.0016,
+      "step": 5237
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.736483195323916e-07,
+      "loss": 0.0011,
+      "step": 5238
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.7362396492937166e-07,
+      "loss": 0.0015,
+      "step": 5239
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.7359961032635166e-07,
+      "loss": 0.0014,
+      "step": 5240
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.735752557233317e-07,
+      "loss": 0.0016,
+      "step": 5241
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 3.7355090112031176e-07,
+      "loss": 0.0016,
+      "step": 5242
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.7352654651729176e-07,
+      "loss": 0.0012,
+      "step": 5243
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.735021919142718e-07,
+      "loss": 0.0012,
+      "step": 5244
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.734778373112518e-07,
+      "loss": 0.0014,
+      "step": 5245
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.7345348270823186e-07,
+      "loss": 0.001,
+      "step": 5246
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.7342912810521186e-07,
+      "loss": 0.0013,
+      "step": 5247
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.734047735021919e-07,
+      "loss": 0.0014,
+      "step": 5248
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.733804188991719e-07,
+      "loss": 0.001,
+      "step": 5249
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.7335606429615196e-07,
+      "loss": 0.0012,
+      "step": 5250
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.73331709693132e-07,
+      "loss": 0.0018,
+      "step": 5251
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.73307355090112e-07,
+      "loss": 0.0016,
+      "step": 5252
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 3.7328300048709206e-07,
+      "loss": 0.0011,
+      "step": 5253
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.7325864588407206e-07,
+      "loss": 0.0014,
+      "step": 5254
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.732342912810521e-07,
+      "loss": 0.0011,
+      "step": 5255
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.732099366780321e-07,
+      "loss": 0.0016,
+      "step": 5256
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.7318558207501216e-07,
+      "loss": 0.0057,
+      "step": 5257
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.731612274719922e-07,
+      "loss": 0.0013,
+      "step": 5258
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.731368728689722e-07,
+      "loss": 0.0084,
+      "step": 5259
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.7311251826595226e-07,
+      "loss": 0.0013,
+      "step": 5260
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.7308816366293226e-07,
+      "loss": 0.0097,
+      "step": 5261
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.730638090599123e-07,
+      "loss": 0.0015,
+      "step": 5262
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 3.730394544568923e-07,
+      "loss": 0.0013,
+      "step": 5263
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.7301509985387236e-07,
+      "loss": 0.0013,
+      "step": 5264
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.7299074525085236e-07,
+      "loss": 0.001,
+      "step": 5265
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.729663906478324e-07,
+      "loss": 0.0011,
+      "step": 5266
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.7294203604481247e-07,
+      "loss": 0.0014,
+      "step": 5267
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.7291768144179247e-07,
+      "loss": 0.0011,
+      "step": 5268
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.728933268387725e-07,
+      "loss": 0.0011,
+      "step": 5269
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.728689722357525e-07,
+      "loss": 0.0012,
+      "step": 5270
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.7284461763273257e-07,
+      "loss": 0.0013,
+      "step": 5271
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.7282026302971257e-07,
+      "loss": 0.0014,
+      "step": 5272
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 3.727959084266926e-07,
+      "loss": 0.001,
+      "step": 5273
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.7277155382367267e-07,
+      "loss": 0.001,
+      "step": 5274
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.7274719922065267e-07,
+      "loss": 0.0012,
+      "step": 5275
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.727228446176327e-07,
+      "loss": 0.0018,
+      "step": 5276
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.726984900146127e-07,
+      "loss": 0.0013,
+      "step": 5277
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.7267413541159277e-07,
+      "loss": 0.0009,
+      "step": 5278
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.7264978080857277e-07,
+      "loss": 0.0012,
+      "step": 5279
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.726254262055528e-07,
+      "loss": 0.0009,
+      "step": 5280
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.7260107160253287e-07,
+      "loss": 0.0012,
+      "step": 5281
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.7257671699951287e-07,
+      "loss": 0.0012,
+      "step": 5282
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 3.725523623964929e-07,
+      "loss": 0.0014,
+      "step": 5283
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.725280077934729e-07,
+      "loss": 0.0014,
+      "step": 5284
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.7250365319045297e-07,
+      "loss": 0.001,
+      "step": 5285
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.7247929858743297e-07,
+      "loss": 0.1951,
+      "step": 5286
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.72454943984413e-07,
+      "loss": 0.0014,
+      "step": 5287
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.72430589381393e-07,
+      "loss": 0.0012,
+      "step": 5288
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.7240623477837307e-07,
+      "loss": 0.0088,
+      "step": 5289
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.723818801753531e-07,
+      "loss": 0.1589,
+      "step": 5290
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.723575255723331e-07,
+      "loss": 0.0014,
+      "step": 5291
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.723331709693132e-07,
+      "loss": 0.0012,
+      "step": 5292
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.723088163662932e-07,
+      "loss": 0.0016,
+      "step": 5293
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 3.722844617632732e-07,
+      "loss": 0.0012,
+      "step": 5294
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.722601071602532e-07,
+      "loss": 0.001,
+      "step": 5295
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.722357525572333e-07,
+      "loss": 0.0011,
+      "step": 5296
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.7221139795421333e-07,
+      "loss": 0.0012,
+      "step": 5297
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.721870433511934e-07,
+      "loss": 0.0013,
+      "step": 5298
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.7216268874817343e-07,
+      "loss": 0.0016,
+      "step": 5299
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.7213833414515343e-07,
+      "loss": 0.001,
+      "step": 5300
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.721139795421335e-07,
+      "loss": 0.0012,
+      "step": 5301
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.720896249391135e-07,
+      "loss": 0.0011,
+      "step": 5302
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.7206527033609353e-07,
+      "loss": 0.0014,
+      "step": 5303
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 3.720409157330736e-07,
+      "loss": 0.0014,
+      "step": 5304
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.720165611300536e-07,
+      "loss": 0.0013,
+      "step": 5305
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.7199220652703363e-07,
+      "loss": 0.0015,
+      "step": 5306
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.7196785192401363e-07,
+      "loss": 0.0014,
+      "step": 5307
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.719434973209937e-07,
+      "loss": 0.0019,
+      "step": 5308
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.719191427179737e-07,
+      "loss": 0.0011,
+      "step": 5309
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.7189478811495373e-07,
+      "loss": 0.0011,
+      "step": 5310
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.7187043351193373e-07,
+      "loss": 0.001,
+      "step": 5311
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.718460789089138e-07,
+      "loss": 0.001,
+      "step": 5312
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.7182172430589384e-07,
+      "loss": 0.001,
+      "step": 5313
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 3.7179736970287383e-07,
+      "loss": 0.001,
+      "step": 5314
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.717730150998539e-07,
+      "loss": 0.0014,
+      "step": 5315
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.717486604968339e-07,
+      "loss": 0.0013,
+      "step": 5316
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.7172430589381394e-07,
+      "loss": 0.0015,
+      "step": 5317
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.7169995129079394e-07,
+      "loss": 0.001,
+      "step": 5318
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.71675596687774e-07,
+      "loss": 0.0014,
+      "step": 5319
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.7165124208475404e-07,
+      "loss": 0.0011,
+      "step": 5320
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.7162688748173404e-07,
+      "loss": 0.001,
+      "step": 5321
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.716025328787141e-07,
+      "loss": 0.0011,
+      "step": 5322
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.715781782756941e-07,
+      "loss": 0.001,
+      "step": 5323
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.7155382367267414e-07,
+      "loss": 0.0009,
+      "step": 5324
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 3.7152946906965414e-07,
+      "loss": 0.0014,
+      "step": 5325
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.715051144666342e-07,
+      "loss": 0.0013,
+      "step": 5326
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.714807598636142e-07,
+      "loss": 0.0009,
+      "step": 5327
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.7145640526059424e-07,
+      "loss": 0.0013,
+      "step": 5328
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.714320506575743e-07,
+      "loss": 0.0012,
+      "step": 5329
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.714076960545543e-07,
+      "loss": 0.0011,
+      "step": 5330
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.7138334145153434e-07,
+      "loss": 0.0014,
+      "step": 5331
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.7135898684851434e-07,
+      "loss": 0.0011,
+      "step": 5332
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.713346322454944e-07,
+      "loss": 0.001,
+      "step": 5333
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.713102776424744e-07,
+      "loss": 0.0009,
+      "step": 5334
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 3.7128592303945444e-07,
+      "loss": 0.0013,
+      "step": 5335
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.712615684364345e-07,
+      "loss": 0.0012,
+      "step": 5336
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.712372138334145e-07,
+      "loss": 0.0012,
+      "step": 5337
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.7121285923039454e-07,
+      "loss": 0.0012,
+      "step": 5338
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.7118850462737454e-07,
+      "loss": 0.002,
+      "step": 5339
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.711641500243546e-07,
+      "loss": 0.0011,
+      "step": 5340
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.711397954213346e-07,
+      "loss": 0.0014,
+      "step": 5341
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.7111544081831465e-07,
+      "loss": 0.0012,
+      "step": 5342
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.7109108621529464e-07,
+      "loss": 0.0013,
+      "step": 5343
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.710667316122747e-07,
+      "loss": 0.001,
+      "step": 5344
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 3.7104237700925475e-07,
+      "loss": 0.001,
+      "step": 5345
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.7101802240623475e-07,
+      "loss": 0.0012,
+      "step": 5346
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.709936678032148e-07,
+      "loss": 0.0014,
+      "step": 5347
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.709693132001948e-07,
+      "loss": 0.001,
+      "step": 5348
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.7094495859717485e-07,
+      "loss": 0.0012,
+      "step": 5349
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.7092060399415485e-07,
+      "loss": 0.0018,
+      "step": 5350
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.708962493911349e-07,
+      "loss": 1.2943,
+      "step": 5351
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.7087189478811495e-07,
+      "loss": 0.0012,
+      "step": 5352
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.7084754018509495e-07,
+      "loss": 0.0011,
+      "step": 5353
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.70823185582075e-07,
+      "loss": 0.0015,
+      "step": 5354
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 3.70798830979055e-07,
+      "loss": 0.0009,
+      "step": 5355
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.7077447637603505e-07,
+      "loss": 0.0018,
+      "step": 5356
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.7075012177301505e-07,
+      "loss": 0.0011,
+      "step": 5357
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.707257671699951e-07,
+      "loss": 0.001,
+      "step": 5358
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.707014125669751e-07,
+      "loss": 0.0011,
+      "step": 5359
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.7067705796395515e-07,
+      "loss": 0.0012,
+      "step": 5360
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.706527033609352e-07,
+      "loss": 0.0011,
+      "step": 5361
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.706283487579152e-07,
+      "loss": 0.001,
+      "step": 5362
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.7060399415489525e-07,
+      "loss": 0.0013,
+      "step": 5363
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.7057963955187525e-07,
+      "loss": 0.0012,
+      "step": 5364
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.705552849488553e-07,
+      "loss": 0.0011,
+      "step": 5365
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 3.705309303458353e-07,
+      "loss": 0.0013,
+      "step": 5366
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.7050657574281535e-07,
+      "loss": 0.001,
+      "step": 5367
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.704822211397954e-07,
+      "loss": 0.0014,
+      "step": 5368
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.704578665367754e-07,
+      "loss": 0.0012,
+      "step": 5369
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.7043351193375546e-07,
+      "loss": 0.0013,
+      "step": 5370
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.7040915733073545e-07,
+      "loss": 0.001,
+      "step": 5371
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.703848027277155e-07,
+      "loss": 0.0012,
+      "step": 5372
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.703604481246955e-07,
+      "loss": 0.0008,
+      "step": 5373
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.7033609352167556e-07,
+      "loss": 0.0012,
+      "step": 5374
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.7031173891865566e-07,
+      "loss": 0.0012,
+      "step": 5375
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 3.7028738431563566e-07,
+      "loss": 0.001,
+      "step": 5376
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.702630297126157e-07,
+      "loss": 0.0011,
+      "step": 5377
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.702386751095957e-07,
+      "loss": 0.0011,
+      "step": 5378
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.7021432050657576e-07,
+      "loss": 0.001,
+      "step": 5379
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.7018996590355576e-07,
+      "loss": 0.0014,
+      "step": 5380
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.701656113005358e-07,
+      "loss": 0.001,
+      "step": 5381
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.701412566975158e-07,
+      "loss": 0.0012,
+      "step": 5382
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.7011690209449586e-07,
+      "loss": 0.001,
+      "step": 5383
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.700925474914759e-07,
+      "loss": 0.0016,
+      "step": 5384
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.700681928884559e-07,
+      "loss": 0.0019,
+      "step": 5385
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 3.7004383828543596e-07,
+      "loss": 0.0011,
+      "step": 5386
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.7001948368241596e-07,
+      "loss": 0.001,
+      "step": 5387
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.69995129079396e-07,
+      "loss": 1.7135,
+      "step": 5388
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.69970774476376e-07,
+      "loss": 0.0011,
+      "step": 5389
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.6994641987335607e-07,
+      "loss": 0.0013,
+      "step": 5390
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.699220652703361e-07,
+      "loss": 0.0013,
+      "step": 5391
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.698977106673161e-07,
+      "loss": 0.0009,
+      "step": 5392
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.6987335606429617e-07,
+      "loss": 0.001,
+      "step": 5393
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.6984900146127617e-07,
+      "loss": 0.0013,
+      "step": 5394
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.698246468582562e-07,
+      "loss": 0.0012,
+      "step": 5395
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.698002922552362e-07,
+      "loss": 0.0013,
+      "step": 5396
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 3.6977593765221627e-07,
+      "loss": 0.0011,
+      "step": 5397
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.6975158304919627e-07,
+      "loss": 0.001,
+      "step": 5398
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.697272284461763e-07,
+      "loss": 0.001,
+      "step": 5399
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.6970287384315637e-07,
+      "loss": 0.0012,
+      "step": 5400
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.6967851924013637e-07,
+      "loss": 0.001,
+      "step": 5401
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.696541646371164e-07,
+      "loss": 0.0011,
+      "step": 5402
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.696298100340964e-07,
+      "loss": 0.0014,
+      "step": 5403
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.6960545543107647e-07,
+      "loss": 0.0011,
+      "step": 5404
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.6958110082805647e-07,
+      "loss": 0.001,
+      "step": 5405
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.695567462250365e-07,
+      "loss": 0.0011,
+      "step": 5406
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 3.6953239162201657e-07,
+      "loss": 0.0015,
+      "step": 5407
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.6950803701899657e-07,
+      "loss": 0.828,
+      "step": 5408
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.694836824159766e-07,
+      "loss": 0.0016,
+      "step": 5409
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.694593278129566e-07,
+      "loss": 0.0011,
+      "step": 5410
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.6943497320993667e-07,
+      "loss": 0.0012,
+      "step": 5411
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.6941061860691667e-07,
+      "loss": 0.0013,
+      "step": 5412
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.693862640038967e-07,
+      "loss": 0.0013,
+      "step": 5413
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.693619094008767e-07,
+      "loss": 0.0011,
+      "step": 5414
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.693375547978568e-07,
+      "loss": 0.001,
+      "step": 5415
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.693132001948368e-07,
+      "loss": 0.0013,
+      "step": 5416
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 3.692888455918168e-07,
+      "loss": 0.001,
+      "step": 5417
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.692644909887969e-07,
+      "loss": 0.0012,
+      "step": 5418
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.692401363857769e-07,
+      "loss": 0.0015,
+      "step": 5419
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.692157817827569e-07,
+      "loss": 0.0014,
+      "step": 5420
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.691914271797369e-07,
+      "loss": 0.0009,
+      "step": 5421
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.69167072576717e-07,
+      "loss": 0.0011,
+      "step": 5422
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.6914271797369703e-07,
+      "loss": 0.0014,
+      "step": 5423
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.6911836337067703e-07,
+      "loss": 0.0011,
+      "step": 5424
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.690940087676571e-07,
+      "loss": 0.0009,
+      "step": 5425
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.690696541646371e-07,
+      "loss": 0.0011,
+      "step": 5426
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 3.6904529956161713e-07,
+      "loss": 0.0012,
+      "step": 5427
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.6902094495859713e-07,
+      "loss": 0.0013,
+      "step": 5428
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.689965903555772e-07,
+      "loss": 0.0015,
+      "step": 5429
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.689722357525572e-07,
+      "loss": 0.0031,
+      "step": 5430
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.6894788114953723e-07,
+      "loss": 0.0011,
+      "step": 5431
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.689235265465173e-07,
+      "loss": 0.0011,
+      "step": 5432
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.688991719434973e-07,
+      "loss": 0.0014,
+      "step": 5433
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.6887481734047733e-07,
+      "loss": 0.0029,
+      "step": 5434
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.6885046273745733e-07,
+      "loss": 0.0012,
+      "step": 5435
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.688261081344374e-07,
+      "loss": 0.0012,
+      "step": 5436
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.688017535314174e-07,
+      "loss": 0.001,
+      "step": 5437
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 3.6877739892839743e-07,
+      "loss": 0.0011,
+      "step": 5438
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.687530443253775e-07,
+      "loss": 0.0012,
+      "step": 5439
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.687286897223575e-07,
+      "loss": 0.0011,
+      "step": 5440
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.6870433511933753e-07,
+      "loss": 0.0012,
+      "step": 5441
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.6867998051631753e-07,
+      "loss": 0.0011,
+      "step": 5442
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.686556259132976e-07,
+      "loss": 0.0013,
+      "step": 5443
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.686312713102776e-07,
+      "loss": 0.0011,
+      "step": 5444
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.6860691670725763e-07,
+      "loss": 0.0009,
+      "step": 5445
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.685825621042377e-07,
+      "loss": 0.2502,
+      "step": 5446
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.685582075012177e-07,
+      "loss": 0.0011,
+      "step": 5447
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 3.6853385289819774e-07,
+      "loss": 0.0011,
+      "step": 5448
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.6850949829517774e-07,
+      "loss": 0.0012,
+      "step": 5449
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.684851436921578e-07,
+      "loss": 0.0013,
+      "step": 5450
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.6846078908913784e-07,
+      "loss": 0.0013,
+      "step": 5451
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.684364344861179e-07,
+      "loss": 0.0015,
+      "step": 5452
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.684120798830979e-07,
+      "loss": 0.001,
+      "step": 5453
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.6838772528007794e-07,
+      "loss": 0.0013,
+      "step": 5454
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.68363370677058e-07,
+      "loss": 0.0012,
+      "step": 5455
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.68339016074038e-07,
+      "loss": 0.0012,
+      "step": 5456
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.6831466147101804e-07,
+      "loss": 0.0014,
+      "step": 5457
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 3.6829030686799804e-07,
+      "loss": 0.0011,
+      "step": 5458
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.682659522649781e-07,
+      "loss": 0.0015,
+      "step": 5459
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.682415976619581e-07,
+      "loss": 0.0012,
+      "step": 5460
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.6821724305893814e-07,
+      "loss": 0.0013,
+      "step": 5461
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.681928884559182e-07,
+      "loss": 0.0013,
+      "step": 5462
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.681685338528982e-07,
+      "loss": 0.0012,
+      "step": 5463
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.6814417924987824e-07,
+      "loss": 0.7705,
+      "step": 5464
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.6811982464685824e-07,
+      "loss": 0.001,
+      "step": 5465
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.680954700438383e-07,
+      "loss": 0.0011,
+      "step": 5466
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.680711154408183e-07,
+      "loss": 0.0016,
+      "step": 5467
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.6804676083779835e-07,
+      "loss": 0.0009,
+      "step": 5468
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 3.680224062347784e-07,
+      "loss": 0.0008,
+      "step": 5469
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.679980516317584e-07,
+      "loss": 0.0022,
+      "step": 5470
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.6797369702873845e-07,
+      "loss": 0.001,
+      "step": 5471
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.6794934242571845e-07,
+      "loss": 0.0015,
+      "step": 5472
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.679249878226985e-07,
+      "loss": 0.0011,
+      "step": 5473
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.679006332196785e-07,
+      "loss": 0.0013,
+      "step": 5474
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.6787627861665855e-07,
+      "loss": 0.0017,
+      "step": 5475
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.6785192401363855e-07,
+      "loss": 0.001,
+      "step": 5476
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.678275694106186e-07,
+      "loss": 0.0012,
+      "step": 5477
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.6780321480759865e-07,
+      "loss": 0.001,
+      "step": 5478
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 3.6777886020457865e-07,
+      "loss": 0.0012,
+      "step": 5479
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.677545056015587e-07,
+      "loss": 0.0013,
+      "step": 5480
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.677301509985387e-07,
+      "loss": 0.0014,
+      "step": 5481
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.6770579639551875e-07,
+      "loss": 0.0012,
+      "step": 5482
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.6768144179249875e-07,
+      "loss": 0.0014,
+      "step": 5483
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.676570871894788e-07,
+      "loss": 0.001,
+      "step": 5484
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.6763273258645885e-07,
+      "loss": 0.0026,
+      "step": 5485
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.6760837798343885e-07,
+      "loss": 0.001,
+      "step": 5486
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.675840233804189e-07,
+      "loss": 0.0012,
+      "step": 5487
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.675596687773989e-07,
+      "loss": 0.001,
+      "step": 5488
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 3.6753531417437895e-07,
+      "loss": 0.0012,
+      "step": 5489
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.6751095957135895e-07,
+      "loss": 0.0012,
+      "step": 5490
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.67486604968339e-07,
+      "loss": 0.0012,
+      "step": 5491
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.67462250365319e-07,
+      "loss": 0.001,
+      "step": 5492
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.6743789576229905e-07,
+      "loss": 0.001,
+      "step": 5493
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.674135411592791e-07,
+      "loss": 0.0013,
+      "step": 5494
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.673891865562591e-07,
+      "loss": 1.4269,
+      "step": 5495
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.6736483195323916e-07,
+      "loss": 0.0011,
+      "step": 5496
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.6734047735021915e-07,
+      "loss": 0.0008,
+      "step": 5497
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.673161227471992e-07,
+      "loss": 0.0011,
+      "step": 5498
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.672917681441792e-07,
+      "loss": 0.0083,
+      "step": 5499
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 3.6726741354115926e-07,
+      "loss": 0.0012,
+      "step": 5500
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.672430589381393e-07,
+      "loss": 0.001,
+      "step": 5501
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.672187043351193e-07,
+      "loss": 0.0009,
+      "step": 5502
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.6719434973209936e-07,
+      "loss": 0.0012,
+      "step": 5503
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.6716999512907936e-07,
+      "loss": 0.001,
+      "step": 5504
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.671456405260594e-07,
+      "loss": 0.0012,
+      "step": 5505
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.671212859230394e-07,
+      "loss": 0.0012,
+      "step": 5506
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.6709693132001946e-07,
+      "loss": 0.0011,
+      "step": 5507
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.6707257671699946e-07,
+      "loss": 0.0012,
+      "step": 5508
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.670482221139795e-07,
+      "loss": 0.0013,
+      "step": 5509
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 3.6702386751095956e-07,
+      "loss": 0.0009,
+      "step": 5510
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.6699951290793956e-07,
+      "loss": 0.001,
+      "step": 5511
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.669751583049196e-07,
+      "loss": 0.0012,
+      "step": 5512
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.669508037018996e-07,
+      "loss": 0.0009,
+      "step": 5513
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.6692644909887966e-07,
+      "loss": 0.0008,
+      "step": 5514
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.6690209449585966e-07,
+      "loss": 0.0013,
+      "step": 5515
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.668777398928397e-07,
+      "loss": 0.0011,
+      "step": 5516
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.6685338528981976e-07,
+      "loss": 0.0012,
+      "step": 5517
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.6682903068679976e-07,
+      "loss": 0.0009,
+      "step": 5518
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.668046760837798e-07,
+      "loss": 0.0011,
+      "step": 5519
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 3.667803214807598e-07,
+      "loss": 0.0009,
+      "step": 5520
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.6675596687773986e-07,
+      "loss": 0.0012,
+      "step": 5521
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.6673161227471986e-07,
+      "loss": 0.0012,
+      "step": 5522
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.667072576716999e-07,
+      "loss": 0.001,
+      "step": 5523
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.666829030686799e-07,
+      "loss": 0.0011,
+      "step": 5524
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.6665854846565997e-07,
+      "loss": 0.0015,
+      "step": 5525
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.6663419386264e-07,
+      "loss": 0.0009,
+      "step": 5526
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.6660983925962e-07,
+      "loss": 0.0015,
+      "step": 5527
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.665854846566001e-07,
+      "loss": 0.0012,
+      "step": 5528
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.665611300535801e-07,
+      "loss": 0.0011,
+      "step": 5529
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 3.6653677545056017e-07,
+      "loss": 0.0014,
+      "step": 5530
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.6651242084754017e-07,
+      "loss": 0.0009,
+      "step": 5531
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.664880662445202e-07,
+      "loss": 0.0011,
+      "step": 5532
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.6646371164150027e-07,
+      "loss": 0.0011,
+      "step": 5533
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.6643935703848027e-07,
+      "loss": 0.0011,
+      "step": 5534
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.664150024354603e-07,
+      "loss": 0.001,
+      "step": 5535
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.663906478324403e-07,
+      "loss": 0.0013,
+      "step": 5536
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.6636629322942037e-07,
+      "loss": 0.001,
+      "step": 5537
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.6634193862640037e-07,
+      "loss": 0.0011,
+      "step": 5538
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.663175840233804e-07,
+      "loss": 0.0012,
+      "step": 5539
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.662932294203605e-07,
+      "loss": 0.0014,
+      "step": 5540
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 3.662688748173405e-07,
+      "loss": 0.001,
+      "step": 5541
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.662445202143205e-07,
+      "loss": 0.001,
+      "step": 5542
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.662201656113005e-07,
+      "loss": 0.0013,
+      "step": 5543
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.661958110082806e-07,
+      "loss": 0.0011,
+      "step": 5544
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.661714564052606e-07,
+      "loss": 0.001,
+      "step": 5545
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.661471018022406e-07,
+      "loss": 0.0013,
+      "step": 5546
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.661227471992206e-07,
+      "loss": 0.001,
+      "step": 5547
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.660983925962007e-07,
+      "loss": 0.001,
+      "step": 5548
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.6607403799318073e-07,
+      "loss": 0.0014,
+      "step": 5549
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.6604968339016073e-07,
+      "loss": 0.0015,
+      "step": 5550
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 3.660253287871408e-07,
+      "loss": 0.001,
+      "step": 5551
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.660009741841208e-07,
+      "loss": 0.0008,
+      "step": 5552
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.6597661958110083e-07,
+      "loss": 0.0009,
+      "step": 5553
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.6595226497808083e-07,
+      "loss": 0.0009,
+      "step": 5554
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.659279103750609e-07,
+      "loss": 0.0011,
+      "step": 5555
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.6590355577204093e-07,
+      "loss": 0.0015,
+      "step": 5556
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.6587920116902093e-07,
+      "loss": 0.0011,
+      "step": 5557
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.65854846566001e-07,
+      "loss": 0.0012,
+      "step": 5558
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.65830491962981e-07,
+      "loss": 0.0016,
+      "step": 5559
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.6580613735996103e-07,
+      "loss": 0.0011,
+      "step": 5560
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 3.6578178275694103e-07,
+      "loss": 0.0011,
+      "step": 5561
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.657574281539211e-07,
+      "loss": 0.001,
+      "step": 5562
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.657330735509011e-07,
+      "loss": 0.0018,
+      "step": 5563
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.6570871894788113e-07,
+      "loss": 0.001,
+      "step": 5564
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.656843643448612e-07,
+      "loss": 0.0011,
+      "step": 5565
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.656600097418412e-07,
+      "loss": 0.0011,
+      "step": 5566
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.6563565513882123e-07,
+      "loss": 0.0011,
+      "step": 5567
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.6561130053580123e-07,
+      "loss": 0.001,
+      "step": 5568
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.655869459327813e-07,
+      "loss": 0.0011,
+      "step": 5569
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.655625913297613e-07,
+      "loss": 0.0009,
+      "step": 5570
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.6553823672674133e-07,
+      "loss": 0.0009,
+      "step": 5571
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 3.655138821237214e-07,
+      "loss": 0.0036,
+      "step": 5572
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.654895275207014e-07,
+      "loss": 0.0011,
+      "step": 5573
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.6546517291768144e-07,
+      "loss": 0.0012,
+      "step": 5574
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.6544081831466144e-07,
+      "loss": 0.0291,
+      "step": 5575
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.654164637116415e-07,
+      "loss": 0.0013,
+      "step": 5576
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.653921091086215e-07,
+      "loss": 0.0012,
+      "step": 5577
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.6536775450560154e-07,
+      "loss": 0.001,
+      "step": 5578
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.6534339990258154e-07,
+      "loss": 0.0017,
+      "step": 5579
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.653190452995616e-07,
+      "loss": 0.001,
+      "step": 5580
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.6529469069654164e-07,
+      "loss": 0.0032,
+      "step": 5581
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 3.6527033609352164e-07,
+      "loss": 0.001,
+      "step": 5582
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.652459814905017e-07,
+      "loss": 0.0012,
+      "step": 5583
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.652216268874817e-07,
+      "loss": 0.0013,
+      "step": 5584
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.6519727228446174e-07,
+      "loss": 0.0015,
+      "step": 5585
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.6517291768144174e-07,
+      "loss": 0.0012,
+      "step": 5586
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.651485630784218e-07,
+      "loss": 0.0013,
+      "step": 5587
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.6512420847540184e-07,
+      "loss": 0.0009,
+      "step": 5588
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.6509985387238184e-07,
+      "loss": 0.0011,
+      "step": 5589
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.650754992693619e-07,
+      "loss": 0.0012,
+      "step": 5590
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.650511446663419e-07,
+      "loss": 0.0009,
+      "step": 5591
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 3.6502679006332194e-07,
+      "loss": 0.0009,
+      "step": 5592
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.6500243546030194e-07,
+      "loss": 0.0011,
+      "step": 5593
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.64978080857282e-07,
+      "loss": 0.0011,
+      "step": 5594
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.64953726254262e-07,
+      "loss": 0.001,
+      "step": 5595
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.6492937165124204e-07,
+      "loss": 0.001,
+      "step": 5596
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.649050170482221e-07,
+      "loss": 0.0012,
+      "step": 5597
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.648806624452021e-07,
+      "loss": 0.001,
+      "step": 5598
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.6485630784218215e-07,
+      "loss": 0.001,
+      "step": 5599
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.6483195323916214e-07,
+      "loss": 0.0041,
+      "step": 5600
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.648075986361422e-07,
+      "loss": 0.1548,
+      "step": 5601
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 3.647832440331222e-07,
+      "loss": 0.0012,
+      "step": 5602
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.6475888943010225e-07,
+      "loss": 0.0013,
+      "step": 5603
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.6473453482708235e-07,
+      "loss": 0.001,
+      "step": 5604
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.6471018022406235e-07,
+      "loss": 0.0009,
+      "step": 5605
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.646858256210424e-07,
+      "loss": 0.0012,
+      "step": 5606
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.646614710180224e-07,
+      "loss": 0.0056,
+      "step": 5607
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.6463711641500245e-07,
+      "loss": 0.0012,
+      "step": 5608
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.6461276181198245e-07,
+      "loss": 0.0011,
+      "step": 5609
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.645884072089625e-07,
+      "loss": 0.0013,
+      "step": 5610
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.6456405260594255e-07,
+      "loss": 0.0011,
+      "step": 5611
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.6453969800292255e-07,
+      "loss": 0.0088,
+      "step": 5612
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 3.645153433999026e-07,
+      "loss": 0.0011,
+      "step": 5613
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.644909887968826e-07,
+      "loss": 0.0011,
+      "step": 5614
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.6446663419386265e-07,
+      "loss": 0.001,
+      "step": 5615
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.6444227959084265e-07,
+      "loss": 0.001,
+      "step": 5616
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.644179249878227e-07,
+      "loss": 0.0022,
+      "step": 5617
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.643935703848027e-07,
+      "loss": 0.0011,
+      "step": 5618
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.6436921578178275e-07,
+      "loss": 0.0009,
+      "step": 5619
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.643448611787628e-07,
+      "loss": 0.001,
+      "step": 5620
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.643205065757428e-07,
+      "loss": 0.0015,
+      "step": 5621
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.6429615197272286e-07,
+      "loss": 0.0009,
+      "step": 5622
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 3.6427179736970286e-07,
+      "loss": 0.0013,
+      "step": 5623
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.642474427666829e-07,
+      "loss": 0.0013,
+      "step": 5624
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.642230881636629e-07,
+      "loss": 0.0011,
+      "step": 5625
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.6419873356064296e-07,
+      "loss": 0.0012,
+      "step": 5626
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.64174378957623e-07,
+      "loss": 0.0011,
+      "step": 5627
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.64150024354603e-07,
+      "loss": 0.0013,
+      "step": 5628
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.6412566975158306e-07,
+      "loss": 0.0014,
+      "step": 5629
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.6410131514856306e-07,
+      "loss": 0.0012,
+      "step": 5630
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.640769605455431e-07,
+      "loss": 0.0011,
+      "step": 5631
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.640526059425231e-07,
+      "loss": 0.001,
+      "step": 5632
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 3.6402825133950316e-07,
+      "loss": 0.0013,
+      "step": 5633
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.6400389673648316e-07,
+      "loss": 0.0015,
+      "step": 5634
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.639795421334632e-07,
+      "loss": 0.0009,
+      "step": 5635
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.6395518753044326e-07,
+      "loss": 0.0008,
+      "step": 5636
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.6393083292742326e-07,
+      "loss": 0.001,
+      "step": 5637
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.639064783244033e-07,
+      "loss": 0.0009,
+      "step": 5638
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.638821237213833e-07,
+      "loss": 0.0011,
+      "step": 5639
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.6385776911836336e-07,
+      "loss": 0.0008,
+      "step": 5640
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.6383341451534336e-07,
+      "loss": 0.3637,
+      "step": 5641
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.638090599123234e-07,
+      "loss": 0.0017,
+      "step": 5642
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.6378470530930346e-07,
+      "loss": 0.4188,
+      "step": 5643
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 3.6376035070628346e-07,
+      "loss": 0.0016,
+      "step": 5644
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.637359961032635e-07,
+      "loss": 0.0011,
+      "step": 5645
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.637116415002435e-07,
+      "loss": 0.0009,
+      "step": 5646
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.6368728689722356e-07,
+      "loss": 0.001,
+      "step": 5647
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.6366293229420356e-07,
+      "loss": 0.0015,
+      "step": 5648
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.636385776911836e-07,
+      "loss": 0.001,
+      "step": 5649
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.6361422308816367e-07,
+      "loss": 0.0012,
+      "step": 5650
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.6358986848514367e-07,
+      "loss": 0.001,
+      "step": 5651
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.635655138821237e-07,
+      "loss": 0.0012,
+      "step": 5652
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.635411592791037e-07,
+      "loss": 0.001,
+      "step": 5653
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 3.6351680467608377e-07,
+      "loss": 0.0012,
+      "step": 5654
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.6349245007306377e-07,
+      "loss": 0.0009,
+      "step": 5655
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.634680954700438e-07,
+      "loss": 0.001,
+      "step": 5656
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.634437408670238e-07,
+      "loss": 0.0012,
+      "step": 5657
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.6341938626400387e-07,
+      "loss": 0.0009,
+      "step": 5658
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.633950316609839e-07,
+      "loss": 0.212,
+      "step": 5659
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.633706770579639e-07,
+      "loss": 0.0012,
+      "step": 5660
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.6334632245494397e-07,
+      "loss": 0.0011,
+      "step": 5661
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.6332196785192397e-07,
+      "loss": 1.5369,
+      "step": 5662
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.63297613248904e-07,
+      "loss": 0.001,
+      "step": 5663
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 3.63273258645884e-07,
+      "loss": 0.0012,
+      "step": 5664
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.6324890404286407e-07,
+      "loss": 0.0011,
+      "step": 5665
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.632245494398441e-07,
+      "loss": 0.0012,
+      "step": 5666
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.632001948368241e-07,
+      "loss": 0.0011,
+      "step": 5667
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.6317584023380417e-07,
+      "loss": 0.0011,
+      "step": 5668
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.6315148563078417e-07,
+      "loss": 0.0009,
+      "step": 5669
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.631271310277642e-07,
+      "loss": 0.001,
+      "step": 5670
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.631027764247442e-07,
+      "loss": 0.001,
+      "step": 5671
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.6307842182172427e-07,
+      "loss": 0.0009,
+      "step": 5672
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.6305406721870427e-07,
+      "loss": 0.0011,
+      "step": 5673
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 3.630297126156843e-07,
+      "loss": 0.0009,
+      "step": 5674
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.630053580126644e-07,
+      "loss": 0.0011,
+      "step": 5675
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.629810034096444e-07,
+      "loss": 0.0011,
+      "step": 5676
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.629566488066244e-07,
+      "loss": 0.0008,
+      "step": 5677
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.629322942036044e-07,
+      "loss": 0.0008,
+      "step": 5678
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.629079396005845e-07,
+      "loss": 0.0009,
+      "step": 5679
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.628835849975645e-07,
+      "loss": 0.0012,
+      "step": 5680
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.628592303945446e-07,
+      "loss": 0.0013,
+      "step": 5681
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.6283487579152463e-07,
+      "loss": 0.001,
+      "step": 5682
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.6281052118850463e-07,
+      "loss": 0.0011,
+      "step": 5683
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.627861665854847e-07,
+      "loss": 0.001,
+      "step": 5684
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 3.627618119824647e-07,
+      "loss": 0.3245,
+      "step": 5685
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.6273745737944473e-07,
+      "loss": 0.001,
+      "step": 5686
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.6271310277642473e-07,
+      "loss": 0.0011,
+      "step": 5687
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.626887481734048e-07,
+      "loss": 0.0012,
+      "step": 5688
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.6266439357038483e-07,
+      "loss": 0.0009,
+      "step": 5689
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.6264003896736483e-07,
+      "loss": 0.001,
+      "step": 5690
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.626156843643449e-07,
+      "loss": 0.0008,
+      "step": 5691
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.625913297613249e-07,
+      "loss": 0.0008,
+      "step": 5692
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.6256697515830493e-07,
+      "loss": 0.001,
+      "step": 5693
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.6254262055528493e-07,
+      "loss": 0.0013,
+      "step": 5694
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 3.62518265952265e-07,
+      "loss": 0.0012,
+      "step": 5695
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.62493911349245e-07,
+      "loss": 0.001,
+      "step": 5696
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.6246955674622503e-07,
+      "loss": 0.001,
+      "step": 5697
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.624452021432051e-07,
+      "loss": 0.0009,
+      "step": 5698
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.624208475401851e-07,
+      "loss": 0.001,
+      "step": 5699
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.6239649293716514e-07,
+      "loss": 0.0011,
+      "step": 5700
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.6237213833414514e-07,
+      "loss": 0.001,
+      "step": 5701
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.623477837311252e-07,
+      "loss": 0.0012,
+      "step": 5702
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.623234291281052e-07,
+      "loss": 0.0011,
+      "step": 5703
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.6229907452508524e-07,
+      "loss": 0.0012,
+      "step": 5704
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 3.622747199220653e-07,
+      "loss": 0.0013,
+      "step": 5705
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.622503653190453e-07,
+      "loss": 0.0013,
+      "step": 5706
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.6222601071602534e-07,
+      "loss": 0.001,
+      "step": 5707
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.6220165611300534e-07,
+      "loss": 0.0009,
+      "step": 5708
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.621773015099854e-07,
+      "loss": 0.0013,
+      "step": 5709
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.621529469069654e-07,
+      "loss": 0.001,
+      "step": 5710
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.6212859230394544e-07,
+      "loss": 0.0009,
+      "step": 5711
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.6210423770092544e-07,
+      "loss": 0.0008,
+      "step": 5712
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.620798830979055e-07,
+      "loss": 0.0012,
+      "step": 5713
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.6205552849488554e-07,
+      "loss": 0.0012,
+      "step": 5714
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.6203117389186554e-07,
+      "loss": 0.0013,
+      "step": 5715
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 3.620068192888456e-07,
+      "loss": 0.0076,
+      "step": 5716
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.619824646858256e-07,
+      "loss": 0.8025,
+      "step": 5717
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.6195811008280564e-07,
+      "loss": 0.0013,
+      "step": 5718
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.6193375547978564e-07,
+      "loss": 0.001,
+      "step": 5719
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.619094008767657e-07,
+      "loss": 0.001,
+      "step": 5720
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.6188504627374574e-07,
+      "loss": 0.0009,
+      "step": 5721
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.6186069167072574e-07,
+      "loss": 0.2355,
+      "step": 5722
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.618363370677058e-07,
+      "loss": 0.001,
+      "step": 5723
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.618119824646858e-07,
+      "loss": 0.0012,
+      "step": 5724
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.6178762786166585e-07,
+      "loss": 0.002,
+      "step": 5725
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 3.6176327325864584e-07,
+      "loss": 0.0013,
+      "step": 5726
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.617389186556259e-07,
+      "loss": 0.001,
+      "step": 5727
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.617145640526059e-07,
+      "loss": 0.0009,
+      "step": 5728
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.6169020944958595e-07,
+      "loss": 0.0009,
+      "step": 5729
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.61665854846566e-07,
+      "loss": 0.0011,
+      "step": 5730
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.61641500243546e-07,
+      "loss": 0.001,
+      "step": 5731
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.6161714564052605e-07,
+      "loss": 0.0009,
+      "step": 5732
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.6159279103750605e-07,
+      "loss": 0.001,
+      "step": 5733
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.615684364344861e-07,
+      "loss": 0.0009,
+      "step": 5734
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.615440818314661e-07,
+      "loss": 0.0024,
+      "step": 5735
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 3.6151972722844615e-07,
+      "loss": 0.0008,
+      "step": 5736
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.614953726254262e-07,
+      "loss": 0.0009,
+      "step": 5737
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.614710180224062e-07,
+      "loss": 0.0011,
+      "step": 5738
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.6144666341938625e-07,
+      "loss": 0.0013,
+      "step": 5739
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.6142230881636625e-07,
+      "loss": 0.0014,
+      "step": 5740
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.613979542133463e-07,
+      "loss": 0.0016,
+      "step": 5741
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.613735996103263e-07,
+      "loss": 0.0008,
+      "step": 5742
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.6134924500730635e-07,
+      "loss": 0.0009,
+      "step": 5743
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.6132489040428635e-07,
+      "loss": 0.0011,
+      "step": 5744
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.613005358012664e-07,
+      "loss": 0.0012,
+      "step": 5745
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 3.6127618119824645e-07,
+      "loss": 0.001,
+      "step": 5746
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.6125182659522645e-07,
+      "loss": 0.0011,
+      "step": 5747
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.612274719922065e-07,
+      "loss": 0.001,
+      "step": 5748
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.612031173891865e-07,
+      "loss": 0.0012,
+      "step": 5749
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.6117876278616655e-07,
+      "loss": 0.0009,
+      "step": 5750
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.6115440818314655e-07,
+      "loss": 0.001,
+      "step": 5751
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.611300535801266e-07,
+      "loss": 0.0011,
+      "step": 5752
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.6110569897710666e-07,
+      "loss": 0.0012,
+      "step": 5753
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.6108134437408665e-07,
+      "loss": 0.0009,
+      "step": 5754
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.610569897710667e-07,
+      "loss": 0.0015,
+      "step": 5755
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.610326351680467e-07,
+      "loss": 0.001,
+      "step": 5756
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 3.6100828056502676e-07,
+      "loss": 0.001,
+      "step": 5757
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.609839259620068e-07,
+      "loss": 0.0009,
+      "step": 5758
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.6095957135898686e-07,
+      "loss": 0.0009,
+      "step": 5759
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.609352167559669e-07,
+      "loss": 0.0009,
+      "step": 5760
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.609108621529469e-07,
+      "loss": 0.0012,
+      "step": 5761
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.6088650754992696e-07,
+      "loss": 0.0016,
+      "step": 5762
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.6086215294690696e-07,
+      "loss": 0.0008,
+      "step": 5763
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.60837798343887e-07,
+      "loss": 0.0011,
+      "step": 5764
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.60813443740867e-07,
+      "loss": 0.0011,
+      "step": 5765
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.6078908913784706e-07,
+      "loss": 0.0132,
+      "step": 5766
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 3.6076473453482706e-07,
+      "loss": 0.0009,
+      "step": 5767
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.607403799318071e-07,
+      "loss": 0.0012,
+      "step": 5768
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.6071602532878716e-07,
+      "loss": 0.0009,
+      "step": 5769
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.6069167072576716e-07,
+      "loss": 0.001,
+      "step": 5770
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.606673161227472e-07,
+      "loss": 0.0013,
+      "step": 5771
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.606429615197272e-07,
+      "loss": 0.258,
+      "step": 5772
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.6061860691670727e-07,
+      "loss": 0.0007,
+      "step": 5773
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.6059425231368726e-07,
+      "loss": 0.0009,
+      "step": 5774
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.605698977106673e-07,
+      "loss": 0.0008,
+      "step": 5775
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.6054554310764737e-07,
+      "loss": 0.0008,
+      "step": 5776
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 3.6052118850462737e-07,
+      "loss": 0.0011,
+      "step": 5777
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.604968339016074e-07,
+      "loss": 0.0011,
+      "step": 5778
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.604724792985874e-07,
+      "loss": 0.0009,
+      "step": 5779
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.6044812469556747e-07,
+      "loss": 0.0018,
+      "step": 5780
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.6042377009254747e-07,
+      "loss": 0.0012,
+      "step": 5781
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.603994154895275e-07,
+      "loss": 0.0009,
+      "step": 5782
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.603750608865075e-07,
+      "loss": 0.0009,
+      "step": 5783
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.6035070628348757e-07,
+      "loss": 0.001,
+      "step": 5784
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.603263516804676e-07,
+      "loss": 0.0009,
+      "step": 5785
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.603019970774476e-07,
+      "loss": 0.0009,
+      "step": 5786
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.6027764247442767e-07,
+      "loss": 0.0011,
+      "step": 5787
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 3.6025328787140767e-07,
+      "loss": 0.001,
+      "step": 5788
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.602289332683877e-07,
+      "loss": 0.0011,
+      "step": 5789
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.602045786653677e-07,
+      "loss": 0.0011,
+      "step": 5790
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.6018022406234777e-07,
+      "loss": 0.6585,
+      "step": 5791
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.601558694593278e-07,
+      "loss": 0.0008,
+      "step": 5792
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.601315148563078e-07,
+      "loss": 0.0011,
+      "step": 5793
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.6010716025328787e-07,
+      "loss": 0.0011,
+      "step": 5794
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.6008280565026787e-07,
+      "loss": 0.0009,
+      "step": 5795
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.600584510472479e-07,
+      "loss": 0.001,
+      "step": 5796
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.600340964442279e-07,
+      "loss": 0.001,
+      "step": 5797
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 3.60009741841208e-07,
+      "loss": 0.001,
+      "step": 5798
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.5998538723818797e-07,
+      "loss": 0.0008,
+      "step": 5799
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.59961032635168e-07,
+      "loss": 0.0011,
+      "step": 5800
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.599366780321481e-07,
+      "loss": 1.3154,
+      "step": 5801
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.599123234291281e-07,
+      "loss": 0.0009,
+      "step": 5802
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.598879688261081e-07,
+      "loss": 0.001,
+      "step": 5803
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.598636142230881e-07,
+      "loss": 0.002,
+      "step": 5804
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.598392596200682e-07,
+      "loss": 0.001,
+      "step": 5805
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.598149050170482e-07,
+      "loss": 1.5724,
+      "step": 5806
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.5979055041402823e-07,
+      "loss": 0.0009,
+      "step": 5807
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 3.597661958110083e-07,
+      "loss": 0.0011,
+      "step": 5808
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.597418412079883e-07,
+      "loss": 0.0007,
+      "step": 5809
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.5971748660496833e-07,
+      "loss": 0.0008,
+      "step": 5810
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.5969313200194833e-07,
+      "loss": 0.003,
+      "step": 5811
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.596687773989284e-07,
+      "loss": 0.0014,
+      "step": 5812
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.596444227959084e-07,
+      "loss": 0.001,
+      "step": 5813
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.5962006819288843e-07,
+      "loss": 0.0009,
+      "step": 5814
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.595957135898685e-07,
+      "loss": 0.001,
+      "step": 5815
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.595713589868485e-07,
+      "loss": 0.0015,
+      "step": 5816
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.5954700438382853e-07,
+      "loss": 0.0008,
+      "step": 5817
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 3.5952264978080853e-07,
+      "loss": 0.0013,
+      "step": 5818
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.594982951777886e-07,
+      "loss": 0.0013,
+      "step": 5819
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.594739405747686e-07,
+      "loss": 0.0013,
+      "step": 5820
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.5944958597174863e-07,
+      "loss": 0.0015,
+      "step": 5821
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.5942523136872863e-07,
+      "loss": 0.0009,
+      "step": 5822
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.594008767657087e-07,
+      "loss": 0.001,
+      "step": 5823
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.5937652216268873e-07,
+      "loss": 0.001,
+      "step": 5824
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.5935216755966873e-07,
+      "loss": 0.001,
+      "step": 5825
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.593278129566488e-07,
+      "loss": 0.0012,
+      "step": 5826
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.593034583536288e-07,
+      "loss": 0.0011,
+      "step": 5827
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.5927910375060883e-07,
+      "loss": 0.0011,
+      "step": 5828
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 3.5925474914758883e-07,
+      "loss": 0.0009,
+      "step": 5829
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.592303945445689e-07,
+      "loss": 0.0012,
+      "step": 5830
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.5920603994154894e-07,
+      "loss": 0.0008,
+      "step": 5831
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.5918168533852894e-07,
+      "loss": 0.001,
+      "step": 5832
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.59157330735509e-07,
+      "loss": 0.0011,
+      "step": 5833
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.5913297613248904e-07,
+      "loss": 0.0019,
+      "step": 5834
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.591086215294691e-07,
+      "loss": 0.0011,
+      "step": 5835
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.590842669264491e-07,
+      "loss": 0.0009,
+      "step": 5836
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.5905991232342914e-07,
+      "loss": 0.0009,
+      "step": 5837
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.590355577204092e-07,
+      "loss": 0.0011,
+      "step": 5838
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 3.590112031173892e-07,
+      "loss": 0.0008,
+      "step": 5839
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.5898684851436924e-07,
+      "loss": 0.0018,
+      "step": 5840
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.5896249391134924e-07,
+      "loss": 0.001,
+      "step": 5841
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.589381393083293e-07,
+      "loss": 0.001,
+      "step": 5842
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.589137847053093e-07,
+      "loss": 0.0008,
+      "step": 5843
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.5888943010228934e-07,
+      "loss": 0.001,
+      "step": 5844
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.5886507549926934e-07,
+      "loss": 0.0009,
+      "step": 5845
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.588407208962494e-07,
+      "loss": 0.0013,
+      "step": 5846
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.5881636629322944e-07,
+      "loss": 0.0012,
+      "step": 5847
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.5879201169020944e-07,
+      "loss": 1.0375,
+      "step": 5848
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 3.587676570871895e-07,
+      "loss": 0.001,
+      "step": 5849
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.587433024841695e-07,
+      "loss": 0.001,
+      "step": 5850
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.5871894788114955e-07,
+      "loss": 0.0008,
+      "step": 5851
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.5869459327812954e-07,
+      "loss": 0.0011,
+      "step": 5852
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.586702386751096e-07,
+      "loss": 0.001,
+      "step": 5853
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.5864588407208965e-07,
+      "loss": 0.0019,
+      "step": 5854
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.5862152946906965e-07,
+      "loss": 0.0007,
+      "step": 5855
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.585971748660497e-07,
+      "loss": 0.0011,
+      "step": 5856
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.585728202630297e-07,
+      "loss": 0.0009,
+      "step": 5857
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.5854846566000975e-07,
+      "loss": 0.0011,
+      "step": 5858
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.5852411105698975e-07,
+      "loss": 0.001,
+      "step": 5859
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 3.584997564539698e-07,
+      "loss": 0.0009,
+      "step": 5860
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.584754018509498e-07,
+      "loss": 0.0011,
+      "step": 5861
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.5845104724792985e-07,
+      "loss": 0.0008,
+      "step": 5862
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.584266926449099e-07,
+      "loss": 0.0009,
+      "step": 5863
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.584023380418899e-07,
+      "loss": 0.001,
+      "step": 5864
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.5837798343886995e-07,
+      "loss": 0.001,
+      "step": 5865
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.5835362883584995e-07,
+      "loss": 0.0007,
+      "step": 5866
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.5832927423283e-07,
+      "loss": 0.0011,
+      "step": 5867
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.5830491962981e-07,
+      "loss": 0.0009,
+      "step": 5868
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.5828056502679005e-07,
+      "loss": 0.001,
+      "step": 5869
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 3.582562104237701e-07,
+      "loss": 0.0012,
+      "step": 5870
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.582318558207501e-07,
+      "loss": 0.0009,
+      "step": 5871
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.5820750121773015e-07,
+      "loss": 0.0012,
+      "step": 5872
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.5818314661471015e-07,
+      "loss": 0.001,
+      "step": 5873
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.581587920116902e-07,
+      "loss": 0.001,
+      "step": 5874
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.581344374086702e-07,
+      "loss": 0.0012,
+      "step": 5875
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.5811008280565025e-07,
+      "loss": 0.001,
+      "step": 5876
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.5808572820263025e-07,
+      "loss": 0.0011,
+      "step": 5877
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.580613735996103e-07,
+      "loss": 0.0009,
+      "step": 5878
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.5803701899659036e-07,
+      "loss": 0.0009,
+      "step": 5879
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 3.5801266439357035e-07,
+      "loss": 0.0009,
+      "step": 5880
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.579883097905504e-07,
+      "loss": 0.0011,
+      "step": 5881
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.579639551875304e-07,
+      "loss": 0.001,
+      "step": 5882
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.5793960058451046e-07,
+      "loss": 0.001,
+      "step": 5883
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.5791524598149046e-07,
+      "loss": 0.001,
+      "step": 5884
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.578908913784705e-07,
+      "loss": 0.0011,
+      "step": 5885
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.5786653677545056e-07,
+      "loss": 0.0008,
+      "step": 5886
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.5784218217243056e-07,
+      "loss": 0.001,
+      "step": 5887
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.578178275694106e-07,
+      "loss": 0.0018,
+      "step": 5888
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.577934729663906e-07,
+      "loss": 0.001,
+      "step": 5889
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.5776911836337066e-07,
+      "loss": 0.0009,
+      "step": 5890
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 3.5774476376035066e-07,
+      "loss": 0.0009,
+      "step": 5891
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.577204091573307e-07,
+      "loss": 0.001,
+      "step": 5892
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.576960545543107e-07,
+      "loss": 0.001,
+      "step": 5893
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.5767169995129076e-07,
+      "loss": 0.0008,
+      "step": 5894
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.576473453482708e-07,
+      "loss": 0.001,
+      "step": 5895
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.576229907452508e-07,
+      "loss": 0.0008,
+      "step": 5896
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.5759863614223086e-07,
+      "loss": 0.0009,
+      "step": 5897
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.5757428153921086e-07,
+      "loss": 0.0011,
+      "step": 5898
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.575499269361909e-07,
+      "loss": 0.001,
+      "step": 5899
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.575255723331709e-07,
+      "loss": 0.001,
+      "step": 5900
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 3.5750121773015096e-07,
+      "loss": 0.001,
+      "step": 5901
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.57476863127131e-07,
+      "loss": 0.0011,
+      "step": 5902
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.57452508524111e-07,
+      "loss": 0.001,
+      "step": 5903
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.5742815392109106e-07,
+      "loss": 0.0011,
+      "step": 5904
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.5740379931807106e-07,
+      "loss": 0.001,
+      "step": 5905
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.573794447150511e-07,
+      "loss": 0.0014,
+      "step": 5906
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.573550901120311e-07,
+      "loss": 0.0008,
+      "step": 5907
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.5733073550901117e-07,
+      "loss": 0.0011,
+      "step": 5908
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.5730638090599116e-07,
+      "loss": 0.0011,
+      "step": 5909
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.572820263029712e-07,
+      "loss": 0.0012,
+      "step": 5910
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 3.572576716999513e-07,
+      "loss": 0.0008,
+      "step": 5911
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.572333170969313e-07,
+      "loss": 0.0015,
+      "step": 5912
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.5720896249391137e-07,
+      "loss": 0.0012,
+      "step": 5913
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.5718460789089137e-07,
+      "loss": 0.0009,
+      "step": 5914
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.571602532878714e-07,
+      "loss": 0.001,
+      "step": 5915
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.571358986848514e-07,
+      "loss": 0.001,
+      "step": 5916
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.5711154408183147e-07,
+      "loss": 0.001,
+      "step": 5917
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.570871894788115e-07,
+      "loss": 0.0008,
+      "step": 5918
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.570628348757915e-07,
+      "loss": 0.0008,
+      "step": 5919
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.5703848027277157e-07,
+      "loss": 0.0009,
+      "step": 5920
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 3.5701412566975157e-07,
+      "loss": 0.0105,
+      "step": 5921
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.569897710667316e-07,
+      "loss": 0.0012,
+      "step": 5922
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.569654164637116e-07,
+      "loss": 0.0009,
+      "step": 5923
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.569410618606917e-07,
+      "loss": 0.001,
+      "step": 5924
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.569167072576717e-07,
+      "loss": 0.001,
+      "step": 5925
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.568923526546517e-07,
+      "loss": 0.0016,
+      "step": 5926
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.568679980516318e-07,
+      "loss": 0.0013,
+      "step": 5927
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.568436434486118e-07,
+      "loss": 0.0009,
+      "step": 5928
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.568192888455918e-07,
+      "loss": 0.0008,
+      "step": 5929
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.567949342425718e-07,
+      "loss": 0.0014,
+      "step": 5930
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.567705796395519e-07,
+      "loss": 0.0015,
+      "step": 5931
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 3.567462250365319e-07,
+      "loss": 0.0008,
+      "step": 5932
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.5672187043351193e-07,
+      "loss": 0.0015,
+      "step": 5933
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.56697515830492e-07,
+      "loss": 0.0009,
+      "step": 5934
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.56673161227472e-07,
+      "loss": 1.8343,
+      "step": 5935
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.5664880662445203e-07,
+      "loss": 0.0013,
+      "step": 5936
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.5662445202143203e-07,
+      "loss": 0.0011,
+      "step": 5937
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.566000974184121e-07,
+      "loss": 0.0012,
+      "step": 5938
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.565757428153921e-07,
+      "loss": 0.0017,
+      "step": 5939
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.5655138821237213e-07,
+      "loss": 0.0009,
+      "step": 5940
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.565270336093522e-07,
+      "loss": 0.018,
+      "step": 5941
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 3.565026790063322e-07,
+      "loss": 0.7305,
+      "step": 5942
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.5647832440331223e-07,
+      "loss": 0.0011,
+      "step": 5943
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.5645396980029223e-07,
+      "loss": 0.0012,
+      "step": 5944
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.564296151972723e-07,
+      "loss": 0.0029,
+      "step": 5945
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.564052605942523e-07,
+      "loss": 0.0012,
+      "step": 5946
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.5638090599123233e-07,
+      "loss": 0.0013,
+      "step": 5947
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.5635655138821233e-07,
+      "loss": 0.0008,
+      "step": 5948
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.563321967851924e-07,
+      "loss": 0.0008,
+      "step": 5949
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.5630784218217243e-07,
+      "loss": 0.0014,
+      "step": 5950
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.5628348757915243e-07,
+      "loss": 0.0018,
+      "step": 5951
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 3.562591329761325e-07,
+      "loss": 0.0009,
+      "step": 5952
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.562347783731125e-07,
+      "loss": 0.0009,
+      "step": 5953
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.5621042377009253e-07,
+      "loss": 0.0008,
+      "step": 5954
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.5618606916707253e-07,
+      "loss": 2.1872,
+      "step": 5955
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.561617145640526e-07,
+      "loss": 0.0013,
+      "step": 5956
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.5613735996103264e-07,
+      "loss": 0.0009,
+      "step": 5957
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.5611300535801264e-07,
+      "loss": 0.001,
+      "step": 5958
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.560886507549927e-07,
+      "loss": 0.0009,
+      "step": 5959
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.560642961519727e-07,
+      "loss": 0.0009,
+      "step": 5960
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.5603994154895274e-07,
+      "loss": 0.0011,
+      "step": 5961
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.5601558694593274e-07,
+      "loss": 0.0012,
+      "step": 5962
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 3.559912323429128e-07,
+      "loss": 0.001,
+      "step": 5963
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.559668777398928e-07,
+      "loss": 0.0281,
+      "step": 5964
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.5594252313687284e-07,
+      "loss": 0.0012,
+      "step": 5965
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.559181685338529e-07,
+      "loss": 0.001,
+      "step": 5966
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.558938139308329e-07,
+      "loss": 0.001,
+      "step": 5967
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.5586945932781294e-07,
+      "loss": 0.0011,
+      "step": 5968
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.5584510472479294e-07,
+      "loss": 0.0009,
+      "step": 5969
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.55820750121773e-07,
+      "loss": 0.0009,
+      "step": 5970
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.55796395518753e-07,
+      "loss": 0.0013,
+      "step": 5971
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.5577204091573304e-07,
+      "loss": 0.0009,
+      "step": 5972
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 3.557476863127131e-07,
+      "loss": 0.001,
+      "step": 5973
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.557233317096931e-07,
+      "loss": 0.001,
+      "step": 5974
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.5569897710667314e-07,
+      "loss": 0.0013,
+      "step": 5975
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.5567462250365314e-07,
+      "loss": 0.0011,
+      "step": 5976
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.556502679006332e-07,
+      "loss": 1.9522,
+      "step": 5977
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.556259132976132e-07,
+      "loss": 0.0008,
+      "step": 5978
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.5560155869459324e-07,
+      "loss": 0.0011,
+      "step": 5979
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.555772040915733e-07,
+      "loss": 0.0008,
+      "step": 5980
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.555528494885533e-07,
+      "loss": 0.0011,
+      "step": 5981
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.5552849488553335e-07,
+      "loss": 0.001,
+      "step": 5982
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 3.5550414028251334e-07,
+      "loss": 0.0012,
+      "step": 5983
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.554797856794934e-07,
+      "loss": 0.0011,
+      "step": 5984
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.554554310764734e-07,
+      "loss": 0.0007,
+      "step": 5985
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.5543107647345345e-07,
+      "loss": 0.0054,
+      "step": 5986
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.554067218704335e-07,
+      "loss": 0.0009,
+      "step": 5987
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.5538236726741355e-07,
+      "loss": 0.001,
+      "step": 5988
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.553580126643936e-07,
+      "loss": 0.0232,
+      "step": 5989
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.553336580613736e-07,
+      "loss": 0.0102,
+      "step": 5990
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.5530930345835365e-07,
+      "loss": 0.0013,
+      "step": 5991
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.5528494885533365e-07,
+      "loss": 0.001,
+      "step": 5992
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 3.552605942523137e-07,
+      "loss": 0.0015,
+      "step": 5993
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.552362396492937e-07,
+      "loss": 0.0011,
+      "step": 5994
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.5521188504627375e-07,
+      "loss": 0.0009,
+      "step": 5995
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.551875304432538e-07,
+      "loss": 0.0009,
+      "step": 5996
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.551631758402338e-07,
+      "loss": 0.001,
+      "step": 5997
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.5513882123721385e-07,
+      "loss": 0.0009,
+      "step": 5998
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.5511446663419385e-07,
+      "loss": 0.0008,
+      "step": 5999
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.550901120311739e-07,
+      "loss": 0.0008,
+      "step": 6000
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.550657574281539e-07,
+      "loss": 0.001,
+      "step": 6001
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.5504140282513395e-07,
+      "loss": 0.001,
+      "step": 6002
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.5501704822211395e-07,
+      "loss": 0.0009,
+      "step": 6003
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 3.54992693619094e-07,
+      "loss": 0.0007,
+      "step": 6004
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.5496833901607406e-07,
+      "loss": 0.0807,
+      "step": 6005
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.5494398441305406e-07,
+      "loss": 0.001,
+      "step": 6006
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.549196298100341e-07,
+      "loss": 0.0008,
+      "step": 6007
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.548952752070141e-07,
+      "loss": 0.001,
+      "step": 6008
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.5487092060399416e-07,
+      "loss": 0.0014,
+      "step": 6009
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.5484656600097416e-07,
+      "loss": 0.0009,
+      "step": 6010
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.548222113979542e-07,
+      "loss": 0.001,
+      "step": 6011
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.5479785679493426e-07,
+      "loss": 0.0009,
+      "step": 6012
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.5477350219191426e-07,
+      "loss": 0.0009,
+      "step": 6013
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 3.547491475888943e-07,
+      "loss": 0.0008,
+      "step": 6014
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.547247929858743e-07,
+      "loss": 0.0009,
+      "step": 6015
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.5470043838285436e-07,
+      "loss": 0.0023,
+      "step": 6016
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.5467608377983436e-07,
+      "loss": 0.0009,
+      "step": 6017
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.546517291768144e-07,
+      "loss": 0.0012,
+      "step": 6018
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.5462737457379446e-07,
+      "loss": 0.0008,
+      "step": 6019
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.5460301997077446e-07,
+      "loss": 0.0009,
+      "step": 6020
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.545786653677545e-07,
+      "loss": 0.0009,
+      "step": 6021
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.545543107647345e-07,
+      "loss": 0.0011,
+      "step": 6022
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.5452995616171456e-07,
+      "loss": 0.0011,
+      "step": 6023
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 3.5450560155869456e-07,
+      "loss": 0.001,
+      "step": 6024
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.544812469556746e-07,
+      "loss": 0.001,
+      "step": 6025
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.544568923526546e-07,
+      "loss": 0.0009,
+      "step": 6026
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.5443253774963466e-07,
+      "loss": 0.0008,
+      "step": 6027
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.544081831466147e-07,
+      "loss": 0.0009,
+      "step": 6028
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.543838285435947e-07,
+      "loss": 0.0009,
+      "step": 6029
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.5435947394057476e-07,
+      "loss": 0.0008,
+      "step": 6030
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.5433511933755476e-07,
+      "loss": 0.001,
+      "step": 6031
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.543107647345348e-07,
+      "loss": 0.0014,
+      "step": 6032
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.542864101315148e-07,
+      "loss": 0.0009,
+      "step": 6033
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.5426205552849487e-07,
+      "loss": 0.001,
+      "step": 6034
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 3.542377009254749e-07,
+      "loss": 0.0008,
+      "step": 6035
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.542133463224549e-07,
+      "loss": 0.0009,
+      "step": 6036
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.5418899171943497e-07,
+      "loss": 0.0009,
+      "step": 6037
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.5416463711641497e-07,
+      "loss": 0.0007,
+      "step": 6038
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.54140282513395e-07,
+      "loss": 0.001,
+      "step": 6039
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.54115927910375e-07,
+      "loss": 0.0009,
+      "step": 6040
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.5409157330735507e-07,
+      "loss": 0.0013,
+      "step": 6041
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.5406721870433507e-07,
+      "loss": 0.001,
+      "step": 6042
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.540428641013151e-07,
+      "loss": 0.001,
+      "step": 6043
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.5401850949829517e-07,
+      "loss": 0.0012,
+      "step": 6044
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 3.5399415489527517e-07,
+      "loss": 0.001,
+      "step": 6045
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.539698002922552e-07,
+      "loss": 0.001,
+      "step": 6046
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.539454456892352e-07,
+      "loss": 0.0009,
+      "step": 6047
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.5392109108621527e-07,
+      "loss": 0.0007,
+      "step": 6048
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.5389673648319527e-07,
+      "loss": 0.0009,
+      "step": 6049
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.538723818801753e-07,
+      "loss": 0.001,
+      "step": 6050
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.5384802727715537e-07,
+      "loss": 0.0007,
+      "step": 6051
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.5382367267413537e-07,
+      "loss": 0.0009,
+      "step": 6052
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.537993180711154e-07,
+      "loss": 0.0008,
+      "step": 6053
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.537749634680954e-07,
+      "loss": 0.0008,
+      "step": 6054
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 3.5375060886507547e-07,
+      "loss": 0.0011,
+      "step": 6055
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.5372625426205547e-07,
+      "loss": 0.0009,
+      "step": 6056
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.537018996590355e-07,
+      "loss": 0.0009,
+      "step": 6057
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.536775450560155e-07,
+      "loss": 0.0011,
+      "step": 6058
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.536531904529956e-07,
+      "loss": 0.0008,
+      "step": 6059
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.536288358499756e-07,
+      "loss": 0.001,
+      "step": 6060
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.536044812469556e-07,
+      "loss": 0.0025,
+      "step": 6061
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.535801266439357e-07,
+      "loss": 0.001,
+      "step": 6062
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.535557720409157e-07,
+      "loss": 0.0008,
+      "step": 6063
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.535314174378958e-07,
+      "loss": 0.0011,
+      "step": 6064
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 3.535070628348758e-07,
+      "loss": 0.0012,
+      "step": 6065
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.5348270823185583e-07,
+      "loss": 0.0009,
+      "step": 6066
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.534583536288359e-07,
+      "loss": 0.001,
+      "step": 6067
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.534339990258159e-07,
+      "loss": 0.0007,
+      "step": 6068
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.5340964442279593e-07,
+      "loss": 0.0009,
+      "step": 6069
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.5338528981977593e-07,
+      "loss": 0.8213,
+      "step": 6070
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.53360935216756e-07,
+      "loss": 0.0011,
+      "step": 6071
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.53336580613736e-07,
+      "loss": 0.001,
+      "step": 6072
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.5331222601071603e-07,
+      "loss": 0.0009,
+      "step": 6073
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.532878714076961e-07,
+      "loss": 0.0008,
+      "step": 6074
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.532635168046761e-07,
+      "loss": 0.001,
+      "step": 6075
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 3.5323916220165613e-07,
+      "loss": 0.0011,
+      "step": 6076
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.5321480759863613e-07,
+      "loss": 0.0011,
+      "step": 6077
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.531904529956162e-07,
+      "loss": 0.001,
+      "step": 6078
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.531660983925962e-07,
+      "loss": 0.0008,
+      "step": 6079
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.5314174378957623e-07,
+      "loss": 0.0013,
+      "step": 6080
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.5311738918655623e-07,
+      "loss": 1.1028,
+      "step": 6081
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.530930345835363e-07,
+      "loss": 0.001,
+      "step": 6082
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.5306867998051634e-07,
+      "loss": 0.0009,
+      "step": 6083
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.5304432537749634e-07,
+      "loss": 0.0009,
+      "step": 6084
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.530199707744764e-07,
+      "loss": 0.0033,
+      "step": 6085
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 3.529956161714564e-07,
+      "loss": 0.0009,
+      "step": 6086
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.5297126156843644e-07,
+      "loss": 0.0012,
+      "step": 6087
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.5294690696541644e-07,
+      "loss": 0.0009,
+      "step": 6088
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.529225523623965e-07,
+      "loss": 0.0009,
+      "step": 6089
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.5289819775937654e-07,
+      "loss": 0.001,
+      "step": 6090
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.5287384315635654e-07,
+      "loss": 0.001,
+      "step": 6091
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.528494885533366e-07,
+      "loss": 0.001,
+      "step": 6092
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.528251339503166e-07,
+      "loss": 0.0008,
+      "step": 6093
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.5280077934729664e-07,
+      "loss": 0.001,
+      "step": 6094
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.5277642474427664e-07,
+      "loss": 0.0007,
+      "step": 6095
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 3.527520701412567e-07,
+      "loss": 0.0007,
+      "step": 6096
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.527277155382367e-07,
+      "loss": 0.001,
+      "step": 6097
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.5270336093521674e-07,
+      "loss": 0.001,
+      "step": 6098
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.526790063321968e-07,
+      "loss": 0.0009,
+      "step": 6099
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.526546517291768e-07,
+      "loss": 0.0011,
+      "step": 6100
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.5263029712615684e-07,
+      "loss": 0.0009,
+      "step": 6101
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.5260594252313684e-07,
+      "loss": 0.0008,
+      "step": 6102
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.525815879201169e-07,
+      "loss": 0.0009,
+      "step": 6103
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.525572333170969e-07,
+      "loss": 0.0008,
+      "step": 6104
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.5253287871407694e-07,
+      "loss": 0.0008,
+      "step": 6105
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.52508524111057e-07,
+      "loss": 0.0008,
+      "step": 6106
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 3.52484169508037e-07,
+      "loss": 0.001,
+      "step": 6107
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.5245981490501705e-07,
+      "loss": 0.0009,
+      "step": 6108
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.5243546030199704e-07,
+      "loss": 0.0084,
+      "step": 6109
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.524111056989771e-07,
+      "loss": 0.0016,
+      "step": 6110
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.523867510959571e-07,
+      "loss": 0.0009,
+      "step": 6111
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.5236239649293715e-07,
+      "loss": 0.0011,
+      "step": 6112
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.5233804188991714e-07,
+      "loss": 0.0012,
+      "step": 6113
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.523136872868972e-07,
+      "loss": 0.0008,
+      "step": 6114
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.5228933268387725e-07,
+      "loss": 0.0009,
+      "step": 6115
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.5226497808085725e-07,
+      "loss": 0.0019,
+      "step": 6116
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 3.522406234778373e-07,
+      "loss": 0.001,
+      "step": 6117
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.522162688748173e-07,
+      "loss": 0.001,
+      "step": 6118
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.5219191427179735e-07,
+      "loss": 0.0009,
+      "step": 6119
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.5216755966877735e-07,
+      "loss": 0.001,
+      "step": 6120
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.521432050657574e-07,
+      "loss": 0.0008,
+      "step": 6121
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.5211885046273745e-07,
+      "loss": 0.001,
+      "step": 6122
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.5209449585971745e-07,
+      "loss": 0.0008,
+      "step": 6123
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.520701412566975e-07,
+      "loss": 0.0009,
+      "step": 6124
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.520457866536775e-07,
+      "loss": 0.001,
+      "step": 6125
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.5202143205065755e-07,
+      "loss": 0.0012,
+      "step": 6126
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 3.5199707744763755e-07,
+      "loss": 0.0008,
+      "step": 6127
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.519727228446176e-07,
+      "loss": 0.0009,
+      "step": 6128
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.519483682415976e-07,
+      "loss": 0.0008,
+      "step": 6129
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.5192401363857765e-07,
+      "loss": 0.0009,
+      "step": 6130
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.518996590355577e-07,
+      "loss": 0.0011,
+      "step": 6131
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.518753044325377e-07,
+      "loss": 0.0009,
+      "step": 6132
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.5185094982951775e-07,
+      "loss": 0.0014,
+      "step": 6133
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.5182659522649775e-07,
+      "loss": 0.0022,
+      "step": 6134
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.518022406234778e-07,
+      "loss": 0.0012,
+      "step": 6135
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.517778860204578e-07,
+      "loss": 0.0011,
+      "step": 6136
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 3.5175353141743785e-07,
+      "loss": 0.0009,
+      "step": 6137
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.517291768144179e-07,
+      "loss": 0.0007,
+      "step": 6138
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.517048222113979e-07,
+      "loss": 0.0017,
+      "step": 6139
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.5168046760837796e-07,
+      "loss": 0.001,
+      "step": 6140
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.51656113005358e-07,
+      "loss": 0.0007,
+      "step": 6141
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.5163175840233806e-07,
+      "loss": 0.0009,
+      "step": 6142
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.5160740379931806e-07,
+      "loss": 0.0009,
+      "step": 6143
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.515830491962981e-07,
+      "loss": 0.0009,
+      "step": 6144
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.5155869459327816e-07,
+      "loss": 0.0012,
+      "step": 6145
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.5153433999025816e-07,
+      "loss": 0.0009,
+      "step": 6146
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.515099853872382e-07,
+      "loss": 1.526,
+      "step": 6147
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 3.514856307842182e-07,
+      "loss": 0.0008,
+      "step": 6148
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.5146127618119826e-07,
+      "loss": 0.0007,
+      "step": 6149
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.5143692157817826e-07,
+      "loss": 0.0007,
+      "step": 6150
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.514125669751583e-07,
+      "loss": 0.0009,
+      "step": 6151
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.513882123721383e-07,
+      "loss": 0.0008,
+      "step": 6152
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.5136385776911836e-07,
+      "loss": 0.0008,
+      "step": 6153
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.513395031660984e-07,
+      "loss": 0.0013,
+      "step": 6154
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.513151485630784e-07,
+      "loss": 0.0008,
+      "step": 6155
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.5129079396005847e-07,
+      "loss": 0.0009,
+      "step": 6156
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.5126643935703846e-07,
+      "loss": 0.0008,
+      "step": 6157
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 3.512420847540185e-07,
+      "loss": 0.0008,
+      "step": 6158
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.512177301509985e-07,
+      "loss": 0.0009,
+      "step": 6159
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.5119337554797857e-07,
+      "loss": 0.001,
+      "step": 6160
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.511690209449586e-07,
+      "loss": 0.001,
+      "step": 6161
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.511446663419386e-07,
+      "loss": 0.001,
+      "step": 6162
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.5112031173891867e-07,
+      "loss": 0.0026,
+      "step": 6163
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.5109595713589867e-07,
+      "loss": 0.0008,
+      "step": 6164
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.510716025328787e-07,
+      "loss": 0.0007,
+      "step": 6165
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.510472479298587e-07,
+      "loss": 0.1475,
+      "step": 6166
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.5102289332683877e-07,
+      "loss": 0.0011,
+      "step": 6167
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 3.5099853872381877e-07,
+      "loss": 0.0011,
+      "step": 6168
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.509741841207988e-07,
+      "loss": 0.0013,
+      "step": 6169
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.5094982951777887e-07,
+      "loss": 0.001,
+      "step": 6170
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.5092547491475887e-07,
+      "loss": 0.0016,
+      "step": 6171
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.509011203117389e-07,
+      "loss": 0.0006,
+      "step": 6172
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.508767657087189e-07,
+      "loss": 0.0011,
+      "step": 6173
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.5085241110569897e-07,
+      "loss": 0.0008,
+      "step": 6174
+    },
+    {
+      "epoch": 6.0,
+      "eval_accuracy": 0.9980554205153136,
+      "eval_loss": 0.01545155979692936,
+      "eval_runtime": 2.7932,
+      "eval_samples_per_second": 736.427,
+      "eval_steps_per_second": 11.814,
+      "step": 6174
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.5082805650267897e-07,
+      "loss": 0.001,
+      "step": 6175
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.50803701899659e-07,
+      "loss": 0.0008,
+      "step": 6176
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.5077934729663907e-07,
+      "loss": 0.0008,
+      "step": 6177
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.5075499269361907e-07,
+      "loss": 0.0008,
+      "step": 6178
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 3.507306380905991e-07,
+      "loss": 0.0013,
+      "step": 6179
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.507062834875791e-07,
+      "loss": 0.0027,
+      "step": 6180
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.506819288845592e-07,
+      "loss": 0.0009,
+      "step": 6181
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.5065757428153917e-07,
+      "loss": 0.0008,
+      "step": 6182
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.506332196785192e-07,
+      "loss": 0.0014,
+      "step": 6183
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.506088650754993e-07,
+      "loss": 0.0015,
+      "step": 6184
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.505845104724793e-07,
+      "loss": 0.0009,
+      "step": 6185
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.505601558694593e-07,
+      "loss": 0.0012,
+      "step": 6186
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.505358012664393e-07,
+      "loss": 0.0012,
+      "step": 6187
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.505114466634194e-07,
+      "loss": 0.0008,
+      "step": 6188
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 3.504870920603994e-07,
+      "loss": 0.0007,
+      "step": 6189
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.5046273745737943e-07,
+      "loss": 0.0009,
+      "step": 6190
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.504383828543594e-07,
+      "loss": 0.0009,
+      "step": 6191
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.504140282513395e-07,
+      "loss": 0.0012,
+      "step": 6192
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.5038967364831953e-07,
+      "loss": 0.0007,
+      "step": 6193
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.5036531904529953e-07,
+      "loss": 0.001,
+      "step": 6194
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.503409644422796e-07,
+      "loss": 0.0008,
+      "step": 6195
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.503166098392596e-07,
+      "loss": 0.8901,
+      "step": 6196
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.5029225523623963e-07,
+      "loss": 0.0008,
+      "step": 6197
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.5026790063321963e-07,
+      "loss": 0.0009,
+      "step": 6198
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 3.502435460301997e-07,
+      "loss": 0.0009,
+      "step": 6199
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.5021919142717973e-07,
+      "loss": 0.001,
+      "step": 6200
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.5019483682415973e-07,
+      "loss": 0.0008,
+      "step": 6201
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.501704822211398e-07,
+      "loss": 0.0011,
+      "step": 6202
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.501461276181198e-07,
+      "loss": 0.0011,
+      "step": 6203
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.5012177301509983e-07,
+      "loss": 0.0007,
+      "step": 6204
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.5009741841207983e-07,
+      "loss": 0.0009,
+      "step": 6205
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.500730638090599e-07,
+      "loss": 0.0009,
+      "step": 6206
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.500487092060399e-07,
+      "loss": 0.0011,
+      "step": 6207
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.5002435460301993e-07,
+      "loss": 0.0009,
+      "step": 6208
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.5e-07,
+      "loss": 0.0012,
+      "step": 6209
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 3.4997564539698e-07,
+      "loss": 0.0014,
+      "step": 6210
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.4995129079396003e-07,
+      "loss": 0.0009,
+      "step": 6211
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.4992693619094003e-07,
+      "loss": 0.0017,
+      "step": 6212
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.499025815879201e-07,
+      "loss": 0.0012,
+      "step": 6213
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.498782269849001e-07,
+      "loss": 0.001,
+      "step": 6214
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.4985387238188014e-07,
+      "loss": 0.0009,
+      "step": 6215
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.498295177788602e-07,
+      "loss": 0.0016,
+      "step": 6216
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.4980516317584024e-07,
+      "loss": 0.0009,
+      "step": 6217
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.497808085728203e-07,
+      "loss": 0.001,
+      "step": 6218
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.497564539698003e-07,
+      "loss": 0.0007,
+      "step": 6219
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 3.4973209936678034e-07,
+      "loss": 0.0009,
+      "step": 6220
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.4970774476376034e-07,
+      "loss": 0.0008,
+      "step": 6221
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.496833901607404e-07,
+      "loss": 0.0008,
+      "step": 6222
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.4965903555772044e-07,
+      "loss": 0.0009,
+      "step": 6223
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.4963468095470044e-07,
+      "loss": 0.0023,
+      "step": 6224
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.496103263516805e-07,
+      "loss": 0.0009,
+      "step": 6225
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.495859717486605e-07,
+      "loss": 0.001,
+      "step": 6226
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.4956161714564054e-07,
+      "loss": 0.001,
+      "step": 6227
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.4953726254262054e-07,
+      "loss": 0.0009,
+      "step": 6228
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.495129079396006e-07,
+      "loss": 0.001,
+      "step": 6229
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 3.494885533365806e-07,
+      "loss": 0.0012,
+      "step": 6230
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.4946419873356064e-07,
+      "loss": 0.0008,
+      "step": 6231
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.494398441305407e-07,
+      "loss": 0.0008,
+      "step": 6232
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.494154895275207e-07,
+      "loss": 0.001,
+      "step": 6233
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.4939113492450075e-07,
+      "loss": 1.5433,
+      "step": 6234
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.4936678032148074e-07,
+      "loss": 0.0009,
+      "step": 6235
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.493424257184608e-07,
+      "loss": 0.0014,
+      "step": 6236
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.493180711154408e-07,
+      "loss": 0.001,
+      "step": 6237
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.4929371651242085e-07,
+      "loss": 0.0008,
+      "step": 6238
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.492693619094009e-07,
+      "loss": 0.0008,
+      "step": 6239
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 3.492450073063809e-07,
+      "loss": 0.0008,
+      "step": 6240
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.4922065270336095e-07,
+      "loss": 0.0011,
+      "step": 6241
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.4919629810034095e-07,
+      "loss": 0.0013,
+      "step": 6242
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.49171943497321e-07,
+      "loss": 0.0007,
+      "step": 6243
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.49147588894301e-07,
+      "loss": 0.0008,
+      "step": 6244
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.4912323429128105e-07,
+      "loss": 0.0008,
+      "step": 6245
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.4909887968826105e-07,
+      "loss": 2.1122,
+      "step": 6246
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.490745250852411e-07,
+      "loss": 0.0007,
+      "step": 6247
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.4905017048222115e-07,
+      "loss": 0.0009,
+      "step": 6248
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.4902581587920115e-07,
+      "loss": 0.0011,
+      "step": 6249
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.490014612761812e-07,
+      "loss": 0.0008,
+      "step": 6250
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 3.489771066731612e-07,
+      "loss": 0.0009,
+      "step": 6251
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.4895275207014125e-07,
+      "loss": 0.0008,
+      "step": 6252
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.4892839746712125e-07,
+      "loss": 0.0007,
+      "step": 6253
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.489040428641013e-07,
+      "loss": 0.0008,
+      "step": 6254
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.4887968826108135e-07,
+      "loss": 0.0009,
+      "step": 6255
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.4885533365806135e-07,
+      "loss": 0.001,
+      "step": 6256
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.488309790550414e-07,
+      "loss": 0.0013,
+      "step": 6257
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.488066244520214e-07,
+      "loss": 0.0008,
+      "step": 6258
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.4878226984900145e-07,
+      "loss": 0.001,
+      "step": 6259
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.4875791524598145e-07,
+      "loss": 0.0013,
+      "step": 6260
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 3.487335606429615e-07,
+      "loss": 0.0009,
+      "step": 6261
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.487092060399415e-07,
+      "loss": 0.0008,
+      "step": 6262
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.4868485143692155e-07,
+      "loss": 0.0007,
+      "step": 6263
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.486604968339016e-07,
+      "loss": 1.7974,
+      "step": 6264
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.486361422308816e-07,
+      "loss": 0.001,
+      "step": 6265
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.4861178762786166e-07,
+      "loss": 0.001,
+      "step": 6266
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.4858743302484166e-07,
+      "loss": 0.001,
+      "step": 6267
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.485630784218217e-07,
+      "loss": 0.001,
+      "step": 6268
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.485387238188017e-07,
+      "loss": 0.0007,
+      "step": 6269
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.4851436921578176e-07,
+      "loss": 0.0009,
+      "step": 6270
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 3.484900146127618e-07,
+      "loss": 0.0012,
+      "step": 6271
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.484656600097418e-07,
+      "loss": 0.0008,
+      "step": 6272
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.4844130540672186e-07,
+      "loss": 0.0011,
+      "step": 6273
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.4841695080370186e-07,
+      "loss": 0.0009,
+      "step": 6274
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.483925962006819e-07,
+      "loss": 0.001,
+      "step": 6275
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.483682415976619e-07,
+      "loss": 0.0009,
+      "step": 6276
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.4834388699464196e-07,
+      "loss": 0.0009,
+      "step": 6277
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.4831953239162196e-07,
+      "loss": 0.001,
+      "step": 6278
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.48295177788602e-07,
+      "loss": 0.0008,
+      "step": 6279
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.4827082318558206e-07,
+      "loss": 0.001,
+      "step": 6280
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.4824646858256206e-07,
+      "loss": 0.001,
+      "step": 6281
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 3.482221139795421e-07,
+      "loss": 0.0009,
+      "step": 6282
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.481977593765221e-07,
+      "loss": 0.001,
+      "step": 6283
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.4817340477350216e-07,
+      "loss": 0.0017,
+      "step": 6284
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.4814905017048216e-07,
+      "loss": 0.001,
+      "step": 6285
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.481246955674622e-07,
+      "loss": 0.0029,
+      "step": 6286
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.4810034096444226e-07,
+      "loss": 0.0011,
+      "step": 6287
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.4807598636142226e-07,
+      "loss": 0.001,
+      "step": 6288
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.480516317584023e-07,
+      "loss": 0.0008,
+      "step": 6289
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.480272771553823e-07,
+      "loss": 0.0008,
+      "step": 6290
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.4800292255236237e-07,
+      "loss": 0.001,
+      "step": 6291
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 3.4797856794934236e-07,
+      "loss": 0.0007,
+      "step": 6292
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.479542133463224e-07,
+      "loss": 0.4556,
+      "step": 6293
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.479298587433025e-07,
+      "loss": 0.0009,
+      "step": 6294
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.479055041402825e-07,
+      "loss": 0.0008,
+      "step": 6295
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.4788114953726257e-07,
+      "loss": 0.0009,
+      "step": 6296
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.4785679493424257e-07,
+      "loss": 0.0009,
+      "step": 6297
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.478324403312226e-07,
+      "loss": 0.0007,
+      "step": 6298
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.478080857282026e-07,
+      "loss": 0.0011,
+      "step": 6299
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.4778373112518267e-07,
+      "loss": 0.0013,
+      "step": 6300
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.4775937652216267e-07,
+      "loss": 0.0008,
+      "step": 6301
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 3.477350219191427e-07,
+      "loss": 0.0013,
+      "step": 6302
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.4771066731612277e-07,
+      "loss": 0.001,
+      "step": 6303
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.4768631271310277e-07,
+      "loss": 0.0006,
+      "step": 6304
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.476619581100828e-07,
+      "loss": 0.0009,
+      "step": 6305
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.476376035070628e-07,
+      "loss": 0.0009,
+      "step": 6306
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.476132489040429e-07,
+      "loss": 0.0008,
+      "step": 6307
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.4758889430102287e-07,
+      "loss": 0.0009,
+      "step": 6308
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.475645396980029e-07,
+      "loss": 0.0007,
+      "step": 6309
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.47540185094983e-07,
+      "loss": 0.0927,
+      "step": 6310
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.47515830491963e-07,
+      "loss": 0.0009,
+      "step": 6311
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 3.47491475888943e-07,
+      "loss": 0.0007,
+      "step": 6312
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.47467121285923e-07,
+      "loss": 0.0007,
+      "step": 6313
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.474427666829031e-07,
+      "loss": 0.0009,
+      "step": 6314
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.474184120798831e-07,
+      "loss": 0.001,
+      "step": 6315
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.4739405747686313e-07,
+      "loss": 0.0011,
+      "step": 6316
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.473697028738431e-07,
+      "loss": 0.0008,
+      "step": 6317
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.473453482708232e-07,
+      "loss": 0.0008,
+      "step": 6318
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.4732099366780323e-07,
+      "loss": 0.0008,
+      "step": 6319
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.4729663906478323e-07,
+      "loss": 0.0009,
+      "step": 6320
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.472722844617633e-07,
+      "loss": 0.0007,
+      "step": 6321
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.472479298587433e-07,
+      "loss": 0.0013,
+      "step": 6322
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 3.4722357525572333e-07,
+      "loss": 0.0009,
+      "step": 6323
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.4719922065270333e-07,
+      "loss": 0.0013,
+      "step": 6324
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.471748660496834e-07,
+      "loss": 0.0008,
+      "step": 6325
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.4715051144666343e-07,
+      "loss": 0.0007,
+      "step": 6326
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.4712615684364343e-07,
+      "loss": 0.0012,
+      "step": 6327
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.471018022406235e-07,
+      "loss": 0.0011,
+      "step": 6328
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.470774476376035e-07,
+      "loss": 0.0006,
+      "step": 6329
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.4705309303458353e-07,
+      "loss": 0.0009,
+      "step": 6330
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.4702873843156353e-07,
+      "loss": 0.0007,
+      "step": 6331
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.470043838285436e-07,
+      "loss": 0.0009,
+      "step": 6332
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 3.469800292255236e-07,
+      "loss": 0.0008,
+      "step": 6333
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.4695567462250363e-07,
+      "loss": 0.0009,
+      "step": 6334
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.469313200194837e-07,
+      "loss": 0.0009,
+      "step": 6335
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.469069654164637e-07,
+      "loss": 0.0008,
+      "step": 6336
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.4688261081344373e-07,
+      "loss": 0.0009,
+      "step": 6337
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.4685825621042373e-07,
+      "loss": 0.0009,
+      "step": 6338
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.468339016074038e-07,
+      "loss": 0.0008,
+      "step": 6339
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.468095470043838e-07,
+      "loss": 0.0009,
+      "step": 6340
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.4678519240136384e-07,
+      "loss": 0.0008,
+      "step": 6341
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.467608377983439e-07,
+      "loss": 0.0009,
+      "step": 6342
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 3.467364831953239e-07,
+      "loss": 0.0011,
+      "step": 6343
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.4671212859230394e-07,
+      "loss": 0.0008,
+      "step": 6344
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.4668777398928394e-07,
+      "loss": 0.001,
+      "step": 6345
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.46663419386264e-07,
+      "loss": 0.001,
+      "step": 6346
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.46639064783244e-07,
+      "loss": 0.0009,
+      "step": 6347
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.4661471018022404e-07,
+      "loss": 0.0012,
+      "step": 6348
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.465903555772041e-07,
+      "loss": 0.0007,
+      "step": 6349
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.465660009741841e-07,
+      "loss": 0.0007,
+      "step": 6350
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.4654164637116414e-07,
+      "loss": 0.001,
+      "step": 6351
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.4651729176814414e-07,
+      "loss": 0.0008,
+      "step": 6352
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.464929371651242e-07,
+      "loss": 0.0011,
+      "step": 6353
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 3.464685825621042e-07,
+      "loss": 0.0009,
+      "step": 6354
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.4644422795908424e-07,
+      "loss": 0.001,
+      "step": 6355
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.4641987335606424e-07,
+      "loss": 0.0009,
+      "step": 6356
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.463955187530443e-07,
+      "loss": 0.0009,
+      "step": 6357
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.4637116415002434e-07,
+      "loss": 0.0008,
+      "step": 6358
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.4634680954700434e-07,
+      "loss": 0.001,
+      "step": 6359
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.463224549439844e-07,
+      "loss": 0.0009,
+      "step": 6360
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.462981003409644e-07,
+      "loss": 0.0007,
+      "step": 6361
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.4627374573794444e-07,
+      "loss": 0.0008,
+      "step": 6362
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.4624939113492444e-07,
+      "loss": 0.0012,
+      "step": 6363
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 3.462250365319045e-07,
+      "loss": 0.0007,
+      "step": 6364
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.4620068192888455e-07,
+      "loss": 0.002,
+      "step": 6365
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.4617632732586454e-07,
+      "loss": 0.0008,
+      "step": 6366
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.461519727228446e-07,
+      "loss": 0.0008,
+      "step": 6367
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.461276181198246e-07,
+      "loss": 0.0012,
+      "step": 6368
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.4610326351680465e-07,
+      "loss": 0.9029,
+      "step": 6369
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.460789089137847e-07,
+      "loss": 0.0073,
+      "step": 6370
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.4605455431076475e-07,
+      "loss": 0.0009,
+      "step": 6371
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.4603019970774475e-07,
+      "loss": 0.0008,
+      "step": 6372
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.460058451047248e-07,
+      "loss": 0.0008,
+      "step": 6373
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 3.4598149050170485e-07,
+      "loss": 0.0007,
+      "step": 6374
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.4595713589868485e-07,
+      "loss": 0.0007,
+      "step": 6375
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.459327812956649e-07,
+      "loss": 0.0008,
+      "step": 6376
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.459084266926449e-07,
+      "loss": 0.001,
+      "step": 6377
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.4588407208962495e-07,
+      "loss": 0.0012,
+      "step": 6378
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.4585971748660495e-07,
+      "loss": 0.001,
+      "step": 6379
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.45835362883585e-07,
+      "loss": 0.0011,
+      "step": 6380
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.4581100828056505e-07,
+      "loss": 0.0009,
+      "step": 6381
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.4578665367754505e-07,
+      "loss": 0.0009,
+      "step": 6382
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.457622990745251e-07,
+      "loss": 0.0009,
+      "step": 6383
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 3.457379444715051e-07,
+      "loss": 0.0009,
+      "step": 6384
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.4571358986848515e-07,
+      "loss": 1.1881,
+      "step": 6385
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.4568923526546515e-07,
+      "loss": 0.0007,
+      "step": 6386
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.456648806624452e-07,
+      "loss": 0.0009,
+      "step": 6387
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.4564052605942526e-07,
+      "loss": 0.0008,
+      "step": 6388
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.4561617145640526e-07,
+      "loss": 0.0007,
+      "step": 6389
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.455918168533853e-07,
+      "loss": 0.0009,
+      "step": 6390
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.455674622503653e-07,
+      "loss": 0.0012,
+      "step": 6391
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.4554310764734536e-07,
+      "loss": 0.0009,
+      "step": 6392
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.4551875304432536e-07,
+      "loss": 0.0011,
+      "step": 6393
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.454943984413054e-07,
+      "loss": 0.0011,
+      "step": 6394
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 3.454700438382854e-07,
+      "loss": 0.0008,
+      "step": 6395
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.4544568923526546e-07,
+      "loss": 0.0009,
+      "step": 6396
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.454213346322455e-07,
+      "loss": 0.0009,
+      "step": 6397
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.453969800292255e-07,
+      "loss": 0.0011,
+      "step": 6398
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.4537262542620556e-07,
+      "loss": 0.0009,
+      "step": 6399
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.4534827082318556e-07,
+      "loss": 0.0009,
+      "step": 6400
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.453239162201656e-07,
+      "loss": 0.0009,
+      "step": 6401
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.452995616171456e-07,
+      "loss": 0.0008,
+      "step": 6402
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.4527520701412566e-07,
+      "loss": 0.0008,
+      "step": 6403
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.452508524111057e-07,
+      "loss": 0.0019,
+      "step": 6404
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 3.452264978080857e-07,
+      "loss": 0.0008,
+      "step": 6405
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.4520214320506576e-07,
+      "loss": 0.0008,
+      "step": 6406
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.4517778860204576e-07,
+      "loss": 0.0012,
+      "step": 6407
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.451534339990258e-07,
+      "loss": 0.001,
+      "step": 6408
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.451290793960058e-07,
+      "loss": 0.0008,
+      "step": 6409
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.4510472479298586e-07,
+      "loss": 0.0009,
+      "step": 6410
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.4508037018996586e-07,
+      "loss": 0.0008,
+      "step": 6411
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.450560155869459e-07,
+      "loss": 0.001,
+      "step": 6412
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.4503166098392596e-07,
+      "loss": 0.0008,
+      "step": 6413
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.4500730638090596e-07,
+      "loss": 0.0007,
+      "step": 6414
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 3.44982951777886e-07,
+      "loss": 0.0009,
+      "step": 6415
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.44958597174866e-07,
+      "loss": 0.0008,
+      "step": 6416
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.4493424257184607e-07,
+      "loss": 0.0008,
+      "step": 6417
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.4490988796882606e-07,
+      "loss": 0.0008,
+      "step": 6418
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.448855333658061e-07,
+      "loss": 0.0009,
+      "step": 6419
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.4486117876278617e-07,
+      "loss": 0.0007,
+      "step": 6420
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.4483682415976617e-07,
+      "loss": 0.0007,
+      "step": 6421
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.448124695567462e-07,
+      "loss": 0.0009,
+      "step": 6422
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.447881149537262e-07,
+      "loss": 0.001,
+      "step": 6423
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.4476376035070627e-07,
+      "loss": 0.0008,
+      "step": 6424
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.4473940574768627e-07,
+      "loss": 0.0016,
+      "step": 6425
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 3.447150511446663e-07,
+      "loss": 0.0007,
+      "step": 6426
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.446906965416463e-07,
+      "loss": 0.001,
+      "step": 6427
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.4466634193862637e-07,
+      "loss": 0.0008,
+      "step": 6428
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.446419873356064e-07,
+      "loss": 0.0007,
+      "step": 6429
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.446176327325864e-07,
+      "loss": 0.0009,
+      "step": 6430
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.4459327812956647e-07,
+      "loss": 0.001,
+      "step": 6431
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.4456892352654647e-07,
+      "loss": 0.0009,
+      "step": 6432
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.445445689235265e-07,
+      "loss": 0.0009,
+      "step": 6433
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.445202143205065e-07,
+      "loss": 0.0013,
+      "step": 6434
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.4449585971748657e-07,
+      "loss": 0.0007,
+      "step": 6435
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 3.444715051144666e-07,
+      "loss": 0.0023,
+      "step": 6436
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.444471505114466e-07,
+      "loss": 0.0016,
+      "step": 6437
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.4442279590842667e-07,
+      "loss": 0.0007,
+      "step": 6438
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.4439844130540667e-07,
+      "loss": 0.0007,
+      "step": 6439
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.443740867023867e-07,
+      "loss": 0.0014,
+      "step": 6440
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.443497320993667e-07,
+      "loss": 0.0008,
+      "step": 6441
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.443253774963468e-07,
+      "loss": 0.0008,
+      "step": 6442
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.4430102289332677e-07,
+      "loss": 0.0011,
+      "step": 6443
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.442766682903068e-07,
+      "loss": 0.0009,
+      "step": 6444
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.442523136872869e-07,
+      "loss": 0.001,
+      "step": 6445
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 3.442279590842669e-07,
+      "loss": 0.0009,
+      "step": 6446
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.44203604481247e-07,
+      "loss": 0.0009,
+      "step": 6447
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.44179249878227e-07,
+      "loss": 0.0008,
+      "step": 6448
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.4415489527520703e-07,
+      "loss": 0.0007,
+      "step": 6449
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.4413054067218703e-07,
+      "loss": 0.001,
+      "step": 6450
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.441061860691671e-07,
+      "loss": 0.0009,
+      "step": 6451
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.4408183146614713e-07,
+      "loss": 0.0007,
+      "step": 6452
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.4405747686312713e-07,
+      "loss": 0.0008,
+      "step": 6453
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.440331222601072e-07,
+      "loss": 0.001,
+      "step": 6454
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.440087676570872e-07,
+      "loss": 0.0008,
+      "step": 6455
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 3.4398441305406723e-07,
+      "loss": 0.0008,
+      "step": 6456
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.4396005845104723e-07,
+      "loss": 0.0008,
+      "step": 6457
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.439357038480273e-07,
+      "loss": 0.0174,
+      "step": 6458
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.4391134924500733e-07,
+      "loss": 0.0009,
+      "step": 6459
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.4388699464198733e-07,
+      "loss": 0.0009,
+      "step": 6460
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.438626400389674e-07,
+      "loss": 0.001,
+      "step": 6461
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.438382854359474e-07,
+      "loss": 0.0008,
+      "step": 6462
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.4381393083292743e-07,
+      "loss": 0.0007,
+      "step": 6463
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.4378957622990743e-07,
+      "loss": 0.0011,
+      "step": 6464
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.437652216268875e-07,
+      "loss": 0.0009,
+      "step": 6465
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.437408670238675e-07,
+      "loss": 0.0009,
+      "step": 6466
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 3.4371651242084754e-07,
+      "loss": 0.0009,
+      "step": 6467
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.436921578178276e-07,
+      "loss": 0.0009,
+      "step": 6468
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.436678032148076e-07,
+      "loss": 0.0007,
+      "step": 6469
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.4364344861178764e-07,
+      "loss": 0.0009,
+      "step": 6470
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.4361909400876764e-07,
+      "loss": 0.0008,
+      "step": 6471
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.435947394057477e-07,
+      "loss": 0.0008,
+      "step": 6472
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.435703848027277e-07,
+      "loss": 0.001,
+      "step": 6473
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.4354603019970774e-07,
+      "loss": 0.0007,
+      "step": 6474
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.435216755966878e-07,
+      "loss": 0.0009,
+      "step": 6475
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.434973209936678e-07,
+      "loss": 0.0008,
+      "step": 6476
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 3.4347296639064784e-07,
+      "loss": 0.0008,
+      "step": 6477
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.4344861178762784e-07,
+      "loss": 0.0009,
+      "step": 6478
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.434242571846079e-07,
+      "loss": 0.0007,
+      "step": 6479
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.433999025815879e-07,
+      "loss": 0.0008,
+      "step": 6480
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.4337554797856794e-07,
+      "loss": 0.0008,
+      "step": 6481
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.4335119337554794e-07,
+      "loss": 0.0007,
+      "step": 6482
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.43326838772528e-07,
+      "loss": 0.0008,
+      "step": 6483
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.4330248416950804e-07,
+      "loss": 0.0009,
+      "step": 6484
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.4327812956648804e-07,
+      "loss": 0.001,
+      "step": 6485
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.432537749634681e-07,
+      "loss": 0.0009,
+      "step": 6486
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 3.432294203604481e-07,
+      "loss": 0.0006,
+      "step": 6487
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.4320506575742814e-07,
+      "loss": 0.0049,
+      "step": 6488
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.4318071115440814e-07,
+      "loss": 0.0029,
+      "step": 6489
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.431563565513882e-07,
+      "loss": 0.0007,
+      "step": 6490
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.4313200194836825e-07,
+      "loss": 0.0009,
+      "step": 6491
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.4310764734534824e-07,
+      "loss": 0.0008,
+      "step": 6492
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.430832927423283e-07,
+      "loss": 0.0007,
+      "step": 6493
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.430589381393083e-07,
+      "loss": 0.001,
+      "step": 6494
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.4303458353628835e-07,
+      "loss": 0.0009,
+      "step": 6495
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.4301022893326834e-07,
+      "loss": 0.0007,
+      "step": 6496
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.429858743302484e-07,
+      "loss": 0.0014,
+      "step": 6497
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 3.429615197272284e-07,
+      "loss": 0.0009,
+      "step": 6498
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.4293716512420845e-07,
+      "loss": 1.4096,
+      "step": 6499
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.429128105211885e-07,
+      "loss": 0.0007,
+      "step": 6500
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.428884559181685e-07,
+      "loss": 0.0009,
+      "step": 6501
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.4286410131514855e-07,
+      "loss": 0.0006,
+      "step": 6502
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.4283974671212855e-07,
+      "loss": 1.4006,
+      "step": 6503
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.428153921091086e-07,
+      "loss": 0.0009,
+      "step": 6504
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.427910375060886e-07,
+      "loss": 0.0008,
+      "step": 6505
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.4276668290306865e-07,
+      "loss": 0.0009,
+      "step": 6506
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.427423283000487e-07,
+      "loss": 0.0008,
+      "step": 6507
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 3.427179736970287e-07,
+      "loss": 0.0009,
+      "step": 6508
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.4269361909400875e-07,
+      "loss": 0.0008,
+      "step": 6509
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.4266926449098875e-07,
+      "loss": 0.0009,
+      "step": 6510
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.426449098879688e-07,
+      "loss": 0.0009,
+      "step": 6511
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.426205552849488e-07,
+      "loss": 0.0007,
+      "step": 6512
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.4259620068192885e-07,
+      "loss": 0.0007,
+      "step": 6513
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.4257184607890885e-07,
+      "loss": 0.0007,
+      "step": 6514
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.425474914758889e-07,
+      "loss": 0.0008,
+      "step": 6515
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.4252313687286895e-07,
+      "loss": 0.0009,
+      "step": 6516
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.4249878226984895e-07,
+      "loss": 0.0008,
+      "step": 6517
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 3.42474427666829e-07,
+      "loss": 0.0007,
+      "step": 6518
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.42450073063809e-07,
+      "loss": 0.0013,
+      "step": 6519
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.4242571846078905e-07,
+      "loss": 0.0008,
+      "step": 6520
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.4240136385776905e-07,
+      "loss": 0.001,
+      "step": 6521
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.423770092547491e-07,
+      "loss": 0.0022,
+      "step": 6522
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.4235265465172916e-07,
+      "loss": 0.0015,
+      "step": 6523
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.423283000487092e-07,
+      "loss": 0.0013,
+      "step": 6524
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.4230394544568926e-07,
+      "loss": 0.0022,
+      "step": 6525
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.4227959084266926e-07,
+      "loss": 0.0006,
+      "step": 6526
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.422552362396493e-07,
+      "loss": 0.0007,
+      "step": 6527
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.422308816366293e-07,
+      "loss": 0.0009,
+      "step": 6528
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 3.4220652703360936e-07,
+      "loss": 0.0011,
+      "step": 6529
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.421821724305894e-07,
+      "loss": 0.0008,
+      "step": 6530
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.421578178275694e-07,
+      "loss": 0.0009,
+      "step": 6531
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.4213346322454946e-07,
+      "loss": 0.0008,
+      "step": 6532
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.4210910862152946e-07,
+      "loss": 0.0008,
+      "step": 6533
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.420847540185095e-07,
+      "loss": 0.0009,
+      "step": 6534
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.420603994154895e-07,
+      "loss": 0.0007,
+      "step": 6535
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.4203604481246956e-07,
+      "loss": 0.0008,
+      "step": 6536
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.4201169020944956e-07,
+      "loss": 0.4101,
+      "step": 6537
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.419873356064296e-07,
+      "loss": 0.0006,
+      "step": 6538
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 3.4196298100340967e-07,
+      "loss": 0.0008,
+      "step": 6539
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.4193862640038966e-07,
+      "loss": 0.0006,
+      "step": 6540
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.419142717973697e-07,
+      "loss": 0.0011,
+      "step": 6541
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.418899171943497e-07,
+      "loss": 0.0009,
+      "step": 6542
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.4186556259132977e-07,
+      "loss": 0.0011,
+      "step": 6543
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.4184120798830976e-07,
+      "loss": 0.0006,
+      "step": 6544
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.418168533852898e-07,
+      "loss": 0.0014,
+      "step": 6545
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.4179249878226987e-07,
+      "loss": 0.0009,
+      "step": 6546
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.4176814417924987e-07,
+      "loss": 0.0007,
+      "step": 6547
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.417437895762299e-07,
+      "loss": 0.0006,
+      "step": 6548
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 3.417194349732099e-07,
+      "loss": 0.0008,
+      "step": 6549
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.4169508037018997e-07,
+      "loss": 0.0008,
+      "step": 6550
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.4167072576716997e-07,
+      "loss": 0.0012,
+      "step": 6551
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.4164637116415e-07,
+      "loss": 0.0162,
+      "step": 6552
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.4162201656113007e-07,
+      "loss": 0.0008,
+      "step": 6553
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.4159766195811007e-07,
+      "loss": 0.0009,
+      "step": 6554
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.415733073550901e-07,
+      "loss": 0.001,
+      "step": 6555
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.415489527520701e-07,
+      "loss": 0.0009,
+      "step": 6556
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.4152459814905017e-07,
+      "loss": 0.0008,
+      "step": 6557
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.4150024354603017e-07,
+      "loss": 0.0007,
+      "step": 6558
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 3.414758889430102e-07,
+      "loss": 0.0008,
+      "step": 6559
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.414515343399902e-07,
+      "loss": 0.0012,
+      "step": 6560
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.4142717973697027e-07,
+      "loss": 0.0008,
+      "step": 6561
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.414028251339503e-07,
+      "loss": 0.4299,
+      "step": 6562
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.413784705309303e-07,
+      "loss": 0.0008,
+      "step": 6563
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.413541159279104e-07,
+      "loss": 0.0008,
+      "step": 6564
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.4132976132489037e-07,
+      "loss": 0.0009,
+      "step": 6565
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.413054067218704e-07,
+      "loss": 0.0009,
+      "step": 6566
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.412810521188504e-07,
+      "loss": 0.0008,
+      "step": 6567
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.412566975158305e-07,
+      "loss": 0.0009,
+      "step": 6568
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.412323429128105e-07,
+      "loss": 0.0008,
+      "step": 6569
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 3.412079883097905e-07,
+      "loss": 0.001,
+      "step": 6570
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.411836337067706e-07,
+      "loss": 0.0114,
+      "step": 6571
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.411592791037506e-07,
+      "loss": 0.0008,
+      "step": 6572
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.4113492450073063e-07,
+      "loss": 0.0007,
+      "step": 6573
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.411105698977106e-07,
+      "loss": 0.0006,
+      "step": 6574
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.410862152946907e-07,
+      "loss": 0.001,
+      "step": 6575
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.410618606916707e-07,
+      "loss": 0.0007,
+      "step": 6576
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.4103750608865073e-07,
+      "loss": 0.0008,
+      "step": 6577
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.410131514856308e-07,
+      "loss": 0.0008,
+      "step": 6578
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.409887968826108e-07,
+      "loss": 0.0009,
+      "step": 6579
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 3.4096444227959083e-07,
+      "loss": 0.001,
+      "step": 6580
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.4094008767657083e-07,
+      "loss": 0.0008,
+      "step": 6581
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.409157330735509e-07,
+      "loss": 0.0007,
+      "step": 6582
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.408913784705309e-07,
+      "loss": 0.0014,
+      "step": 6583
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.4086702386751093e-07,
+      "loss": 0.0006,
+      "step": 6584
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.40842669264491e-07,
+      "loss": 0.0009,
+      "step": 6585
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.40818314661471e-07,
+      "loss": 1.7832,
+      "step": 6586
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.4079396005845103e-07,
+      "loss": 0.0008,
+      "step": 6587
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.4076960545543103e-07,
+      "loss": 0.0008,
+      "step": 6588
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.407452508524111e-07,
+      "loss": 0.001,
+      "step": 6589
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 3.407208962493911e-07,
+      "loss": 0.0009,
+      "step": 6590
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.4069654164637113e-07,
+      "loss": 0.0009,
+      "step": 6591
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.4067218704335113e-07,
+      "loss": 0.0007,
+      "step": 6592
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.406478324403312e-07,
+      "loss": 0.0009,
+      "step": 6593
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.4062347783731123e-07,
+      "loss": 0.0007,
+      "step": 6594
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.4059912323429123e-07,
+      "loss": 0.0009,
+      "step": 6595
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.405747686312713e-07,
+      "loss": 0.001,
+      "step": 6596
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.405504140282513e-07,
+      "loss": 0.0009,
+      "step": 6597
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.4052605942523134e-07,
+      "loss": 0.0009,
+      "step": 6598
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.4050170482221133e-07,
+      "loss": 0.0008,
+      "step": 6599
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.4047735021919144e-07,
+      "loss": 0.0007,
+      "step": 6600
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 3.404529956161715e-07,
+      "loss": 0.0006,
+      "step": 6601
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.404286410131515e-07,
+      "loss": 0.0009,
+      "step": 6602
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.4040428641013154e-07,
+      "loss": 0.0009,
+      "step": 6603
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.4037993180711154e-07,
+      "loss": 0.001,
+      "step": 6604
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.403555772040916e-07,
+      "loss": 0.0007,
+      "step": 6605
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.403312226010716e-07,
+      "loss": 0.0029,
+      "step": 6606
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.4030686799805164e-07,
+      "loss": 0.0008,
+      "step": 6607
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.402825133950317e-07,
+      "loss": 0.0008,
+      "step": 6608
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.402581587920117e-07,
+      "loss": 0.0009,
+      "step": 6609
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.4023380418899174e-07,
+      "loss": 0.0009,
+      "step": 6610
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 3.4020944958597174e-07,
+      "loss": 0.0008,
+      "step": 6611
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.401850949829518e-07,
+      "loss": 0.0007,
+      "step": 6612
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.401607403799318e-07,
+      "loss": 0.0007,
+      "step": 6613
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.4013638577691184e-07,
+      "loss": 0.0196,
+      "step": 6614
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.4011203117389184e-07,
+      "loss": 0.0013,
+      "step": 6615
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.400876765708719e-07,
+      "loss": 0.0008,
+      "step": 6616
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.4006332196785195e-07,
+      "loss": 0.0008,
+      "step": 6617
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.4003896736483194e-07,
+      "loss": 0.0047,
+      "step": 6618
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.40014612761812e-07,
+      "loss": 0.001,
+      "step": 6619
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.39990258158792e-07,
+      "loss": 0.0014,
+      "step": 6620
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 3.3996590355577205e-07,
+      "loss": 0.0007,
+      "step": 6621
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.3994154895275205e-07,
+      "loss": 0.0008,
+      "step": 6622
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.399171943497321e-07,
+      "loss": 0.0009,
+      "step": 6623
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.3989283974671215e-07,
+      "loss": 0.0007,
+      "step": 6624
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.3986848514369215e-07,
+      "loss": 0.0009,
+      "step": 6625
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.398441305406722e-07,
+      "loss": 0.0009,
+      "step": 6626
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.398197759376522e-07,
+      "loss": 0.0008,
+      "step": 6627
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.3979542133463225e-07,
+      "loss": 0.0007,
+      "step": 6628
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.3977106673161225e-07,
+      "loss": 0.0008,
+      "step": 6629
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.397467121285923e-07,
+      "loss": 0.0008,
+      "step": 6630
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 3.397223575255723e-07,
+      "loss": 0.0007,
+      "step": 6631
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.3969800292255235e-07,
+      "loss": 0.0009,
+      "step": 6632
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.396736483195324e-07,
+      "loss": 0.0008,
+      "step": 6633
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.396492937165124e-07,
+      "loss": 0.0008,
+      "step": 6634
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.3962493911349245e-07,
+      "loss": 0.4717,
+      "step": 6635
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.3960058451047245e-07,
+      "loss": 0.0007,
+      "step": 6636
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.395762299074525e-07,
+      "loss": 0.0008,
+      "step": 6637
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.395518753044325e-07,
+      "loss": 0.0008,
+      "step": 6638
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.3952752070141255e-07,
+      "loss": 0.001,
+      "step": 6639
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.395031660983926e-07,
+      "loss": 0.0007,
+      "step": 6640
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.394788114953726e-07,
+      "loss": 0.0007,
+      "step": 6641
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 3.3945445689235265e-07,
+      "loss": 0.001,
+      "step": 6642
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.3943010228933265e-07,
+      "loss": 0.0024,
+      "step": 6643
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.394057476863127e-07,
+      "loss": 0.0008,
+      "step": 6644
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.393813930832927e-07,
+      "loss": 0.0007,
+      "step": 6645
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.3935703848027275e-07,
+      "loss": 0.0009,
+      "step": 6646
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.3933268387725275e-07,
+      "loss": 0.1641,
+      "step": 6647
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.393083292742328e-07,
+      "loss": 0.9784,
+      "step": 6648
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.3928397467121286e-07,
+      "loss": 0.0007,
+      "step": 6649
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.3925962006819286e-07,
+      "loss": 0.5024,
+      "step": 6650
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.392352654651729e-07,
+      "loss": 0.0007,
+      "step": 6651
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 3.392109108621529e-07,
+      "loss": 0.0008,
+      "step": 6652
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.3918655625913296e-07,
+      "loss": 0.001,
+      "step": 6653
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.3916220165611296e-07,
+      "loss": 0.0009,
+      "step": 6654
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.39137847053093e-07,
+      "loss": 0.0006,
+      "step": 6655
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.3911349245007306e-07,
+      "loss": 0.0008,
+      "step": 6656
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.3908913784705306e-07,
+      "loss": 0.0006,
+      "step": 6657
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.390647832440331e-07,
+      "loss": 0.0007,
+      "step": 6658
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.390404286410131e-07,
+      "loss": 0.0008,
+      "step": 6659
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.3901607403799316e-07,
+      "loss": 0.0007,
+      "step": 6660
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.3899171943497316e-07,
+      "loss": 0.0011,
+      "step": 6661
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 3.389673648319532e-07,
+      "loss": 0.0011,
+      "step": 6662
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.389430102289332e-07,
+      "loss": 0.0006,
+      "step": 6663
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.3891865562591326e-07,
+      "loss": 0.0006,
+      "step": 6664
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.388943010228933e-07,
+      "loss": 0.0009,
+      "step": 6665
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.388699464198733e-07,
+      "loss": 0.0009,
+      "step": 6666
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.3884559181685336e-07,
+      "loss": 0.0006,
+      "step": 6667
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.3882123721383336e-07,
+      "loss": 0.0007,
+      "step": 6668
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.387968826108134e-07,
+      "loss": 0.0009,
+      "step": 6669
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.387725280077934e-07,
+      "loss": 0.001,
+      "step": 6670
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.3874817340477346e-07,
+      "loss": 0.0007,
+      "step": 6671
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.387238188017535e-07,
+      "loss": 0.0009,
+      "step": 6672
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 3.386994641987335e-07,
+      "loss": 0.001,
+      "step": 6673
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.3867510959571357e-07,
+      "loss": 0.001,
+      "step": 6674
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.3865075499269356e-07,
+      "loss": 0.0006,
+      "step": 6675
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.386264003896736e-07,
+      "loss": 0.6795,
+      "step": 6676
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.3860204578665367e-07,
+      "loss": 0.0007,
+      "step": 6677
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.385776911836337e-07,
+      "loss": 0.001,
+      "step": 6678
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.3855333658061377e-07,
+      "loss": 0.0008,
+      "step": 6679
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.3852898197759377e-07,
+      "loss": 0.0008,
+      "step": 6680
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.385046273745738e-07,
+      "loss": 0.0007,
+      "step": 6681
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.384802727715538e-07,
+      "loss": 0.0009,
+      "step": 6682
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 3.3845591816853387e-07,
+      "loss": 0.0011,
+      "step": 6683
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.3843156356551387e-07,
+      "loss": 0.0007,
+      "step": 6684
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.384072089624939e-07,
+      "loss": 0.0008,
+      "step": 6685
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.383828543594739e-07,
+      "loss": 0.0008,
+      "step": 6686
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.3835849975645397e-07,
+      "loss": 0.0008,
+      "step": 6687
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.38334145153434e-07,
+      "loss": 0.0011,
+      "step": 6688
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.38309790550414e-07,
+      "loss": 0.0009,
+      "step": 6689
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.382854359473941e-07,
+      "loss": 0.0009,
+      "step": 6690
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.3826108134437407e-07,
+      "loss": 0.0007,
+      "step": 6691
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.382367267413541e-07,
+      "loss": 0.0007,
+      "step": 6692
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 3.382123721383341e-07,
+      "loss": 0.0007,
+      "step": 6693
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.381880175353142e-07,
+      "loss": 0.0008,
+      "step": 6694
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.381636629322942e-07,
+      "loss": 0.0007,
+      "step": 6695
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.381393083292742e-07,
+      "loss": 0.0007,
+      "step": 6696
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.381149537262543e-07,
+      "loss": 0.0009,
+      "step": 6697
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.380905991232343e-07,
+      "loss": 0.0006,
+      "step": 6698
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.3806624452021433e-07,
+      "loss": 0.0008,
+      "step": 6699
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.380418899171943e-07,
+      "loss": 0.0006,
+      "step": 6700
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.380175353141744e-07,
+      "loss": 0.0009,
+      "step": 6701
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.379931807111544e-07,
+      "loss": 0.0009,
+      "step": 6702
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 3.3796882610813443e-07,
+      "loss": 0.0012,
+      "step": 6703
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.379444715051145e-07,
+      "loss": 0.0006,
+      "step": 6704
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.379201169020945e-07,
+      "loss": 0.001,
+      "step": 6705
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.3789576229907453e-07,
+      "loss": 0.001,
+      "step": 6706
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.3787140769605453e-07,
+      "loss": 0.0008,
+      "step": 6707
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.378470530930346e-07,
+      "loss": 0.0008,
+      "step": 6708
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.378226984900146e-07,
+      "loss": 0.0011,
+      "step": 6709
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.3779834388699463e-07,
+      "loss": 0.0008,
+      "step": 6710
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.377739892839747e-07,
+      "loss": 0.0008,
+      "step": 6711
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.377496346809547e-07,
+      "loss": 0.0008,
+      "step": 6712
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.3772528007793473e-07,
+      "loss": 0.0008,
+      "step": 6713
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 3.3770092547491473e-07,
+      "loss": 0.0009,
+      "step": 6714
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.376765708718948e-07,
+      "loss": 0.0008,
+      "step": 6715
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.376522162688748e-07,
+      "loss": 0.0007,
+      "step": 6716
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.3762786166585483e-07,
+      "loss": 0.0008,
+      "step": 6717
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.376035070628349e-07,
+      "loss": 0.001,
+      "step": 6718
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.375791524598149e-07,
+      "loss": 0.001,
+      "step": 6719
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.3755479785679493e-07,
+      "loss": 0.0007,
+      "step": 6720
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.3753044325377493e-07,
+      "loss": 0.001,
+      "step": 6721
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.37506088650755e-07,
+      "loss": 0.0007,
+      "step": 6722
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.37481734047735e-07,
+      "loss": 0.0981,
+      "step": 6723
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 3.3745737944471504e-07,
+      "loss": 0.0012,
+      "step": 6724
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.3743302484169503e-07,
+      "loss": 0.0009,
+      "step": 6725
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.374086702386751e-07,
+      "loss": 0.0008,
+      "step": 6726
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.3738431563565514e-07,
+      "loss": 0.0006,
+      "step": 6727
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.3735996103263514e-07,
+      "loss": 0.0007,
+      "step": 6728
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.373356064296152e-07,
+      "loss": 0.0007,
+      "step": 6729
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.373112518265952e-07,
+      "loss": 0.0007,
+      "step": 6730
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.3728689722357524e-07,
+      "loss": 0.001,
+      "step": 6731
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.3726254262055524e-07,
+      "loss": 0.0007,
+      "step": 6732
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.372381880175353e-07,
+      "loss": 0.0028,
+      "step": 6733
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 3.3721383341451534e-07,
+      "loss": 0.0012,
+      "step": 6734
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.3718947881149534e-07,
+      "loss": 0.0007,
+      "step": 6735
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.371651242084754e-07,
+      "loss": 0.0007,
+      "step": 6736
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.371407696054554e-07,
+      "loss": 0.001,
+      "step": 6737
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.3711641500243544e-07,
+      "loss": 0.0007,
+      "step": 6738
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.3709206039941544e-07,
+      "loss": 0.0007,
+      "step": 6739
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.370677057963955e-07,
+      "loss": 0.0009,
+      "step": 6740
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.370433511933755e-07,
+      "loss": 0.0007,
+      "step": 6741
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.3701899659035554e-07,
+      "loss": 0.0009,
+      "step": 6742
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.369946419873356e-07,
+      "loss": 0.0005,
+      "step": 6743
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.369702873843156e-07,
+      "loss": 0.0008,
+      "step": 6744
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 3.3694593278129564e-07,
+      "loss": 0.0007,
+      "step": 6745
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.3692157817827564e-07,
+      "loss": 0.0009,
+      "step": 6746
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.368972235752557e-07,
+      "loss": 0.0009,
+      "step": 6747
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.368728689722357e-07,
+      "loss": 0.0009,
+      "step": 6748
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.3684851436921574e-07,
+      "loss": 0.0008,
+      "step": 6749
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.368241597661958e-07,
+      "loss": 0.0008,
+      "step": 6750
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.367998051631758e-07,
+      "loss": 0.0006,
+      "step": 6751
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.3677545056015585e-07,
+      "loss": 0.001,
+      "step": 6752
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.367510959571359e-07,
+      "loss": 0.001,
+      "step": 6753
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.3672674135411595e-07,
+      "loss": 0.0009,
+      "step": 6754
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 3.3670238675109595e-07,
+      "loss": 0.0007,
+      "step": 6755
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.36678032148076e-07,
+      "loss": 0.0007,
+      "step": 6756
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.3665367754505605e-07,
+      "loss": 0.001,
+      "step": 6757
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.3662932294203605e-07,
+      "loss": 0.0008,
+      "step": 6758
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.366049683390161e-07,
+      "loss": 0.0009,
+      "step": 6759
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.365806137359961e-07,
+      "loss": 0.0008,
+      "step": 6760
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.3655625913297615e-07,
+      "loss": 0.0008,
+      "step": 6761
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.3653190452995615e-07,
+      "loss": 0.0006,
+      "step": 6762
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.365075499269362e-07,
+      "loss": 0.0006,
+      "step": 6763
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.364831953239162e-07,
+      "loss": 0.0007,
+      "step": 6764
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 3.3645884072089625e-07,
+      "loss": 0.0008,
+      "step": 6765
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.364344861178763e-07,
+      "loss": 0.0008,
+      "step": 6766
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.364101315148563e-07,
+      "loss": 0.0006,
+      "step": 6767
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.3638577691183635e-07,
+      "loss": 0.0008,
+      "step": 6768
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.3636142230881635e-07,
+      "loss": 0.0176,
+      "step": 6769
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.363370677057964e-07,
+      "loss": 0.0007,
+      "step": 6770
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.363127131027764e-07,
+      "loss": 0.0009,
+      "step": 6771
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.3628835849975646e-07,
+      "loss": 0.0033,
+      "step": 6772
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.362640038967365e-07,
+      "loss": 0.0007,
+      "step": 6773
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.362396492937165e-07,
+      "loss": 0.0008,
+      "step": 6774
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 3.3621529469069656e-07,
+      "loss": 0.0009,
+      "step": 6775
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.3619094008767656e-07,
+      "loss": 0.0007,
+      "step": 6776
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.361665854846566e-07,
+      "loss": 0.0008,
+      "step": 6777
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.361422308816366e-07,
+      "loss": 0.0008,
+      "step": 6778
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.3611787627861666e-07,
+      "loss": 0.0007,
+      "step": 6779
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.3609352167559666e-07,
+      "loss": 0.0011,
+      "step": 6780
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.360691670725767e-07,
+      "loss": 0.001,
+      "step": 6781
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.3604481246955676e-07,
+      "loss": 0.001,
+      "step": 6782
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.3602045786653676e-07,
+      "loss": 0.0008,
+      "step": 6783
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.359961032635168e-07,
+      "loss": 0.0009,
+      "step": 6784
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.359717486604968e-07,
+      "loss": 0.0007,
+      "step": 6785
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 3.3594739405747686e-07,
+      "loss": 0.0008,
+      "step": 6786
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.3592303945445686e-07,
+      "loss": 0.0008,
+      "step": 6787
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.358986848514369e-07,
+      "loss": 0.0006,
+      "step": 6788
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.3587433024841696e-07,
+      "loss": 0.0009,
+      "step": 6789
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.3584997564539696e-07,
+      "loss": 0.0014,
+      "step": 6790
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.35825621042377e-07,
+      "loss": 0.0009,
+      "step": 6791
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.35801266439357e-07,
+      "loss": 0.0009,
+      "step": 6792
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.3577691183633706e-07,
+      "loss": 0.0008,
+      "step": 6793
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.3575255723331706e-07,
+      "loss": 0.0011,
+      "step": 6794
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.357282026302971e-07,
+      "loss": 0.0007,
+      "step": 6795
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 3.357038480272771e-07,
+      "loss": 0.0007,
+      "step": 6796
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3567949342425716e-07,
+      "loss": 0.0007,
+      "step": 6797
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.356551388212372e-07,
+      "loss": 0.0007,
+      "step": 6798
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.356307842182172e-07,
+      "loss": 0.0009,
+      "step": 6799
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3560642961519727e-07,
+      "loss": 0.034,
+      "step": 6800
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3558207501217726e-07,
+      "loss": 0.0009,
+      "step": 6801
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.355577204091573e-07,
+      "loss": 0.0009,
+      "step": 6802
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.355333658061373e-07,
+      "loss": 0.0008,
+      "step": 6803
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.3550901120311737e-07,
+      "loss": 0.0007,
+      "step": 6804
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.354846566000974e-07,
+      "loss": 0.0008,
+      "step": 6805
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 3.354603019970774e-07,
+      "loss": 0.0008,
+      "step": 6806
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3543594739405747e-07,
+      "loss": 0.0006,
+      "step": 6807
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3541159279103747e-07,
+      "loss": 0.0007,
+      "step": 6808
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.353872381880175e-07,
+      "loss": 0.0006,
+      "step": 6809
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.353628835849975e-07,
+      "loss": 0.0008,
+      "step": 6810
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3533852898197757e-07,
+      "loss": 0.0009,
+      "step": 6811
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3531417437895757e-07,
+      "loss": 0.0007,
+      "step": 6812
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.352898197759376e-07,
+      "loss": 0.001,
+      "step": 6813
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3526546517291767e-07,
+      "loss": 0.0007,
+      "step": 6814
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.3524111056989767e-07,
+      "loss": 0.0007,
+      "step": 6815
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.352167559668777e-07,
+      "loss": 0.0007,
+      "step": 6816
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 3.351924013638577e-07,
+      "loss": 0.0007,
+      "step": 6817
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3516804676083777e-07,
+      "loss": 0.0008,
+      "step": 6818
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3514369215781777e-07,
+      "loss": 0.0007,
+      "step": 6819
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.351193375547978e-07,
+      "loss": 0.0011,
+      "step": 6820
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3509498295177787e-07,
+      "loss": 0.0008,
+      "step": 6821
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3507062834875787e-07,
+      "loss": 0.0008,
+      "step": 6822
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.350462737457379e-07,
+      "loss": 0.0015,
+      "step": 6823
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.350219191427179e-07,
+      "loss": 0.0007,
+      "step": 6824
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.34997564539698e-07,
+      "loss": 0.0008,
+      "step": 6825
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.3497320993667797e-07,
+      "loss": 0.001,
+      "step": 6826
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 3.34948855333658e-07,
+      "loss": 0.0008,
+      "step": 6827
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.34924500730638e-07,
+      "loss": 0.0006,
+      "step": 6828
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.349001461276181e-07,
+      "loss": 0.0008,
+      "step": 6829
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.348757915245982e-07,
+      "loss": 0.0009,
+      "step": 6830
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.348514369215782e-07,
+      "loss": 0.0008,
+      "step": 6831
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3482708231855823e-07,
+      "loss": 0.0007,
+      "step": 6832
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3480272771553823e-07,
+      "loss": 0.0009,
+      "step": 6833
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.347783731125183e-07,
+      "loss": 0.0008,
+      "step": 6834
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.347540185094983e-07,
+      "loss": 0.0009,
+      "step": 6835
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.3472966390647833e-07,
+      "loss": 0.0008,
+      "step": 6836
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 3.347053093034584e-07,
+      "loss": 0.0011,
+      "step": 6837
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.346809547004384e-07,
+      "loss": 0.001,
+      "step": 6838
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3465660009741843e-07,
+      "loss": 0.0006,
+      "step": 6839
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3463224549439843e-07,
+      "loss": 0.0009,
+      "step": 6840
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.346078908913785e-07,
+      "loss": 0.0007,
+      "step": 6841
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.345835362883585e-07,
+      "loss": 0.0006,
+      "step": 6842
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3455918168533853e-07,
+      "loss": 0.0007,
+      "step": 6843
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.345348270823186e-07,
+      "loss": 0.2702,
+      "step": 6844
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.345104724792986e-07,
+      "loss": 0.0006,
+      "step": 6845
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3448611787627863e-07,
+      "loss": 0.0009,
+      "step": 6846
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 3.3446176327325863e-07,
+      "loss": 0.0008,
+      "step": 6847
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.344374086702387e-07,
+      "loss": 0.0007,
+      "step": 6848
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.344130540672187e-07,
+      "loss": 0.0006,
+      "step": 6849
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3438869946419874e-07,
+      "loss": 0.0008,
+      "step": 6850
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3436434486117873e-07,
+      "loss": 0.0007,
+      "step": 6851
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.343399902581588e-07,
+      "loss": 0.0007,
+      "step": 6852
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3431563565513884e-07,
+      "loss": 0.0009,
+      "step": 6853
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3429128105211884e-07,
+      "loss": 0.0007,
+      "step": 6854
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.342669264490989e-07,
+      "loss": 0.0007,
+      "step": 6855
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.342425718460789e-07,
+      "loss": 0.001,
+      "step": 6856
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3421821724305894e-07,
+      "loss": 0.0009,
+      "step": 6857
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 3.3419386264003894e-07,
+      "loss": 0.0008,
+      "step": 6858
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.34169508037019e-07,
+      "loss": 0.001,
+      "step": 6859
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3414515343399904e-07,
+      "loss": 0.0007,
+      "step": 6860
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3412079883097904e-07,
+      "loss": 0.0006,
+      "step": 6861
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.340964442279591e-07,
+      "loss": 0.0008,
+      "step": 6862
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.340720896249391e-07,
+      "loss": 0.0007,
+      "step": 6863
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3404773502191914e-07,
+      "loss": 0.0007,
+      "step": 6864
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3402338041889914e-07,
+      "loss": 0.0009,
+      "step": 6865
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.339990258158792e-07,
+      "loss": 0.0011,
+      "step": 6866
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.339746712128592e-07,
+      "loss": 0.0009,
+      "step": 6867
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 3.3395031660983924e-07,
+      "loss": 0.0008,
+      "step": 6868
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.339259620068193e-07,
+      "loss": 0.0009,
+      "step": 6869
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.339016074037993e-07,
+      "loss": 0.0007,
+      "step": 6870
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3387725280077934e-07,
+      "loss": 0.0007,
+      "step": 6871
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3385289819775934e-07,
+      "loss": 0.0009,
+      "step": 6872
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.338285435947394e-07,
+      "loss": 0.0011,
+      "step": 6873
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.338041889917194e-07,
+      "loss": 0.0013,
+      "step": 6874
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3377983438869944e-07,
+      "loss": 0.0008,
+      "step": 6875
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.337554797856795e-07,
+      "loss": 0.0009,
+      "step": 6876
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.337311251826595e-07,
+      "loss": 0.0008,
+      "step": 6877
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 3.3370677057963955e-07,
+      "loss": 0.0008,
+      "step": 6878
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3368241597661954e-07,
+      "loss": 0.0008,
+      "step": 6879
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.336580613735996e-07,
+      "loss": 0.0007,
+      "step": 6880
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.336337067705796e-07,
+      "loss": 0.0008,
+      "step": 6881
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3360935216755965e-07,
+      "loss": 0.0007,
+      "step": 6882
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3358499756453965e-07,
+      "loss": 0.0006,
+      "step": 6883
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.335606429615197e-07,
+      "loss": 0.0009,
+      "step": 6884
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3353628835849975e-07,
+      "loss": 0.0008,
+      "step": 6885
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3351193375547975e-07,
+      "loss": 0.0007,
+      "step": 6886
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.334875791524598e-07,
+      "loss": 0.0008,
+      "step": 6887
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.334632245494398e-07,
+      "loss": 0.0006,
+      "step": 6888
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 3.3343886994641985e-07,
+      "loss": 0.0008,
+      "step": 6889
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3341451534339985e-07,
+      "loss": 0.0009,
+      "step": 6890
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.333901607403799e-07,
+      "loss": 0.0008,
+      "step": 6891
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3336580613735995e-07,
+      "loss": 0.0005,
+      "step": 6892
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3334145153433995e-07,
+      "loss": 0.0006,
+      "step": 6893
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3331709693132e-07,
+      "loss": 0.0009,
+      "step": 6894
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.332927423283e-07,
+      "loss": 0.0006,
+      "step": 6895
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3326838772528005e-07,
+      "loss": 0.001,
+      "step": 6896
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3324403312226005e-07,
+      "loss": 0.0006,
+      "step": 6897
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.332196785192401e-07,
+      "loss": 0.0008,
+      "step": 6898
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 3.3319532391622015e-07,
+      "loss": 0.0006,
+      "step": 6899
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3317096931320015e-07,
+      "loss": 0.0012,
+      "step": 6900
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.331466147101802e-07,
+      "loss": 0.0007,
+      "step": 6901
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.331222601071602e-07,
+      "loss": 0.0011,
+      "step": 6902
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3309790550414025e-07,
+      "loss": 0.001,
+      "step": 6903
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3307355090112025e-07,
+      "loss": 0.0006,
+      "step": 6904
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.330491962981003e-07,
+      "loss": 0.0009,
+      "step": 6905
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.330248416950803e-07,
+      "loss": 0.0007,
+      "step": 6906
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.330004870920604e-07,
+      "loss": 0.0008,
+      "step": 6907
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3297613248904046e-07,
+      "loss": 0.0006,
+      "step": 6908
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 3.3295177788602046e-07,
+      "loss": 0.0009,
+      "step": 6909
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.329274232830005e-07,
+      "loss": 0.0007,
+      "step": 6910
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.329030686799805e-07,
+      "loss": 0.0007,
+      "step": 6911
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.3287871407696056e-07,
+      "loss": 0.0006,
+      "step": 6912
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.3285435947394056e-07,
+      "loss": 0.0009,
+      "step": 6913
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.328300048709206e-07,
+      "loss": 0.0007,
+      "step": 6914
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.3280565026790066e-07,
+      "loss": 0.0008,
+      "step": 6915
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.3278129566488066e-07,
+      "loss": 0.0009,
+      "step": 6916
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.327569410618607e-07,
+      "loss": 0.0009,
+      "step": 6917
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.327325864588407e-07,
+      "loss": 0.0011,
+      "step": 6918
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.3270823185582076e-07,
+      "loss": 0.0005,
+      "step": 6919
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 3.3268387725280076e-07,
+      "loss": 0.0008,
+      "step": 6920
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.326595226497808e-07,
+      "loss": 0.0006,
+      "step": 6921
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.3263516804676087e-07,
+      "loss": 0.0006,
+      "step": 6922
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.3261081344374086e-07,
+      "loss": 0.0007,
+      "step": 6923
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.325864588407209e-07,
+      "loss": 0.001,
+      "step": 6924
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.325621042377009e-07,
+      "loss": 0.0007,
+      "step": 6925
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.3253774963468097e-07,
+      "loss": 0.0009,
+      "step": 6926
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.3251339503166096e-07,
+      "loss": 0.0009,
+      "step": 6927
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.32489040428641e-07,
+      "loss": 0.0009,
+      "step": 6928
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.32464685825621e-07,
+      "loss": 0.0007,
+      "step": 6929
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 3.3244033122260107e-07,
+      "loss": 0.0007,
+      "step": 6930
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.324159766195811e-07,
+      "loss": 0.0009,
+      "step": 6931
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.323916220165611e-07,
+      "loss": 0.0009,
+      "step": 6932
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.3236726741354117e-07,
+      "loss": 0.0007,
+      "step": 6933
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.3234291281052117e-07,
+      "loss": 0.0009,
+      "step": 6934
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.323185582075012e-07,
+      "loss": 0.0006,
+      "step": 6935
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.322942036044812e-07,
+      "loss": 0.0009,
+      "step": 6936
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.3226984900146127e-07,
+      "loss": 0.0008,
+      "step": 6937
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.322454943984413e-07,
+      "loss": 0.0005,
+      "step": 6938
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.322211397954213e-07,
+      "loss": 0.0007,
+      "step": 6939
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 3.3219678519240137e-07,
+      "loss": 0.0007,
+      "step": 6940
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.3217243058938137e-07,
+      "loss": 0.0008,
+      "step": 6941
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.321480759863614e-07,
+      "loss": 0.0033,
+      "step": 6942
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.321237213833414e-07,
+      "loss": 0.0007,
+      "step": 6943
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.3209936678032147e-07,
+      "loss": 0.0011,
+      "step": 6944
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.3207501217730147e-07,
+      "loss": 0.0012,
+      "step": 6945
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.320506575742815e-07,
+      "loss": 0.0007,
+      "step": 6946
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.320263029712616e-07,
+      "loss": 0.001,
+      "step": 6947
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.3200194836824157e-07,
+      "loss": 0.0006,
+      "step": 6948
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.319775937652216e-07,
+      "loss": 0.0006,
+      "step": 6949
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 3.319532391622016e-07,
+      "loss": 0.0008,
+      "step": 6950
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.319288845591817e-07,
+      "loss": 0.0064,
+      "step": 6951
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.3190452995616167e-07,
+      "loss": 0.0008,
+      "step": 6952
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.318801753531417e-07,
+      "loss": 0.0006,
+      "step": 6953
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.318558207501218e-07,
+      "loss": 0.0011,
+      "step": 6954
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.318314661471018e-07,
+      "loss": 0.0006,
+      "step": 6955
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.3180711154408183e-07,
+      "loss": 0.0005,
+      "step": 6956
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.317827569410618e-07,
+      "loss": 0.0007,
+      "step": 6957
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.317584023380419e-07,
+      "loss": 0.0007,
+      "step": 6958
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.317340477350219e-07,
+      "loss": 0.0005,
+      "step": 6959
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.3170969313200193e-07,
+      "loss": 0.0008,
+      "step": 6960
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 3.316853385289819e-07,
+      "loss": 0.0007,
+      "step": 6961
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.31660983925962e-07,
+      "loss": 0.0008,
+      "step": 6962
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.3163662932294203e-07,
+      "loss": 0.0007,
+      "step": 6963
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.3161227471992203e-07,
+      "loss": 0.0011,
+      "step": 6964
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.315879201169021e-07,
+      "loss": 0.0006,
+      "step": 6965
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.315635655138821e-07,
+      "loss": 0.0006,
+      "step": 6966
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.3153921091086213e-07,
+      "loss": 0.0007,
+      "step": 6967
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.3151485630784213e-07,
+      "loss": 0.0008,
+      "step": 6968
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.314905017048222e-07,
+      "loss": 0.0007,
+      "step": 6969
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.3146614710180223e-07,
+      "loss": 0.0007,
+      "step": 6970
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 3.3144179249878223e-07,
+      "loss": 0.0009,
+      "step": 6971
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.314174378957623e-07,
+      "loss": 0.0009,
+      "step": 6972
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.313930832927423e-07,
+      "loss": 0.0008,
+      "step": 6973
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.3136872868972233e-07,
+      "loss": 0.0008,
+      "step": 6974
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.3134437408670233e-07,
+      "loss": 0.0007,
+      "step": 6975
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.313200194836824e-07,
+      "loss": 0.0008,
+      "step": 6976
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.312956648806624e-07,
+      "loss": 0.0006,
+      "step": 6977
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.3127131027764243e-07,
+      "loss": 0.0016,
+      "step": 6978
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.312469556746225e-07,
+      "loss": 0.0006,
+      "step": 6979
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.312226010716025e-07,
+      "loss": 0.001,
+      "step": 6980
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 3.3119824646858254e-07,
+      "loss": 0.0008,
+      "step": 6981
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.3117389186556253e-07,
+      "loss": 0.0005,
+      "step": 6982
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.3114953726254264e-07,
+      "loss": 0.0006,
+      "step": 6983
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.3112518265952264e-07,
+      "loss": 0.0008,
+      "step": 6984
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.311008280565027e-07,
+      "loss": 0.001,
+      "step": 6985
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.3107647345348274e-07,
+      "loss": 0.0008,
+      "step": 6986
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.3105211885046274e-07,
+      "loss": 0.0008,
+      "step": 6987
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.310277642474428e-07,
+      "loss": 0.0007,
+      "step": 6988
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.310034096444228e-07,
+      "loss": 0.0006,
+      "step": 6989
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.3097905504140284e-07,
+      "loss": 0.0005,
+      "step": 6990
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.3095470043838284e-07,
+      "loss": 0.0005,
+      "step": 6991
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 3.309303458353629e-07,
+      "loss": 0.0007,
+      "step": 6992
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.3090599123234294e-07,
+      "loss": 0.0008,
+      "step": 6993
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.3088163662932294e-07,
+      "loss": 0.0005,
+      "step": 6994
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.30857282026303e-07,
+      "loss": 0.0006,
+      "step": 6995
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.30832927423283e-07,
+      "loss": 0.0006,
+      "step": 6996
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.3080857282026304e-07,
+      "loss": 0.0007,
+      "step": 6997
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.3078421821724304e-07,
+      "loss": 0.0005,
+      "step": 6998
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.307598636142231e-07,
+      "loss": 0.0007,
+      "step": 6999
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.307355090112031e-07,
+      "loss": 0.001,
+      "step": 7000
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.3071115440818314e-07,
+      "loss": 0.0006,
+      "step": 7001
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 3.306867998051632e-07,
+      "loss": 0.001,
+      "step": 7002
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.306624452021432e-07,
+      "loss": 0.0006,
+      "step": 7003
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.3063809059912325e-07,
+      "loss": 0.0006,
+      "step": 7004
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.3061373599610325e-07,
+      "loss": 0.0006,
+      "step": 7005
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.305893813930833e-07,
+      "loss": 0.0007,
+      "step": 7006
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.305650267900633e-07,
+      "loss": 0.0007,
+      "step": 7007
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.3054067218704335e-07,
+      "loss": 0.001,
+      "step": 7008
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.305163175840234e-07,
+      "loss": 1.7046,
+      "step": 7009
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.304919629810034e-07,
+      "loss": 0.0006,
+      "step": 7010
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.3046760837798345e-07,
+      "loss": 0.0007,
+      "step": 7011
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 3.3044325377496345e-07,
+      "loss": 0.0006,
+      "step": 7012
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.304188991719435e-07,
+      "loss": 0.0007,
+      "step": 7013
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.303945445689235e-07,
+      "loss": 0.0006,
+      "step": 7014
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.3037018996590355e-07,
+      "loss": 0.0006,
+      "step": 7015
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.3034583536288355e-07,
+      "loss": 0.0006,
+      "step": 7016
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.303214807598636e-07,
+      "loss": 0.0009,
+      "step": 7017
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.3029712615684365e-07,
+      "loss": 0.001,
+      "step": 7018
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.3027277155382365e-07,
+      "loss": 0.0006,
+      "step": 7019
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.302484169508037e-07,
+      "loss": 0.0007,
+      "step": 7020
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.302240623477837e-07,
+      "loss": 0.0008,
+      "step": 7021
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 3.3019970774476375e-07,
+      "loss": 0.0005,
+      "step": 7022
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.3017535314174375e-07,
+      "loss": 0.001,
+      "step": 7023
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.301509985387238e-07,
+      "loss": 0.0008,
+      "step": 7024
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.3012664393570385e-07,
+      "loss": 0.0008,
+      "step": 7025
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.3010228933268385e-07,
+      "loss": 0.0009,
+      "step": 7026
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.300779347296639e-07,
+      "loss": 0.0007,
+      "step": 7027
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.300535801266439e-07,
+      "loss": 0.0008,
+      "step": 7028
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.3002922552362395e-07,
+      "loss": 0.0008,
+      "step": 7029
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.3000487092060395e-07,
+      "loss": 0.0008,
+      "step": 7030
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.29980516317584e-07,
+      "loss": 0.0008,
+      "step": 7031
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.29956161714564e-07,
+      "loss": 0.0007,
+      "step": 7032
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 3.2993180711154406e-07,
+      "loss": 0.0007,
+      "step": 7033
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.299074525085241e-07,
+      "loss": 0.0007,
+      "step": 7034
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.298830979055041e-07,
+      "loss": 0.0008,
+      "step": 7035
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.2985874330248416e-07,
+      "loss": 0.0007,
+      "step": 7036
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.2983438869946416e-07,
+      "loss": 0.0008,
+      "step": 7037
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.298100340964442e-07,
+      "loss": 0.0007,
+      "step": 7038
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.297856794934242e-07,
+      "loss": 0.0007,
+      "step": 7039
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.2976132489040426e-07,
+      "loss": 0.0008,
+      "step": 7040
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.297369702873843e-07,
+      "loss": 0.0008,
+      "step": 7041
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.297126156843643e-07,
+      "loss": 0.001,
+      "step": 7042
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 3.2968826108134436e-07,
+      "loss": 0.0007,
+      "step": 7043
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.2966390647832436e-07,
+      "loss": 0.0007,
+      "step": 7044
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.296395518753044e-07,
+      "loss": 0.0008,
+      "step": 7045
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.296151972722844e-07,
+      "loss": 0.0006,
+      "step": 7046
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.2959084266926446e-07,
+      "loss": 0.0008,
+      "step": 7047
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.2956648806624446e-07,
+      "loss": 0.0007,
+      "step": 7048
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.295421334632245e-07,
+      "loss": 0.0007,
+      "step": 7049
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.2951777886020456e-07,
+      "loss": 0.0006,
+      "step": 7050
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.2949342425718456e-07,
+      "loss": 0.0008,
+      "step": 7051
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.294690696541646e-07,
+      "loss": 0.0006,
+      "step": 7052
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 3.294447150511446e-07,
+      "loss": 0.0007,
+      "step": 7053
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.2942036044812466e-07,
+      "loss": 0.0009,
+      "step": 7054
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.2939600584510466e-07,
+      "loss": 0.0008,
+      "step": 7055
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.293716512420847e-07,
+      "loss": 0.0008,
+      "step": 7056
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.2934729663906477e-07,
+      "loss": 1.7515,
+      "step": 7057
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.2932294203604476e-07,
+      "loss": 0.0009,
+      "step": 7058
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.292985874330248e-07,
+      "loss": 0.0008,
+      "step": 7059
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.2927423283000487e-07,
+      "loss": 0.001,
+      "step": 7060
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.292498782269849e-07,
+      "loss": 0.0008,
+      "step": 7061
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.292255236239649e-07,
+      "loss": 0.0013,
+      "step": 7062
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.2920116902094497e-07,
+      "loss": 0.0009,
+      "step": 7063
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 3.29176814417925e-07,
+      "loss": 0.0144,
+      "step": 7064
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.29152459814905e-07,
+      "loss": 0.0006,
+      "step": 7065
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.2912810521188507e-07,
+      "loss": 0.0007,
+      "step": 7066
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.2910375060886507e-07,
+      "loss": 0.0009,
+      "step": 7067
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.290793960058451e-07,
+      "loss": 0.001,
+      "step": 7068
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.290550414028251e-07,
+      "loss": 0.0009,
+      "step": 7069
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.2903068679980517e-07,
+      "loss": 0.0005,
+      "step": 7070
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.2900633219678517e-07,
+      "loss": 0.0006,
+      "step": 7071
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.289819775937652e-07,
+      "loss": 0.0007,
+      "step": 7072
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.289576229907453e-07,
+      "loss": 0.0006,
+      "step": 7073
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 3.2893326838772527e-07,
+      "loss": 0.0482,
+      "step": 7074
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.289089137847053e-07,
+      "loss": 0.0007,
+      "step": 7075
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.288845591816853e-07,
+      "loss": 0.0008,
+      "step": 7076
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.288602045786654e-07,
+      "loss": 0.0006,
+      "step": 7077
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.2883584997564537e-07,
+      "loss": 0.0008,
+      "step": 7078
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.288114953726254e-07,
+      "loss": 0.0013,
+      "step": 7079
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.287871407696055e-07,
+      "loss": 0.0006,
+      "step": 7080
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.287627861665855e-07,
+      "loss": 0.0007,
+      "step": 7081
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.2873843156356553e-07,
+      "loss": 0.001,
+      "step": 7082
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.287140769605455e-07,
+      "loss": 0.0008,
+      "step": 7083
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 3.286897223575256e-07,
+      "loss": 0.0011,
+      "step": 7084
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.286653677545056e-07,
+      "loss": 0.0008,
+      "step": 7085
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.2864101315148563e-07,
+      "loss": 0.0007,
+      "step": 7086
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.286166585484657e-07,
+      "loss": 0.0006,
+      "step": 7087
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.285923039454457e-07,
+      "loss": 0.0006,
+      "step": 7088
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.2856794934242573e-07,
+      "loss": 0.0006,
+      "step": 7089
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.2854359473940573e-07,
+      "loss": 0.0006,
+      "step": 7090
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.285192401363858e-07,
+      "loss": 0.0008,
+      "step": 7091
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.284948855333658e-07,
+      "loss": 0.0008,
+      "step": 7092
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.2847053093034583e-07,
+      "loss": 0.0008,
+      "step": 7093
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 3.2844617632732583e-07,
+      "loss": 0.0017,
+      "step": 7094
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.284218217243059e-07,
+      "loss": 0.0007,
+      "step": 7095
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.2839746712128593e-07,
+      "loss": 0.0007,
+      "step": 7096
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.2837311251826593e-07,
+      "loss": 0.0008,
+      "step": 7097
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.28348757915246e-07,
+      "loss": 0.0006,
+      "step": 7098
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.28324403312226e-07,
+      "loss": 0.0007,
+      "step": 7099
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.2830004870920603e-07,
+      "loss": 0.0006,
+      "step": 7100
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.2827569410618603e-07,
+      "loss": 0.0007,
+      "step": 7101
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.282513395031661e-07,
+      "loss": 0.0005,
+      "step": 7102
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.2822698490014613e-07,
+      "loss": 0.0008,
+      "step": 7103
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.2820263029712613e-07,
+      "loss": 0.0006,
+      "step": 7104
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 3.281782756941062e-07,
+      "loss": 0.0007,
+      "step": 7105
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.281539210910862e-07,
+      "loss": 0.0007,
+      "step": 7106
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.2812956648806624e-07,
+      "loss": 0.0006,
+      "step": 7107
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.2810521188504623e-07,
+      "loss": 0.0008,
+      "step": 7108
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.280808572820263e-07,
+      "loss": 0.0007,
+      "step": 7109
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.280565026790063e-07,
+      "loss": 0.0008,
+      "step": 7110
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.2803214807598634e-07,
+      "loss": 0.0006,
+      "step": 7111
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.280077934729664e-07,
+      "loss": 0.0006,
+      "step": 7112
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.279834388699464e-07,
+      "loss": 0.0005,
+      "step": 7113
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.2795908426692644e-07,
+      "loss": 0.0008,
+      "step": 7114
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 3.2793472966390644e-07,
+      "loss": 0.0008,
+      "step": 7115
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.279103750608865e-07,
+      "loss": 0.0005,
+      "step": 7116
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.278860204578665e-07,
+      "loss": 0.0007,
+      "step": 7117
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.2786166585484654e-07,
+      "loss": 0.0008,
+      "step": 7118
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.278373112518266e-07,
+      "loss": 0.0005,
+      "step": 7119
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.278129566488066e-07,
+      "loss": 0.0007,
+      "step": 7120
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.2778860204578664e-07,
+      "loss": 0.0006,
+      "step": 7121
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.2776424744276664e-07,
+      "loss": 0.0008,
+      "step": 7122
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.277398928397467e-07,
+      "loss": 0.0008,
+      "step": 7123
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.277155382367267e-07,
+      "loss": 0.0009,
+      "step": 7124
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 3.2769118363370674e-07,
+      "loss": 0.0008,
+      "step": 7125
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.2766682903068674e-07,
+      "loss": 0.0009,
+      "step": 7126
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.276424744276668e-07,
+      "loss": 0.0008,
+      "step": 7127
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.2761811982464684e-07,
+      "loss": 0.0011,
+      "step": 7128
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.2759376522162684e-07,
+      "loss": 0.0005,
+      "step": 7129
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.275694106186069e-07,
+      "loss": 0.0005,
+      "step": 7130
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.275450560155869e-07,
+      "loss": 0.0007,
+      "step": 7131
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.2752070141256694e-07,
+      "loss": 0.0008,
+      "step": 7132
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.2749634680954694e-07,
+      "loss": 0.0007,
+      "step": 7133
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.27471992206527e-07,
+      "loss": 0.0007,
+      "step": 7134
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.2744763760350705e-07,
+      "loss": 0.0007,
+      "step": 7135
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 3.274232830004871e-07,
+      "loss": 0.0006,
+      "step": 7136
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.2739892839746715e-07,
+      "loss": 0.0006,
+      "step": 7137
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.2737457379444715e-07,
+      "loss": 0.0224,
+      "step": 7138
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.273502191914272e-07,
+      "loss": 0.0007,
+      "step": 7139
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.273258645884072e-07,
+      "loss": 0.4779,
+      "step": 7140
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.2730150998538725e-07,
+      "loss": 0.0006,
+      "step": 7141
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.272771553823673e-07,
+      "loss": 0.0006,
+      "step": 7142
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.272528007793473e-07,
+      "loss": 0.0006,
+      "step": 7143
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.2722844617632735e-07,
+      "loss": 0.0006,
+      "step": 7144
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.2720409157330735e-07,
+      "loss": 0.0006,
+      "step": 7145
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 3.271797369702874e-07,
+      "loss": 0.0008,
+      "step": 7146
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.271553823672674e-07,
+      "loss": 0.0009,
+      "step": 7147
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.2713102776424745e-07,
+      "loss": 0.0007,
+      "step": 7148
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.2710667316122745e-07,
+      "loss": 0.0006,
+      "step": 7149
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.270823185582075e-07,
+      "loss": 0.0007,
+      "step": 7150
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.2705796395518755e-07,
+      "loss": 0.0007,
+      "step": 7151
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.2703360935216755e-07,
+      "loss": 0.0009,
+      "step": 7152
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.270092547491476e-07,
+      "loss": 0.0007,
+      "step": 7153
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.269849001461276e-07,
+      "loss": 0.0007,
+      "step": 7154
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.2696054554310766e-07,
+      "loss": 0.0006,
+      "step": 7155
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 3.2693619094008765e-07,
+      "loss": 0.0007,
+      "step": 7156
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.269118363370677e-07,
+      "loss": 0.0008,
+      "step": 7157
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.2688748173404776e-07,
+      "loss": 0.0009,
+      "step": 7158
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.2686312713102776e-07,
+      "loss": 0.0008,
+      "step": 7159
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.268387725280078e-07,
+      "loss": 0.0006,
+      "step": 7160
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.268144179249878e-07,
+      "loss": 0.0006,
+      "step": 7161
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.2679006332196786e-07,
+      "loss": 0.0008,
+      "step": 7162
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.2676570871894786e-07,
+      "loss": 0.0007,
+      "step": 7163
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.267413541159279e-07,
+      "loss": 0.0006,
+      "step": 7164
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.267169995129079e-07,
+      "loss": 0.0007,
+      "step": 7165
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 3.2669264490988796e-07,
+      "loss": 0.0007,
+      "step": 7166
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.26668290306868e-07,
+      "loss": 0.0091,
+      "step": 7167
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.26643935703848e-07,
+      "loss": 0.0006,
+      "step": 7168
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.2661958110082806e-07,
+      "loss": 0.0007,
+      "step": 7169
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.2659522649780806e-07,
+      "loss": 0.0006,
+      "step": 7170
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.265708718947881e-07,
+      "loss": 0.0015,
+      "step": 7171
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.265465172917681e-07,
+      "loss": 0.309,
+      "step": 7172
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.2652216268874816e-07,
+      "loss": 0.0008,
+      "step": 7173
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.264978080857282e-07,
+      "loss": 0.0006,
+      "step": 7174
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.264734534827082e-07,
+      "loss": 0.0008,
+      "step": 7175
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.2644909887968826e-07,
+      "loss": 0.0006,
+      "step": 7176
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 3.2642474427666826e-07,
+      "loss": 0.0009,
+      "step": 7177
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.264003896736483e-07,
+      "loss": 0.0007,
+      "step": 7178
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.263760350706283e-07,
+      "loss": 0.0008,
+      "step": 7179
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.2635168046760836e-07,
+      "loss": 0.0006,
+      "step": 7180
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.2632732586458836e-07,
+      "loss": 0.0007,
+      "step": 7181
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.263029712615684e-07,
+      "loss": 0.0008,
+      "step": 7182
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.2627861665854847e-07,
+      "loss": 0.0007,
+      "step": 7183
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.2625426205552846e-07,
+      "loss": 0.0009,
+      "step": 7184
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.262299074525085e-07,
+      "loss": 0.0007,
+      "step": 7185
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.262055528494885e-07,
+      "loss": 0.0007,
+      "step": 7186
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 3.2618119824646857e-07,
+      "loss": 0.0007,
+      "step": 7187
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.2615684364344856e-07,
+      "loss": 0.0008,
+      "step": 7188
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.261324890404286e-07,
+      "loss": 0.0007,
+      "step": 7189
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.2610813443740867e-07,
+      "loss": 0.001,
+      "step": 7190
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.2608377983438867e-07,
+      "loss": 0.0006,
+      "step": 7191
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.260594252313687e-07,
+      "loss": 0.0007,
+      "step": 7192
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.260350706283487e-07,
+      "loss": 0.0007,
+      "step": 7193
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.2601071602532877e-07,
+      "loss": 0.0006,
+      "step": 7194
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.2598636142230877e-07,
+      "loss": 0.0008,
+      "step": 7195
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.259620068192888e-07,
+      "loss": 0.0006,
+      "step": 7196
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 3.259376522162688e-07,
+      "loss": 0.0006,
+      "step": 7197
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.2591329761324887e-07,
+      "loss": 0.0006,
+      "step": 7198
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.258889430102289e-07,
+      "loss": 0.001,
+      "step": 7199
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.258645884072089e-07,
+      "loss": 0.0006,
+      "step": 7200
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.2584023380418897e-07,
+      "loss": 2.0563,
+      "step": 7201
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.2581587920116897e-07,
+      "loss": 0.0007,
+      "step": 7202
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.25791524598149e-07,
+      "loss": 0.0004,
+      "step": 7203
+    },
+    {
+      "epoch": 7.0,
+      "eval_accuracy": 0.9985415653864852,
+      "eval_loss": 0.015450915321707726,
+      "eval_runtime": 2.7885,
+      "eval_samples_per_second": 737.675,
+      "eval_steps_per_second": 11.834,
+      "step": 7203
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.25767169995129e-07,
+      "loss": 0.0005,
+      "step": 7204
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.2574281539210907e-07,
+      "loss": 0.0018,
+      "step": 7205
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.257184607890891e-07,
+      "loss": 0.0005,
+      "step": 7206
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.256941061860691e-07,
+      "loss": 0.0023,
+      "step": 7207
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 3.256697515830492e-07,
+      "loss": 0.0006,
+      "step": 7208
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.2564539698002917e-07,
+      "loss": 0.0009,
+      "step": 7209
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.256210423770092e-07,
+      "loss": 0.0007,
+      "step": 7210
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.255966877739892e-07,
+      "loss": 0.001,
+      "step": 7211
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.255723331709693e-07,
+      "loss": 0.0006,
+      "step": 7212
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.255479785679494e-07,
+      "loss": 0.0008,
+      "step": 7213
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.255236239649294e-07,
+      "loss": 0.0006,
+      "step": 7214
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.2549926936190943e-07,
+      "loss": 0.0009,
+      "step": 7215
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.2547491475888943e-07,
+      "loss": 0.0006,
+      "step": 7216
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.254505601558695e-07,
+      "loss": 0.0009,
+      "step": 7217
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 3.254262055528495e-07,
+      "loss": 0.0006,
+      "step": 7218
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.2540185094982953e-07,
+      "loss": 0.0006,
+      "step": 7219
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.2537749634680953e-07,
+      "loss": 0.0006,
+      "step": 7220
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.253531417437896e-07,
+      "loss": 0.0009,
+      "step": 7221
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.2532878714076963e-07,
+      "loss": 0.0006,
+      "step": 7222
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.2530443253774963e-07,
+      "loss": 0.0007,
+      "step": 7223
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.252800779347297e-07,
+      "loss": 0.0007,
+      "step": 7224
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.252557233317097e-07,
+      "loss": 0.0008,
+      "step": 7225
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.2523136872868973e-07,
+      "loss": 0.0006,
+      "step": 7226
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.2520701412566973e-07,
+      "loss": 0.0006,
+      "step": 7227
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 3.251826595226498e-07,
+      "loss": 0.0006,
+      "step": 7228
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.2515830491962983e-07,
+      "loss": 0.0008,
+      "step": 7229
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.2513395031660983e-07,
+      "loss": 0.0006,
+      "step": 7230
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.251095957135899e-07,
+      "loss": 0.0006,
+      "step": 7231
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.250852411105699e-07,
+      "loss": 0.0012,
+      "step": 7232
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.2506088650754994e-07,
+      "loss": 0.0007,
+      "step": 7233
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.2503653190452993e-07,
+      "loss": 0.0007,
+      "step": 7234
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.2501217730151e-07,
+      "loss": 0.0009,
+      "step": 7235
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.2498782269849e-07,
+      "loss": 0.0006,
+      "step": 7236
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.2496346809547004e-07,
+      "loss": 0.0008,
+      "step": 7237
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.249391134924501e-07,
+      "loss": 0.0008,
+      "step": 7238
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 3.249147588894301e-07,
+      "loss": 0.0009,
+      "step": 7239
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.2489040428641014e-07,
+      "loss": 0.0007,
+      "step": 7240
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.2486604968339014e-07,
+      "loss": 0.0007,
+      "step": 7241
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.248416950803702e-07,
+      "loss": 0.0009,
+      "step": 7242
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.248173404773502e-07,
+      "loss": 0.0006,
+      "step": 7243
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.2479298587433024e-07,
+      "loss": 0.0007,
+      "step": 7244
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.247686312713103e-07,
+      "loss": 0.1061,
+      "step": 7245
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.247442766682903e-07,
+      "loss": 0.0006,
+      "step": 7246
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.2471992206527034e-07,
+      "loss": 0.0008,
+      "step": 7247
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.2469556746225034e-07,
+      "loss": 0.0006,
+      "step": 7248
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 3.246712128592304e-07,
+      "loss": 0.0016,
+      "step": 7249
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.246468582562104e-07,
+      "loss": 0.0006,
+      "step": 7250
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.2462250365319044e-07,
+      "loss": 0.0017,
+      "step": 7251
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.2459814905017044e-07,
+      "loss": 0.0006,
+      "step": 7252
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.245737944471505e-07,
+      "loss": 0.0092,
+      "step": 7253
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.2454943984413054e-07,
+      "loss": 0.0008,
+      "step": 7254
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.2452508524111054e-07,
+      "loss": 0.0006,
+      "step": 7255
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.245007306380906e-07,
+      "loss": 0.0008,
+      "step": 7256
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.244763760350706e-07,
+      "loss": 0.0007,
+      "step": 7257
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.2445202143205064e-07,
+      "loss": 0.0007,
+      "step": 7258
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 3.2442766682903064e-07,
+      "loss": 1.3612,
+      "step": 7259
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.244033122260107e-07,
+      "loss": 0.0008,
+      "step": 7260
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.2437895762299075e-07,
+      "loss": 0.0006,
+      "step": 7261
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.2435460301997074e-07,
+      "loss": 0.0005,
+      "step": 7262
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.243302484169508e-07,
+      "loss": 0.0009,
+      "step": 7263
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.243058938139308e-07,
+      "loss": 0.0006,
+      "step": 7264
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.2428153921091085e-07,
+      "loss": 0.0008,
+      "step": 7265
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.2425718460789085e-07,
+      "loss": 0.0007,
+      "step": 7266
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.242328300048709e-07,
+      "loss": 0.0007,
+      "step": 7267
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.2420847540185095e-07,
+      "loss": 0.0006,
+      "step": 7268
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 3.2418412079883095e-07,
+      "loss": 0.0006,
+      "step": 7269
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.24159766195811e-07,
+      "loss": 0.0006,
+      "step": 7270
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.24135411592791e-07,
+      "loss": 0.0007,
+      "step": 7271
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.2411105698977105e-07,
+      "loss": 1.9251,
+      "step": 7272
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.2408670238675105e-07,
+      "loss": 0.0006,
+      "step": 7273
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.240623477837311e-07,
+      "loss": 0.0009,
+      "step": 7274
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.240379931807111e-07,
+      "loss": 0.0007,
+      "step": 7275
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.2401363857769115e-07,
+      "loss": 0.0006,
+      "step": 7276
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.239892839746712e-07,
+      "loss": 0.0005,
+      "step": 7277
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.239649293716512e-07,
+      "loss": 0.0007,
+      "step": 7278
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.2394057476863125e-07,
+      "loss": 0.0008,
+      "step": 7279
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 3.2391622016561125e-07,
+      "loss": 0.0011,
+      "step": 7280
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.238918655625913e-07,
+      "loss": 0.0007,
+      "step": 7281
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.238675109595713e-07,
+      "loss": 0.0007,
+      "step": 7282
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.2384315635655135e-07,
+      "loss": 0.0007,
+      "step": 7283
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.238188017535314e-07,
+      "loss": 0.0006,
+      "step": 7284
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.237944471505114e-07,
+      "loss": 0.0007,
+      "step": 7285
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.2377009254749145e-07,
+      "loss": 0.0005,
+      "step": 7286
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.2374573794447145e-07,
+      "loss": 0.0006,
+      "step": 7287
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.237213833414515e-07,
+      "loss": 0.0008,
+      "step": 7288
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.236970287384315e-07,
+      "loss": 0.0006,
+      "step": 7289
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 3.236726741354116e-07,
+      "loss": 0.0006,
+      "step": 7290
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.2364831953239166e-07,
+      "loss": 0.0007,
+      "step": 7291
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.2362396492937166e-07,
+      "loss": 0.0007,
+      "step": 7292
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.235996103263517e-07,
+      "loss": 0.0007,
+      "step": 7293
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.235752557233317e-07,
+      "loss": 0.0008,
+      "step": 7294
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.2355090112031176e-07,
+      "loss": 0.0007,
+      "step": 7295
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.2352654651729176e-07,
+      "loss": 0.0006,
+      "step": 7296
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.235021919142718e-07,
+      "loss": 0.0006,
+      "step": 7297
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.234778373112518e-07,
+      "loss": 0.0005,
+      "step": 7298
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.2345348270823186e-07,
+      "loss": 0.0006,
+      "step": 7299
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 3.234291281052119e-07,
+      "loss": 0.0006,
+      "step": 7300
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.234047735021919e-07,
+      "loss": 0.0005,
+      "step": 7301
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.2338041889917196e-07,
+      "loss": 1.5824,
+      "step": 7302
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.2335606429615196e-07,
+      "loss": 0.0006,
+      "step": 7303
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.23331709693132e-07,
+      "loss": 0.001,
+      "step": 7304
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.23307355090112e-07,
+      "loss": 0.0008,
+      "step": 7305
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.2328300048709206e-07,
+      "loss": 0.001,
+      "step": 7306
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.232586458840721e-07,
+      "loss": 0.0006,
+      "step": 7307
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.232342912810521e-07,
+      "loss": 0.0007,
+      "step": 7308
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.2320993667803217e-07,
+      "loss": 0.0008,
+      "step": 7309
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.2318558207501216e-07,
+      "loss": 0.0006,
+      "step": 7310
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 3.231612274719922e-07,
+      "loss": 0.0006,
+      "step": 7311
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.231368728689722e-07,
+      "loss": 0.0007,
+      "step": 7312
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.2311251826595227e-07,
+      "loss": 0.0007,
+      "step": 7313
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.2308816366293227e-07,
+      "loss": 0.0009,
+      "step": 7314
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.230638090599123e-07,
+      "loss": 0.0008,
+      "step": 7315
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.2303945445689237e-07,
+      "loss": 0.0007,
+      "step": 7316
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.2301509985387237e-07,
+      "loss": 0.0007,
+      "step": 7317
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.229907452508524e-07,
+      "loss": 0.0008,
+      "step": 7318
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.229663906478324e-07,
+      "loss": 0.0008,
+      "step": 7319
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.2294203604481247e-07,
+      "loss": 0.0009,
+      "step": 7320
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 3.2291768144179247e-07,
+      "loss": 0.0005,
+      "step": 7321
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.228933268387725e-07,
+      "loss": 0.0008,
+      "step": 7322
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.2286897223575257e-07,
+      "loss": 0.0006,
+      "step": 7323
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.2284461763273257e-07,
+      "loss": 0.0005,
+      "step": 7324
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.228202630297126e-07,
+      "loss": 0.0008,
+      "step": 7325
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.227959084266926e-07,
+      "loss": 0.0007,
+      "step": 7326
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.2277155382367267e-07,
+      "loss": 0.0008,
+      "step": 7327
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.2274719922065267e-07,
+      "loss": 0.0006,
+      "step": 7328
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.227228446176327e-07,
+      "loss": 0.0007,
+      "step": 7329
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.226984900146127e-07,
+      "loss": 0.0007,
+      "step": 7330
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 3.2267413541159277e-07,
+      "loss": 0.0006,
+      "step": 7331
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.226497808085728e-07,
+      "loss": 0.001,
+      "step": 7332
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.226254262055528e-07,
+      "loss": 0.0006,
+      "step": 7333
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.226010716025329e-07,
+      "loss": 0.0008,
+      "step": 7334
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.2257671699951287e-07,
+      "loss": 0.0007,
+      "step": 7335
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.225523623964929e-07,
+      "loss": 0.0005,
+      "step": 7336
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.225280077934729e-07,
+      "loss": 0.0007,
+      "step": 7337
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.22503653190453e-07,
+      "loss": 0.0006,
+      "step": 7338
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.2247929858743303e-07,
+      "loss": 0.0005,
+      "step": 7339
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.22454943984413e-07,
+      "loss": 0.0007,
+      "step": 7340
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 3.224305893813931e-07,
+      "loss": 0.0008,
+      "step": 7341
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.224062347783731e-07,
+      "loss": 0.0005,
+      "step": 7342
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.2238188017535313e-07,
+      "loss": 0.0006,
+      "step": 7343
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.223575255723331e-07,
+      "loss": 0.0011,
+      "step": 7344
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.223331709693132e-07,
+      "loss": 0.0007,
+      "step": 7345
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.223088163662932e-07,
+      "loss": 0.0006,
+      "step": 7346
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.2228446176327323e-07,
+      "loss": 0.0007,
+      "step": 7347
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.222601071602533e-07,
+      "loss": 0.0006,
+      "step": 7348
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.222357525572333e-07,
+      "loss": 0.0007,
+      "step": 7349
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.2221139795421333e-07,
+      "loss": 0.0007,
+      "step": 7350
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.2218704335119333e-07,
+      "loss": 0.0008,
+      "step": 7351
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 3.221626887481734e-07,
+      "loss": 0.0005,
+      "step": 7352
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.221383341451534e-07,
+      "loss": 0.0008,
+      "step": 7353
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.2211397954213343e-07,
+      "loss": 0.0006,
+      "step": 7354
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.220896249391135e-07,
+      "loss": 0.0005,
+      "step": 7355
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.220652703360935e-07,
+      "loss": 0.0007,
+      "step": 7356
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.2204091573307353e-07,
+      "loss": 0.0008,
+      "step": 7357
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.2201656113005353e-07,
+      "loss": 0.0007,
+      "step": 7358
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.219922065270336e-07,
+      "loss": 0.0008,
+      "step": 7359
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.219678519240136e-07,
+      "loss": 0.0009,
+      "step": 7360
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.2194349732099363e-07,
+      "loss": 0.0007,
+      "step": 7361
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 3.2191914271797363e-07,
+      "loss": 0.0007,
+      "step": 7362
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.218947881149537e-07,
+      "loss": 0.0007,
+      "step": 7363
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.2187043351193374e-07,
+      "loss": 0.0009,
+      "step": 7364
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.2184607890891373e-07,
+      "loss": 0.001,
+      "step": 7365
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.2182172430589384e-07,
+      "loss": 0.0007,
+      "step": 7366
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.2179736970287384e-07,
+      "loss": 0.0006,
+      "step": 7367
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.217730150998539e-07,
+      "loss": 0.0007,
+      "step": 7368
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.217486604968339e-07,
+      "loss": 0.0007,
+      "step": 7369
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.2172430589381394e-07,
+      "loss": 0.0009,
+      "step": 7370
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.21699951290794e-07,
+      "loss": 0.0008,
+      "step": 7371
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 3.21675596687774e-07,
+      "loss": 0.0005,
+      "step": 7372
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.2165124208475404e-07,
+      "loss": 0.0005,
+      "step": 7373
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.2162688748173404e-07,
+      "loss": 0.0007,
+      "step": 7374
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.216025328787141e-07,
+      "loss": 0.0008,
+      "step": 7375
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.215781782756941e-07,
+      "loss": 0.001,
+      "step": 7376
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.2155382367267414e-07,
+      "loss": 0.0006,
+      "step": 7377
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.215294690696542e-07,
+      "loss": 0.0007,
+      "step": 7378
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.215051144666342e-07,
+      "loss": 0.0006,
+      "step": 7379
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.2148075986361424e-07,
+      "loss": 0.0009,
+      "step": 7380
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.2145640526059424e-07,
+      "loss": 0.0006,
+      "step": 7381
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.214320506575743e-07,
+      "loss": 0.0005,
+      "step": 7382
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 3.214076960545543e-07,
+      "loss": 0.0006,
+      "step": 7383
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.2138334145153434e-07,
+      "loss": 0.001,
+      "step": 7384
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.2135898684851434e-07,
+      "loss": 0.0005,
+      "step": 7385
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.213346322454944e-07,
+      "loss": 0.0006,
+      "step": 7386
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.2131027764247445e-07,
+      "loss": 0.0007,
+      "step": 7387
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.2128592303945445e-07,
+      "loss": 0.0005,
+      "step": 7388
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.212615684364345e-07,
+      "loss": 0.0007,
+      "step": 7389
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.212372138334145e-07,
+      "loss": 0.0006,
+      "step": 7390
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.2121285923039455e-07,
+      "loss": 0.0015,
+      "step": 7391
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.2118850462737455e-07,
+      "loss": 0.0008,
+      "step": 7392
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 3.211641500243546e-07,
+      "loss": 0.0007,
+      "step": 7393
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.2113979542133465e-07,
+      "loss": 0.0008,
+      "step": 7394
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.2111544081831465e-07,
+      "loss": 0.0006,
+      "step": 7395
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.210910862152947e-07,
+      "loss": 0.0009,
+      "step": 7396
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.210667316122747e-07,
+      "loss": 0.0009,
+      "step": 7397
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.2104237700925475e-07,
+      "loss": 0.0004,
+      "step": 7398
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.2101802240623475e-07,
+      "loss": 0.0008,
+      "step": 7399
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.209936678032148e-07,
+      "loss": 0.0007,
+      "step": 7400
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.209693132001948e-07,
+      "loss": 0.0007,
+      "step": 7401
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.2094495859717485e-07,
+      "loss": 0.0034,
+      "step": 7402
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 3.209206039941549e-07,
+      "loss": 0.0078,
+      "step": 7403
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.208962493911349e-07,
+      "loss": 0.0007,
+      "step": 7404
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.2087189478811495e-07,
+      "loss": 0.0005,
+      "step": 7405
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.2084754018509495e-07,
+      "loss": 0.0008,
+      "step": 7406
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.20823185582075e-07,
+      "loss": 0.0007,
+      "step": 7407
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.20798830979055e-07,
+      "loss": 0.0005,
+      "step": 7408
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.2077447637603505e-07,
+      "loss": 0.0005,
+      "step": 7409
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.207501217730151e-07,
+      "loss": 0.0008,
+      "step": 7410
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.207257671699951e-07,
+      "loss": 0.0006,
+      "step": 7411
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.2070141256697515e-07,
+      "loss": 0.0009,
+      "step": 7412
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 3.2067705796395515e-07,
+      "loss": 0.0007,
+      "step": 7413
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.206527033609352e-07,
+      "loss": 0.0006,
+      "step": 7414
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.206283487579152e-07,
+      "loss": 0.0006,
+      "step": 7415
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.2060399415489526e-07,
+      "loss": 0.0008,
+      "step": 7416
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.2057963955187525e-07,
+      "loss": 0.0007,
+      "step": 7417
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.205552849488553e-07,
+      "loss": 0.0007,
+      "step": 7418
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.2053093034583536e-07,
+      "loss": 0.0005,
+      "step": 7419
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.2050657574281536e-07,
+      "loss": 0.0007,
+      "step": 7420
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.204822211397954e-07,
+      "loss": 0.0006,
+      "step": 7421
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.204578665367754e-07,
+      "loss": 0.0006,
+      "step": 7422
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.2043351193375546e-07,
+      "loss": 1.0097,
+      "step": 7423
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 3.2040915733073546e-07,
+      "loss": 0.0007,
+      "step": 7424
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.203848027277155e-07,
+      "loss": 0.0006,
+      "step": 7425
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.2036044812469556e-07,
+      "loss": 0.0007,
+      "step": 7426
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.2033609352167556e-07,
+      "loss": 0.0008,
+      "step": 7427
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.203117389186556e-07,
+      "loss": 0.001,
+      "step": 7428
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.202873843156356e-07,
+      "loss": 0.0013,
+      "step": 7429
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.2026302971261566e-07,
+      "loss": 0.0006,
+      "step": 7430
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.2023867510959566e-07,
+      "loss": 0.0008,
+      "step": 7431
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.202143205065757e-07,
+      "loss": 0.0008,
+      "step": 7432
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.2018996590355576e-07,
+      "loss": 0.0008,
+      "step": 7433
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 3.2016561130053576e-07,
+      "loss": 0.0005,
+      "step": 7434
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.201412566975158e-07,
+      "loss": 0.0005,
+      "step": 7435
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.201169020944958e-07,
+      "loss": 0.0009,
+      "step": 7436
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.2009254749147586e-07,
+      "loss": 0.0006,
+      "step": 7437
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.2006819288845586e-07,
+      "loss": 0.0006,
+      "step": 7438
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.200438382854359e-07,
+      "loss": 0.0008,
+      "step": 7439
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.200194836824159e-07,
+      "loss": 0.0007,
+      "step": 7440
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.1999512907939596e-07,
+      "loss": 0.0006,
+      "step": 7441
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.19970774476376e-07,
+      "loss": 0.0005,
+      "step": 7442
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.1994641987335607e-07,
+      "loss": 0.0006,
+      "step": 7443
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 3.199220652703361e-07,
+      "loss": 0.0007,
+      "step": 7444
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.198977106673161e-07,
+      "loss": 0.0007,
+      "step": 7445
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.1987335606429617e-07,
+      "loss": 0.0007,
+      "step": 7446
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.1984900146127617e-07,
+      "loss": 0.0006,
+      "step": 7447
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.198246468582562e-07,
+      "loss": 0.0008,
+      "step": 7448
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.1980029225523627e-07,
+      "loss": 0.0005,
+      "step": 7449
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.1977593765221627e-07,
+      "loss": 0.0009,
+      "step": 7450
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.197515830491963e-07,
+      "loss": 0.0007,
+      "step": 7451
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.197272284461763e-07,
+      "loss": 0.0007,
+      "step": 7452
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.1970287384315637e-07,
+      "loss": 0.0006,
+      "step": 7453
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.1967851924013637e-07,
+      "loss": 0.0006,
+      "step": 7454
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 3.196541646371164e-07,
+      "loss": 0.0006,
+      "step": 7455
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.196298100340965e-07,
+      "loss": 0.0008,
+      "step": 7456
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.1960545543107647e-07,
+      "loss": 0.0009,
+      "step": 7457
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.195811008280565e-07,
+      "loss": 0.0007,
+      "step": 7458
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.195567462250365e-07,
+      "loss": 0.0008,
+      "step": 7459
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.195323916220166e-07,
+      "loss": 0.8384,
+      "step": 7460
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.1950803701899657e-07,
+      "loss": 0.0007,
+      "step": 7461
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.194836824159766e-07,
+      "loss": 0.0007,
+      "step": 7462
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.194593278129566e-07,
+      "loss": 0.0004,
+      "step": 7463
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.194349732099367e-07,
+      "loss": 0.0007,
+      "step": 7464
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 3.1941061860691673e-07,
+      "loss": 0.0007,
+      "step": 7465
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.193862640038967e-07,
+      "loss": 0.0009,
+      "step": 7466
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.193619094008768e-07,
+      "loss": 0.0006,
+      "step": 7467
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.193375547978568e-07,
+      "loss": 0.0011,
+      "step": 7468
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.1931320019483683e-07,
+      "loss": 0.0005,
+      "step": 7469
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.192888455918168e-07,
+      "loss": 0.0006,
+      "step": 7470
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.192644909887969e-07,
+      "loss": 0.0006,
+      "step": 7471
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.1924013638577693e-07,
+      "loss": 0.001,
+      "step": 7472
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.1921578178275693e-07,
+      "loss": 0.0007,
+      "step": 7473
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.19191427179737e-07,
+      "loss": 0.0007,
+      "step": 7474
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 3.19167072576717e-07,
+      "loss": 0.0007,
+      "step": 7475
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.1914271797369703e-07,
+      "loss": 0.0007,
+      "step": 7476
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.1911836337067703e-07,
+      "loss": 0.0005,
+      "step": 7477
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.190940087676571e-07,
+      "loss": 0.0006,
+      "step": 7478
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.190696541646371e-07,
+      "loss": 0.0007,
+      "step": 7479
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.1904529956161713e-07,
+      "loss": 0.0007,
+      "step": 7480
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.190209449585972e-07,
+      "loss": 0.0009,
+      "step": 7481
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.189965903555772e-07,
+      "loss": 0.0007,
+      "step": 7482
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.1897223575255723e-07,
+      "loss": 0.0006,
+      "step": 7483
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.1894788114953723e-07,
+      "loss": 0.0006,
+      "step": 7484
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 3.189235265465173e-07,
+      "loss": 0.0007,
+      "step": 7485
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.188991719434973e-07,
+      "loss": 0.0009,
+      "step": 7486
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.1887481734047733e-07,
+      "loss": 0.0007,
+      "step": 7487
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.188504627374574e-07,
+      "loss": 0.0005,
+      "step": 7488
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.188261081344374e-07,
+      "loss": 0.0006,
+      "step": 7489
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.1880175353141744e-07,
+      "loss": 0.0008,
+      "step": 7490
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.1877739892839743e-07,
+      "loss": 0.0009,
+      "step": 7491
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.187530443253775e-07,
+      "loss": 0.0006,
+      "step": 7492
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.187286897223575e-07,
+      "loss": 0.0006,
+      "step": 7493
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.1870433511933754e-07,
+      "loss": 0.0007,
+      "step": 7494
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.1867998051631753e-07,
+      "loss": 0.1513,
+      "step": 7495
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 3.186556259132976e-07,
+      "loss": 0.0008,
+      "step": 7496
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.1863127131027764e-07,
+      "loss": 0.0008,
+      "step": 7497
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.1860691670725764e-07,
+      "loss": 0.0007,
+      "step": 7498
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.185825621042377e-07,
+      "loss": 0.0005,
+      "step": 7499
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.185582075012177e-07,
+      "loss": 0.0005,
+      "step": 7500
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.1853385289819774e-07,
+      "loss": 0.0007,
+      "step": 7501
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.1850949829517774e-07,
+      "loss": 0.0007,
+      "step": 7502
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.184851436921578e-07,
+      "loss": 0.0006,
+      "step": 7503
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.1846078908913784e-07,
+      "loss": 0.0005,
+      "step": 7504
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.1843643448611784e-07,
+      "loss": 0.0006,
+      "step": 7505
+    },
+    {
+      "epoch": 7.29,
+      "learning_rate": 3.184120798830979e-07,
+      "loss": 0.0005,
+      "step": 7506
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.183877252800779e-07,
+      "loss": 0.0005,
+      "step": 7507
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.1836337067705794e-07,
+      "loss": 0.0035,
+      "step": 7508
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.1833901607403794e-07,
+      "loss": 0.0006,
+      "step": 7509
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.18314661471018e-07,
+      "loss": 0.0006,
+      "step": 7510
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.18290306867998e-07,
+      "loss": 0.0005,
+      "step": 7511
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.1826595226497804e-07,
+      "loss": 0.0009,
+      "step": 7512
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.182415976619581e-07,
+      "loss": 0.0007,
+      "step": 7513
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.182172430589381e-07,
+      "loss": 0.0005,
+      "step": 7514
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.1819288845591814e-07,
+      "loss": 0.0006,
+      "step": 7515
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 3.1816853385289814e-07,
+      "loss": 0.0005,
+      "step": 7516
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.181441792498782e-07,
+      "loss": 0.0005,
+      "step": 7517
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.181198246468582e-07,
+      "loss": 0.0004,
+      "step": 7518
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.180954700438383e-07,
+      "loss": 0.0005,
+      "step": 7519
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.1807111544081835e-07,
+      "loss": 0.0006,
+      "step": 7520
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.1804676083779835e-07,
+      "loss": 0.0007,
+      "step": 7521
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.180224062347784e-07,
+      "loss": 0.0007,
+      "step": 7522
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.179980516317584e-07,
+      "loss": 0.0006,
+      "step": 7523
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.1797369702873845e-07,
+      "loss": 0.0007,
+      "step": 7524
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.1794934242571845e-07,
+      "loss": 0.0007,
+      "step": 7525
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.179249878226985e-07,
+      "loss": 0.0007,
+      "step": 7526
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 3.1790063321967855e-07,
+      "loss": 0.0006,
+      "step": 7527
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.1787627861665855e-07,
+      "loss": 0.0004,
+      "step": 7528
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.178519240136386e-07,
+      "loss": 0.0007,
+      "step": 7529
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.178275694106186e-07,
+      "loss": 0.0058,
+      "step": 7530
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.1780321480759865e-07,
+      "loss": 0.0007,
+      "step": 7531
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.1777886020457865e-07,
+      "loss": 0.0005,
+      "step": 7532
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.177545056015587e-07,
+      "loss": 0.0008,
+      "step": 7533
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.177301509985387e-07,
+      "loss": 0.0005,
+      "step": 7534
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.1770579639551875e-07,
+      "loss": 0.0005,
+      "step": 7535
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.176814417924988e-07,
+      "loss": 0.0007,
+      "step": 7536
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 3.176570871894788e-07,
+      "loss": 0.0005,
+      "step": 7537
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.1763273258645886e-07,
+      "loss": 0.0007,
+      "step": 7538
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.1760837798343885e-07,
+      "loss": 0.0005,
+      "step": 7539
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.175840233804189e-07,
+      "loss": 0.0006,
+      "step": 7540
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.175596687773989e-07,
+      "loss": 0.0007,
+      "step": 7541
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.1753531417437896e-07,
+      "loss": 0.0006,
+      "step": 7542
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.17510959571359e-07,
+      "loss": 0.0005,
+      "step": 7543
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.17486604968339e-07,
+      "loss": 0.0006,
+      "step": 7544
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.1746225036531906e-07,
+      "loss": 0.0007,
+      "step": 7545
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.1743789576229906e-07,
+      "loss": 0.0012,
+      "step": 7546
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 3.174135411592791e-07,
+      "loss": 0.0006,
+      "step": 7547
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.173891865562591e-07,
+      "loss": 0.0005,
+      "step": 7548
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.1736483195323916e-07,
+      "loss": 1.5284,
+      "step": 7549
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.1734047735021916e-07,
+      "loss": 0.0005,
+      "step": 7550
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.173161227471992e-07,
+      "loss": 0.0006,
+      "step": 7551
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.1729176814417926e-07,
+      "loss": 0.0007,
+      "step": 7552
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.1726741354115926e-07,
+      "loss": 0.0005,
+      "step": 7553
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.172430589381393e-07,
+      "loss": 0.0006,
+      "step": 7554
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.172187043351193e-07,
+      "loss": 0.0006,
+      "step": 7555
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.1719434973209936e-07,
+      "loss": 0.0008,
+      "step": 7556
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.1716999512907936e-07,
+      "loss": 0.0009,
+      "step": 7557
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 3.171456405260594e-07,
+      "loss": 0.0006,
+      "step": 7558
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.1712128592303946e-07,
+      "loss": 0.0009,
+      "step": 7559
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.1709693132001946e-07,
+      "loss": 0.0009,
+      "step": 7560
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.170725767169995e-07,
+      "loss": 0.0007,
+      "step": 7561
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.170482221139795e-07,
+      "loss": 0.0009,
+      "step": 7562
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.1702386751095956e-07,
+      "loss": 0.252,
+      "step": 7563
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.1699951290793956e-07,
+      "loss": 0.0006,
+      "step": 7564
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.169751583049196e-07,
+      "loss": 0.0005,
+      "step": 7565
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.169508037018996e-07,
+      "loss": 0.0006,
+      "step": 7566
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.1692644909887966e-07,
+      "loss": 0.0006,
+      "step": 7567
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 3.169020944958597e-07,
+      "loss": 0.0007,
+      "step": 7568
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.168777398928397e-07,
+      "loss": 0.0005,
+      "step": 7569
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.1685338528981977e-07,
+      "loss": 0.0008,
+      "step": 7570
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.1682903068679976e-07,
+      "loss": 0.0007,
+      "step": 7571
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.168046760837798e-07,
+      "loss": 0.0005,
+      "step": 7572
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.167803214807598e-07,
+      "loss": 0.0006,
+      "step": 7573
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.1675596687773987e-07,
+      "loss": 0.0012,
+      "step": 7574
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.167316122747199e-07,
+      "loss": 0.0006,
+      "step": 7575
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.167072576716999e-07,
+      "loss": 0.0005,
+      "step": 7576
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.1668290306867997e-07,
+      "loss": 0.0005,
+      "step": 7577
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 3.1665854846565997e-07,
+      "loss": 0.0007,
+      "step": 7578
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.1663419386264e-07,
+      "loss": 0.0006,
+      "step": 7579
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.1660983925962e-07,
+      "loss": 0.0005,
+      "step": 7580
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.1658548465660007e-07,
+      "loss": 0.0006,
+      "step": 7581
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.1656113005358007e-07,
+      "loss": 0.0029,
+      "step": 7582
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.165367754505601e-07,
+      "loss": 0.0006,
+      "step": 7583
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.1651242084754017e-07,
+      "loss": 0.0007,
+      "step": 7584
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.1648806624452017e-07,
+      "loss": 0.0006,
+      "step": 7585
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.164637116415002e-07,
+      "loss": 0.0007,
+      "step": 7586
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.164393570384802e-07,
+      "loss": 0.0008,
+      "step": 7587
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 3.1641500243546027e-07,
+      "loss": 0.0005,
+      "step": 7588
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.1639064783244027e-07,
+      "loss": 0.0007,
+      "step": 7589
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.163662932294203e-07,
+      "loss": 0.0006,
+      "step": 7590
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.163419386264004e-07,
+      "loss": 0.0007,
+      "step": 7591
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.1631758402338037e-07,
+      "loss": 0.0006,
+      "step": 7592
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.162932294203604e-07,
+      "loss": 0.0009,
+      "step": 7593
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.162688748173404e-07,
+      "loss": 0.203,
+      "step": 7594
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.162445202143205e-07,
+      "loss": 0.2297,
+      "step": 7595
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.162201656113005e-07,
+      "loss": 0.0006,
+      "step": 7596
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.161958110082806e-07,
+      "loss": 0.0005,
+      "step": 7597
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.1617145640526063e-07,
+      "loss": 0.0182,
+      "step": 7598
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 3.1614710180224063e-07,
+      "loss": 0.0008,
+      "step": 7599
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.161227471992207e-07,
+      "loss": 0.0005,
+      "step": 7600
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.160983925962007e-07,
+      "loss": 0.0009,
+      "step": 7601
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.1607403799318073e-07,
+      "loss": 0.0004,
+      "step": 7602
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.1604968339016073e-07,
+      "loss": 0.0006,
+      "step": 7603
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.160253287871408e-07,
+      "loss": 0.0007,
+      "step": 7604
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.160009741841208e-07,
+      "loss": 0.0007,
+      "step": 7605
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.1597661958110083e-07,
+      "loss": 0.0005,
+      "step": 7606
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.159522649780809e-07,
+      "loss": 0.0006,
+      "step": 7607
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.159279103750609e-07,
+      "loss": 0.0006,
+      "step": 7608
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 3.1590355577204093e-07,
+      "loss": 0.0005,
+      "step": 7609
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.1587920116902093e-07,
+      "loss": 0.0005,
+      "step": 7610
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.15854846566001e-07,
+      "loss": 0.0006,
+      "step": 7611
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.15830491962981e-07,
+      "loss": 0.0006,
+      "step": 7612
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.1580613735996103e-07,
+      "loss": 0.0005,
+      "step": 7613
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.157817827569411e-07,
+      "loss": 0.0007,
+      "step": 7614
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.157574281539211e-07,
+      "loss": 0.0006,
+      "step": 7615
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.1573307355090114e-07,
+      "loss": 0.0009,
+      "step": 7616
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.1570871894788113e-07,
+      "loss": 0.0006,
+      "step": 7617
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.156843643448612e-07,
+      "loss": 0.0008,
+      "step": 7618
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 3.156600097418412e-07,
+      "loss": 0.0007,
+      "step": 7619
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.1563565513882124e-07,
+      "loss": 0.0008,
+      "step": 7620
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.156113005358013e-07,
+      "loss": 0.0009,
+      "step": 7621
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.155869459327813e-07,
+      "loss": 0.0006,
+      "step": 7622
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.1556259132976134e-07,
+      "loss": 2.0114,
+      "step": 7623
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.1553823672674134e-07,
+      "loss": 0.0006,
+      "step": 7624
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.155138821237214e-07,
+      "loss": 0.0007,
+      "step": 7625
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.154895275207014e-07,
+      "loss": 0.0005,
+      "step": 7626
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.1546517291768144e-07,
+      "loss": 0.0006,
+      "step": 7627
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.1544081831466144e-07,
+      "loss": 0.0007,
+      "step": 7628
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.154164637116415e-07,
+      "loss": 0.0005,
+      "step": 7629
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 3.1539210910862154e-07,
+      "loss": 0.0005,
+      "step": 7630
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.1536775450560154e-07,
+      "loss": 0.0005,
+      "step": 7631
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.153433999025816e-07,
+      "loss": 0.0007,
+      "step": 7632
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.153190452995616e-07,
+      "loss": 0.0007,
+      "step": 7633
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.1529469069654164e-07,
+      "loss": 0.0006,
+      "step": 7634
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.1527033609352164e-07,
+      "loss": 0.001,
+      "step": 7635
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.152459814905017e-07,
+      "loss": 0.0013,
+      "step": 7636
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.1522162688748174e-07,
+      "loss": 0.0032,
+      "step": 7637
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.1519727228446174e-07,
+      "loss": 0.0026,
+      "step": 7638
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.151729176814418e-07,
+      "loss": 0.0019,
+      "step": 7639
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 3.151485630784218e-07,
+      "loss": 0.0005,
+      "step": 7640
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.1512420847540184e-07,
+      "loss": 0.0006,
+      "step": 7641
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.1509985387238184e-07,
+      "loss": 0.0005,
+      "step": 7642
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.150754992693619e-07,
+      "loss": 0.0007,
+      "step": 7643
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.150511446663419e-07,
+      "loss": 0.0005,
+      "step": 7644
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.1502679006332194e-07,
+      "loss": 0.0036,
+      "step": 7645
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.15002435460302e-07,
+      "loss": 0.0005,
+      "step": 7646
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.14978080857282e-07,
+      "loss": 0.0007,
+      "step": 7647
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.1495372625426205e-07,
+      "loss": 0.0005,
+      "step": 7648
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.1492937165124205e-07,
+      "loss": 0.0006,
+      "step": 7649
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 3.149050170482221e-07,
+      "loss": 0.0005,
+      "step": 7650
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.148806624452021e-07,
+      "loss": 0.0006,
+      "step": 7651
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.1485630784218215e-07,
+      "loss": 0.0006,
+      "step": 7652
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.148319532391622e-07,
+      "loss": 0.0006,
+      "step": 7653
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.148075986361422e-07,
+      "loss": 0.0145,
+      "step": 7654
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.1478324403312225e-07,
+      "loss": 0.0006,
+      "step": 7655
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.1475888943010225e-07,
+      "loss": 0.0007,
+      "step": 7656
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.147345348270823e-07,
+      "loss": 0.0005,
+      "step": 7657
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.147101802240623e-07,
+      "loss": 0.0006,
+      "step": 7658
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.1468582562104235e-07,
+      "loss": 0.0007,
+      "step": 7659
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 3.1466147101802235e-07,
+      "loss": 0.0006,
+      "step": 7660
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.146371164150024e-07,
+      "loss": 0.0014,
+      "step": 7661
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.1461276181198245e-07,
+      "loss": 0.0006,
+      "step": 7662
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.1458840720896245e-07,
+      "loss": 0.0008,
+      "step": 7663
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.145640526059425e-07,
+      "loss": 0.0007,
+      "step": 7664
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.145396980029225e-07,
+      "loss": 0.0006,
+      "step": 7665
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.1451534339990255e-07,
+      "loss": 0.001,
+      "step": 7666
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.1449098879688255e-07,
+      "loss": 0.0006,
+      "step": 7667
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.144666341938626e-07,
+      "loss": 1.9485,
+      "step": 7668
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.1444227959084265e-07,
+      "loss": 0.0005,
+      "step": 7669
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.1441792498782265e-07,
+      "loss": 0.0004,
+      "step": 7670
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 3.143935703848027e-07,
+      "loss": 0.0006,
+      "step": 7671
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.143692157817827e-07,
+      "loss": 0.0004,
+      "step": 7672
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.143448611787628e-07,
+      "loss": 0.0005,
+      "step": 7673
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.143205065757428e-07,
+      "loss": 0.0006,
+      "step": 7674
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.1429615197272286e-07,
+      "loss": 0.0005,
+      "step": 7675
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.142717973697029e-07,
+      "loss": 0.0006,
+      "step": 7676
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.142474427666829e-07,
+      "loss": 0.0006,
+      "step": 7677
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.1422308816366296e-07,
+      "loss": 0.0005,
+      "step": 7678
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.1419873356064296e-07,
+      "loss": 0.0006,
+      "step": 7679
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.14174378957623e-07,
+      "loss": 0.0093,
+      "step": 7680
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 3.14150024354603e-07,
+      "loss": 0.0006,
+      "step": 7681
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.1412566975158306e-07,
+      "loss": 0.0005,
+      "step": 7682
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.1410131514856306e-07,
+      "loss": 0.0006,
+      "step": 7683
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.140769605455431e-07,
+      "loss": 0.0006,
+      "step": 7684
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.1405260594252316e-07,
+      "loss": 0.0006,
+      "step": 7685
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.1402825133950316e-07,
+      "loss": 0.0005,
+      "step": 7686
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.140038967364832e-07,
+      "loss": 0.0006,
+      "step": 7687
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.139795421334632e-07,
+      "loss": 0.0006,
+      "step": 7688
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.1395518753044326e-07,
+      "loss": 0.0006,
+      "step": 7689
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.1393083292742326e-07,
+      "loss": 0.0006,
+      "step": 7690
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 3.139064783244033e-07,
+      "loss": 0.0008,
+      "step": 7691
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.1388212372138337e-07,
+      "loss": 0.0005,
+      "step": 7692
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.1385776911836336e-07,
+      "loss": 0.0008,
+      "step": 7693
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.138334145153434e-07,
+      "loss": 0.0008,
+      "step": 7694
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.138090599123234e-07,
+      "loss": 0.0007,
+      "step": 7695
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.1378470530930347e-07,
+      "loss": 0.0006,
+      "step": 7696
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.1376035070628347e-07,
+      "loss": 0.0006,
+      "step": 7697
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.137359961032635e-07,
+      "loss": 0.0005,
+      "step": 7698
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.137116415002435e-07,
+      "loss": 0.0006,
+      "step": 7699
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.1368728689722357e-07,
+      "loss": 0.0006,
+      "step": 7700
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.136629322942036e-07,
+      "loss": 0.0007,
+      "step": 7701
+    },
+    {
+      "epoch": 7.48,
+      "learning_rate": 3.136385776911836e-07,
+      "loss": 0.0005,
+      "step": 7702
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.1361422308816367e-07,
+      "loss": 0.0008,
+      "step": 7703
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.1358986848514367e-07,
+      "loss": 0.0006,
+      "step": 7704
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.135655138821237e-07,
+      "loss": 0.0007,
+      "step": 7705
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.135411592791037e-07,
+      "loss": 0.0006,
+      "step": 7706
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.1351680467608377e-07,
+      "loss": 0.0007,
+      "step": 7707
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.134924500730638e-07,
+      "loss": 0.0009,
+      "step": 7708
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.134680954700438e-07,
+      "loss": 0.0006,
+      "step": 7709
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.1344374086702387e-07,
+      "loss": 0.0005,
+      "step": 7710
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.1341938626400387e-07,
+      "loss": 0.0007,
+      "step": 7711
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 3.133950316609839e-07,
+      "loss": 0.0006,
+      "step": 7712
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.133706770579639e-07,
+      "loss": 0.0006,
+      "step": 7713
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.1334632245494397e-07,
+      "loss": 0.0007,
+      "step": 7714
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.1332196785192397e-07,
+      "loss": 0.0005,
+      "step": 7715
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.13297613248904e-07,
+      "loss": 0.0006,
+      "step": 7716
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.132732586458841e-07,
+      "loss": 0.0006,
+      "step": 7717
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.1324890404286407e-07,
+      "loss": 0.0006,
+      "step": 7718
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.132245494398441e-07,
+      "loss": 0.0006,
+      "step": 7719
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.132001948368241e-07,
+      "loss": 0.0005,
+      "step": 7720
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.131758402338042e-07,
+      "loss": 0.0005,
+      "step": 7721
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 3.131514856307842e-07,
+      "loss": 0.0006,
+      "step": 7722
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.131271310277642e-07,
+      "loss": 0.0007,
+      "step": 7723
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.131027764247443e-07,
+      "loss": 0.0006,
+      "step": 7724
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.130784218217243e-07,
+      "loss": 0.0007,
+      "step": 7725
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.1305406721870433e-07,
+      "loss": 0.0009,
+      "step": 7726
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.130297126156843e-07,
+      "loss": 0.0005,
+      "step": 7727
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.130053580126644e-07,
+      "loss": 0.0005,
+      "step": 7728
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.129810034096444e-07,
+      "loss": 0.0007,
+      "step": 7729
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.1295664880662443e-07,
+      "loss": 0.0006,
+      "step": 7730
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.1293229420360443e-07,
+      "loss": 0.0005,
+      "step": 7731
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 3.129079396005845e-07,
+      "loss": 0.0005,
+      "step": 7732
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.1288358499756453e-07,
+      "loss": 0.0005,
+      "step": 7733
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.1285923039454453e-07,
+      "loss": 0.0006,
+      "step": 7734
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.128348757915246e-07,
+      "loss": 0.0006,
+      "step": 7735
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.128105211885046e-07,
+      "loss": 0.0007,
+      "step": 7736
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.1278616658548463e-07,
+      "loss": 0.0006,
+      "step": 7737
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.1276181198246463e-07,
+      "loss": 0.0006,
+      "step": 7738
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.127374573794447e-07,
+      "loss": 0.001,
+      "step": 7739
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.1271310277642473e-07,
+      "loss": 0.0006,
+      "step": 7740
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.1268874817340473e-07,
+      "loss": 0.0006,
+      "step": 7741
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.126643935703848e-07,
+      "loss": 0.0007,
+      "step": 7742
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 3.126400389673648e-07,
+      "loss": 0.0006,
+      "step": 7743
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.1261568436434483e-07,
+      "loss": 0.0009,
+      "step": 7744
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.1259132976132483e-07,
+      "loss": 0.0005,
+      "step": 7745
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.125669751583049e-07,
+      "loss": 0.0006,
+      "step": 7746
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.125426205552849e-07,
+      "loss": 0.0006,
+      "step": 7747
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.1251826595226493e-07,
+      "loss": 0.0007,
+      "step": 7748
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.1249391134924504e-07,
+      "loss": 0.0005,
+      "step": 7749
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.1246955674622504e-07,
+      "loss": 0.0006,
+      "step": 7750
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.124452021432051e-07,
+      "loss": 0.0005,
+      "step": 7751
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.124208475401851e-07,
+      "loss": 0.0005,
+      "step": 7752
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 3.1239649293716514e-07,
+      "loss": 0.0006,
+      "step": 7753
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.1237213833414514e-07,
+      "loss": 0.0007,
+      "step": 7754
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.123477837311252e-07,
+      "loss": 0.0004,
+      "step": 7755
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.1232342912810524e-07,
+      "loss": 0.0005,
+      "step": 7756
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.1229907452508524e-07,
+      "loss": 0.0006,
+      "step": 7757
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.122747199220653e-07,
+      "loss": 0.0005,
+      "step": 7758
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.122503653190453e-07,
+      "loss": 0.0007,
+      "step": 7759
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.1222601071602534e-07,
+      "loss": 0.0005,
+      "step": 7760
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.1220165611300534e-07,
+      "loss": 0.0005,
+      "step": 7761
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.121773015099854e-07,
+      "loss": 0.0006,
+      "step": 7762
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 3.1215294690696544e-07,
+      "loss": 0.0006,
+      "step": 7763
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.1212859230394544e-07,
+      "loss": 0.0006,
+      "step": 7764
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.121042377009255e-07,
+      "loss": 0.0007,
+      "step": 7765
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.120798830979055e-07,
+      "loss": 0.0007,
+      "step": 7766
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.1205552849488554e-07,
+      "loss": 0.0192,
+      "step": 7767
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.1203117389186554e-07,
+      "loss": 0.0007,
+      "step": 7768
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.120068192888456e-07,
+      "loss": 0.0005,
+      "step": 7769
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.119824646858256e-07,
+      "loss": 0.0006,
+      "step": 7770
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.1195811008280565e-07,
+      "loss": 0.0007,
+      "step": 7771
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.119337554797857e-07,
+      "loss": 0.0007,
+      "step": 7772
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.119094008767657e-07,
+      "loss": 0.0005,
+      "step": 7773
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 3.1188504627374575e-07,
+      "loss": 0.0005,
+      "step": 7774
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.1186069167072575e-07,
+      "loss": 0.0007,
+      "step": 7775
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.118363370677058e-07,
+      "loss": 0.0005,
+      "step": 7776
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.118119824646858e-07,
+      "loss": 0.0006,
+      "step": 7777
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.1178762786166585e-07,
+      "loss": 0.0006,
+      "step": 7778
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.117632732586459e-07,
+      "loss": 0.0006,
+      "step": 7779
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.117389186556259e-07,
+      "loss": 0.0006,
+      "step": 7780
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.1171456405260595e-07,
+      "loss": 0.0005,
+      "step": 7781
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.1169020944958595e-07,
+      "loss": 0.173,
+      "step": 7782
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.11665854846566e-07,
+      "loss": 0.0007,
+      "step": 7783
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 3.11641500243546e-07,
+      "loss": 0.0007,
+      "step": 7784
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.1161714564052605e-07,
+      "loss": 0.0006,
+      "step": 7785
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.1159279103750605e-07,
+      "loss": 0.0004,
+      "step": 7786
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.115684364344861e-07,
+      "loss": 0.0005,
+      "step": 7787
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.1154408183146615e-07,
+      "loss": 0.0009,
+      "step": 7788
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.1151972722844615e-07,
+      "loss": 0.0007,
+      "step": 7789
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.114953726254262e-07,
+      "loss": 0.0005,
+      "step": 7790
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.114710180224062e-07,
+      "loss": 0.0006,
+      "step": 7791
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.1144666341938625e-07,
+      "loss": 0.0013,
+      "step": 7792
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.1142230881636625e-07,
+      "loss": 0.0005,
+      "step": 7793
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 3.113979542133463e-07,
+      "loss": 0.0007,
+      "step": 7794
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.1137359961032635e-07,
+      "loss": 0.011,
+      "step": 7795
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.1134924500730635e-07,
+      "loss": 0.0005,
+      "step": 7796
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.113248904042864e-07,
+      "loss": 0.0005,
+      "step": 7797
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.113005358012664e-07,
+      "loss": 0.0008,
+      "step": 7798
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.1127618119824646e-07,
+      "loss": 0.0005,
+      "step": 7799
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.1125182659522645e-07,
+      "loss": 0.0009,
+      "step": 7800
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.112274719922065e-07,
+      "loss": 0.0007,
+      "step": 7801
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.1120311738918656e-07,
+      "loss": 0.0006,
+      "step": 7802
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.1117876278616656e-07,
+      "loss": 0.0005,
+      "step": 7803
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 3.111544081831466e-07,
+      "loss": 0.0006,
+      "step": 7804
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.111300535801266e-07,
+      "loss": 0.0006,
+      "step": 7805
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.1110569897710666e-07,
+      "loss": 0.0007,
+      "step": 7806
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.1108134437408666e-07,
+      "loss": 0.0007,
+      "step": 7807
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.110569897710667e-07,
+      "loss": 0.0007,
+      "step": 7808
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.110326351680467e-07,
+      "loss": 0.0006,
+      "step": 7809
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.1100828056502676e-07,
+      "loss": 0.0005,
+      "step": 7810
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.109839259620068e-07,
+      "loss": 0.0007,
+      "step": 7811
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.109595713589868e-07,
+      "loss": 0.0006,
+      "step": 7812
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.1093521675596686e-07,
+      "loss": 0.0006,
+      "step": 7813
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.1091086215294686e-07,
+      "loss": 0.0011,
+      "step": 7814
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 3.108865075499269e-07,
+      "loss": 0.0005,
+      "step": 7815
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.108621529469069e-07,
+      "loss": 0.0006,
+      "step": 7816
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.1083779834388696e-07,
+      "loss": 0.0006,
+      "step": 7817
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.10813443740867e-07,
+      "loss": 0.0011,
+      "step": 7818
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.10789089137847e-07,
+      "loss": 0.0005,
+      "step": 7819
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.1076473453482706e-07,
+      "loss": 0.0008,
+      "step": 7820
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.1074037993180706e-07,
+      "loss": 0.0007,
+      "step": 7821
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.107160253287871e-07,
+      "loss": 0.0005,
+      "step": 7822
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.106916707257671e-07,
+      "loss": 0.0007,
+      "step": 7823
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.1066731612274716e-07,
+      "loss": 0.0006,
+      "step": 7824
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 3.1064296151972716e-07,
+      "loss": 0.0007,
+      "step": 7825
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.1061860691670727e-07,
+      "loss": 0.0005,
+      "step": 7826
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.105942523136873e-07,
+      "loss": 0.0005,
+      "step": 7827
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.105698977106673e-07,
+      "loss": 0.0006,
+      "step": 7828
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.1054554310764737e-07,
+      "loss": 0.0006,
+      "step": 7829
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.1052118850462737e-07,
+      "loss": 0.0008,
+      "step": 7830
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.104968339016074e-07,
+      "loss": 0.0006,
+      "step": 7831
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.104724792985874e-07,
+      "loss": 0.0006,
+      "step": 7832
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.1044812469556747e-07,
+      "loss": 0.0008,
+      "step": 7833
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.104237700925475e-07,
+      "loss": 0.0006,
+      "step": 7834
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 3.103994154895275e-07,
+      "loss": 0.0007,
+      "step": 7835
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.1037506088650757e-07,
+      "loss": 0.0006,
+      "step": 7836
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.1035070628348757e-07,
+      "loss": 1.0984,
+      "step": 7837
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.103263516804676e-07,
+      "loss": 0.0028,
+      "step": 7838
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.103019970774476e-07,
+      "loss": 0.0231,
+      "step": 7839
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.1027764247442767e-07,
+      "loss": 0.0008,
+      "step": 7840
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.102532878714077e-07,
+      "loss": 0.0008,
+      "step": 7841
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.102289332683877e-07,
+      "loss": 0.0006,
+      "step": 7842
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.102045786653678e-07,
+      "loss": 0.0007,
+      "step": 7843
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.1018022406234777e-07,
+      "loss": 0.0008,
+      "step": 7844
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.101558694593278e-07,
+      "loss": 0.0006,
+      "step": 7845
+    },
+    {
+      "epoch": 7.62,
+      "learning_rate": 3.101315148563078e-07,
+      "loss": 0.0006,
+      "step": 7846
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.101071602532879e-07,
+      "loss": 0.0005,
+      "step": 7847
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.100828056502679e-07,
+      "loss": 0.0005,
+      "step": 7848
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.100584510472479e-07,
+      "loss": 0.0005,
+      "step": 7849
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.10034096444228e-07,
+      "loss": 0.0005,
+      "step": 7850
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.10009741841208e-07,
+      "loss": 0.0008,
+      "step": 7851
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.0998538723818803e-07,
+      "loss": 0.0006,
+      "step": 7852
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.09961032635168e-07,
+      "loss": 0.0005,
+      "step": 7853
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.099366780321481e-07,
+      "loss": 0.0006,
+      "step": 7854
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.099123234291281e-07,
+      "loss": 0.0007,
+      "step": 7855
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 3.0988796882610813e-07,
+      "loss": 0.0006,
+      "step": 7856
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.098636142230882e-07,
+      "loss": 0.0006,
+      "step": 7857
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.098392596200682e-07,
+      "loss": 0.0008,
+      "step": 7858
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.0981490501704823e-07,
+      "loss": 0.0038,
+      "step": 7859
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.0979055041402823e-07,
+      "loss": 0.0006,
+      "step": 7860
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.097661958110083e-07,
+      "loss": 0.0543,
+      "step": 7861
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.097418412079883e-07,
+      "loss": 0.0006,
+      "step": 7862
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.0971748660496833e-07,
+      "loss": 0.0006,
+      "step": 7863
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.0969313200194833e-07,
+      "loss": 0.0008,
+      "step": 7864
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.096687773989284e-07,
+      "loss": 0.0004,
+      "step": 7865
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 3.0964442279590843e-07,
+      "loss": 0.0005,
+      "step": 7866
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.0962006819288843e-07,
+      "loss": 0.0006,
+      "step": 7867
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.095957135898685e-07,
+      "loss": 0.0009,
+      "step": 7868
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.095713589868485e-07,
+      "loss": 0.0006,
+      "step": 7869
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.0954700438382853e-07,
+      "loss": 0.0006,
+      "step": 7870
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.0952264978080853e-07,
+      "loss": 0.0006,
+      "step": 7871
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.094982951777886e-07,
+      "loss": 0.0005,
+      "step": 7872
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.0947394057476864e-07,
+      "loss": 0.0009,
+      "step": 7873
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.0944958597174863e-07,
+      "loss": 0.0005,
+      "step": 7874
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.094252313687287e-07,
+      "loss": 0.0005,
+      "step": 7875
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 3.094008767657087e-07,
+      "loss": 0.0005,
+      "step": 7876
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.0937652216268874e-07,
+      "loss": 0.0007,
+      "step": 7877
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.0935216755966873e-07,
+      "loss": 0.0006,
+      "step": 7878
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.093278129566488e-07,
+      "loss": 0.0005,
+      "step": 7879
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.093034583536288e-07,
+      "loss": 0.0005,
+      "step": 7880
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.0927910375060884e-07,
+      "loss": 0.0006,
+      "step": 7881
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.092547491475889e-07,
+      "loss": 0.0019,
+      "step": 7882
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.092303945445689e-07,
+      "loss": 0.0007,
+      "step": 7883
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.0920603994154894e-07,
+      "loss": 0.0005,
+      "step": 7884
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.0918168533852894e-07,
+      "loss": 0.0006,
+      "step": 7885
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.09157330735509e-07,
+      "loss": 0.0006,
+      "step": 7886
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 3.09132976132489e-07,
+      "loss": 0.0006,
+      "step": 7887
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.0910862152946904e-07,
+      "loss": 0.0329,
+      "step": 7888
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.090842669264491e-07,
+      "loss": 0.0005,
+      "step": 7889
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.090599123234291e-07,
+      "loss": 0.0007,
+      "step": 7890
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.0903555772040914e-07,
+      "loss": 0.0005,
+      "step": 7891
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.0901120311738914e-07,
+      "loss": 0.0006,
+      "step": 7892
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.089868485143692e-07,
+      "loss": 0.0007,
+      "step": 7893
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.089624939113492e-07,
+      "loss": 0.0005,
+      "step": 7894
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.0893813930832924e-07,
+      "loss": 0.0004,
+      "step": 7895
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.0891378470530924e-07,
+      "loss": 0.0007,
+      "step": 7896
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 3.088894301022893e-07,
+      "loss": 0.0005,
+      "step": 7897
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.0886507549926934e-07,
+      "loss": 0.0006,
+      "step": 7898
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.0884072089624934e-07,
+      "loss": 0.0006,
+      "step": 7899
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.088163662932294e-07,
+      "loss": 0.0006,
+      "step": 7900
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.087920116902094e-07,
+      "loss": 0.0008,
+      "step": 7901
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.087676570871895e-07,
+      "loss": 0.0005,
+      "step": 7902
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.087433024841695e-07,
+      "loss": 0.0004,
+      "step": 7903
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.0871894788114955e-07,
+      "loss": 0.0006,
+      "step": 7904
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.086945932781296e-07,
+      "loss": 0.0008,
+      "step": 7905
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.086702386751096e-07,
+      "loss": 0.0005,
+      "step": 7906
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 3.0864588407208965e-07,
+      "loss": 0.0006,
+      "step": 7907
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.0862152946906965e-07,
+      "loss": 0.0004,
+      "step": 7908
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.085971748660497e-07,
+      "loss": 0.0006,
+      "step": 7909
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.085728202630297e-07,
+      "loss": 0.0005,
+      "step": 7910
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.0854846566000975e-07,
+      "loss": 0.0006,
+      "step": 7911
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.085241110569898e-07,
+      "loss": 0.0005,
+      "step": 7912
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.084997564539698e-07,
+      "loss": 1.9179,
+      "step": 7913
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.0847540185094985e-07,
+      "loss": 0.0007,
+      "step": 7914
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.0845104724792985e-07,
+      "loss": 0.0006,
+      "step": 7915
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.084266926449099e-07,
+      "loss": 0.0006,
+      "step": 7916
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.084023380418899e-07,
+      "loss": 0.0005,
+      "step": 7917
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 3.0837798343886995e-07,
+      "loss": 0.0005,
+      "step": 7918
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.0835362883584995e-07,
+      "loss": 0.0005,
+      "step": 7919
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.0832927423283e-07,
+      "loss": 0.0006,
+      "step": 7920
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.0830491962981006e-07,
+      "loss": 0.0007,
+      "step": 7921
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.0828056502679005e-07,
+      "loss": 0.0005,
+      "step": 7922
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.082562104237701e-07,
+      "loss": 0.0007,
+      "step": 7923
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.082318558207501e-07,
+      "loss": 0.0005,
+      "step": 7924
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.0820750121773016e-07,
+      "loss": 0.0007,
+      "step": 7925
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.0818314661471015e-07,
+      "loss": 0.0007,
+      "step": 7926
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.081587920116902e-07,
+      "loss": 0.0007,
+      "step": 7927
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 3.0813443740867026e-07,
+      "loss": 0.0006,
+      "step": 7928
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.0811008280565026e-07,
+      "loss": 0.001,
+      "step": 7929
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.080857282026303e-07,
+      "loss": 0.0006,
+      "step": 7930
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.080613735996103e-07,
+      "loss": 0.0006,
+      "step": 7931
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.0803701899659036e-07,
+      "loss": 0.0007,
+      "step": 7932
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.0801266439357036e-07,
+      "loss": 0.0007,
+      "step": 7933
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.079883097905504e-07,
+      "loss": 0.0006,
+      "step": 7934
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.079639551875304e-07,
+      "loss": 0.0006,
+      "step": 7935
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.0793960058451046e-07,
+      "loss": 0.0005,
+      "step": 7936
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.079152459814905e-07,
+      "loss": 0.0005,
+      "step": 7937
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 3.078908913784705e-07,
+      "loss": 0.0006,
+      "step": 7938
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.0786653677545056e-07,
+      "loss": 0.0005,
+      "step": 7939
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.0784218217243056e-07,
+      "loss": 0.0005,
+      "step": 7940
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.078178275694106e-07,
+      "loss": 0.0006,
+      "step": 7941
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.077934729663906e-07,
+      "loss": 0.0006,
+      "step": 7942
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.0776911836337066e-07,
+      "loss": 0.0007,
+      "step": 7943
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.077447637603507e-07,
+      "loss": 0.0007,
+      "step": 7944
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.077204091573307e-07,
+      "loss": 0.0008,
+      "step": 7945
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.0769605455431076e-07,
+      "loss": 0.0006,
+      "step": 7946
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.0767169995129076e-07,
+      "loss": 0.0005,
+      "step": 7947
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.076473453482708e-07,
+      "loss": 0.0007,
+      "step": 7948
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 3.076229907452508e-07,
+      "loss": 0.0004,
+      "step": 7949
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.0759863614223086e-07,
+      "loss": 0.0006,
+      "step": 7950
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.0757428153921086e-07,
+      "loss": 0.0006,
+      "step": 7951
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.075499269361909e-07,
+      "loss": 0.0005,
+      "step": 7952
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.0752557233317097e-07,
+      "loss": 0.0006,
+      "step": 7953
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.0750121773015096e-07,
+      "loss": 0.0006,
+      "step": 7954
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.07476863127131e-07,
+      "loss": 0.0005,
+      "step": 7955
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.07452508524111e-07,
+      "loss": 0.0006,
+      "step": 7956
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.0742815392109107e-07,
+      "loss": 0.0006,
+      "step": 7957
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.0740379931807107e-07,
+      "loss": 0.0004,
+      "step": 7958
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 3.073794447150511e-07,
+      "loss": 0.002,
+      "step": 7959
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.0735509011203117e-07,
+      "loss": 0.0007,
+      "step": 7960
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.0733073550901117e-07,
+      "loss": 0.0006,
+      "step": 7961
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.073063809059912e-07,
+      "loss": 0.0008,
+      "step": 7962
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.072820263029712e-07,
+      "loss": 0.0005,
+      "step": 7963
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.0725767169995127e-07,
+      "loss": 0.001,
+      "step": 7964
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.0723331709693127e-07,
+      "loss": 0.0006,
+      "step": 7965
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.072089624939113e-07,
+      "loss": 0.0005,
+      "step": 7966
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.0718460789089137e-07,
+      "loss": 0.0008,
+      "step": 7967
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.0716025328787137e-07,
+      "loss": 0.0005,
+      "step": 7968
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 3.071358986848514e-07,
+      "loss": 0.0005,
+      "step": 7969
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.071115440818314e-07,
+      "loss": 0.0006,
+      "step": 7970
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.0708718947881147e-07,
+      "loss": 0.0006,
+      "step": 7971
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.0706283487579147e-07,
+      "loss": 0.0005,
+      "step": 7972
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.070384802727715e-07,
+      "loss": 0.0006,
+      "step": 7973
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.070141256697515e-07,
+      "loss": 0.0005,
+      "step": 7974
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.0698977106673157e-07,
+      "loss": 0.0005,
+      "step": 7975
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.069654164637116e-07,
+      "loss": 0.0006,
+      "step": 7976
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.069410618606916e-07,
+      "loss": 0.0008,
+      "step": 7977
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.069167072576717e-07,
+      "loss": 0.0004,
+      "step": 7978
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 3.068923526546517e-07,
+      "loss": 0.0007,
+      "step": 7979
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.068679980516318e-07,
+      "loss": 0.0008,
+      "step": 7980
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.068436434486118e-07,
+      "loss": 0.0006,
+      "step": 7981
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.0681928884559183e-07,
+      "loss": 0.0005,
+      "step": 7982
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.067949342425719e-07,
+      "loss": 0.775,
+      "step": 7983
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.067705796395519e-07,
+      "loss": 0.0009,
+      "step": 7984
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.0674622503653193e-07,
+      "loss": 0.0006,
+      "step": 7985
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.0672187043351193e-07,
+      "loss": 0.0006,
+      "step": 7986
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.06697515830492e-07,
+      "loss": 0.0005,
+      "step": 7987
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.06673161227472e-07,
+      "loss": 0.0007,
+      "step": 7988
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.0664880662445203e-07,
+      "loss": 0.0005,
+      "step": 7989
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 3.066244520214321e-07,
+      "loss": 0.0008,
+      "step": 7990
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.066000974184121e-07,
+      "loss": 0.0004,
+      "step": 7991
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.0657574281539213e-07,
+      "loss": 0.0006,
+      "step": 7992
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.0655138821237213e-07,
+      "loss": 0.0005,
+      "step": 7993
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.065270336093522e-07,
+      "loss": 0.0005,
+      "step": 7994
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.065026790063322e-07,
+      "loss": 0.0005,
+      "step": 7995
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.0647832440331223e-07,
+      "loss": 0.0006,
+      "step": 7996
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.0645396980029223e-07,
+      "loss": 0.0006,
+      "step": 7997
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.064296151972723e-07,
+      "loss": 0.0006,
+      "step": 7998
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.0640526059425234e-07,
+      "loss": 0.0005,
+      "step": 7999
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 3.0638090599123233e-07,
+      "loss": 0.0005,
+      "step": 8000
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.063565513882124e-07,
+      "loss": 0.0005,
+      "step": 8001
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.063321967851924e-07,
+      "loss": 0.0005,
+      "step": 8002
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.0630784218217244e-07,
+      "loss": 0.0008,
+      "step": 8003
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.0628348757915243e-07,
+      "loss": 0.0005,
+      "step": 8004
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.062591329761325e-07,
+      "loss": 0.0007,
+      "step": 8005
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.0623477837311254e-07,
+      "loss": 0.0006,
+      "step": 8006
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.0621042377009254e-07,
+      "loss": 0.0008,
+      "step": 8007
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.061860691670726e-07,
+      "loss": 0.0006,
+      "step": 8008
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.061617145640526e-07,
+      "loss": 0.0005,
+      "step": 8009
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 3.0613735996103264e-07,
+      "loss": 0.0008,
+      "step": 8010
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.0611300535801264e-07,
+      "loss": 0.0005,
+      "step": 8011
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.060886507549927e-07,
+      "loss": 0.0005,
+      "step": 8012
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.060642961519727e-07,
+      "loss": 0.0005,
+      "step": 8013
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.0603994154895274e-07,
+      "loss": 0.0005,
+      "step": 8014
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.060155869459328e-07,
+      "loss": 0.0007,
+      "step": 8015
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.059912323429128e-07,
+      "loss": 0.001,
+      "step": 8016
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.0596687773989284e-07,
+      "loss": 0.0011,
+      "step": 8017
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.0594252313687284e-07,
+      "loss": 0.0005,
+      "step": 8018
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.059181685338529e-07,
+      "loss": 0.0004,
+      "step": 8019
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.058938139308329e-07,
+      "loss": 0.0006,
+      "step": 8020
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 3.0586945932781294e-07,
+      "loss": 0.0005,
+      "step": 8021
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.05845104724793e-07,
+      "loss": 0.0006,
+      "step": 8022
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.05820750121773e-07,
+      "loss": 0.0004,
+      "step": 8023
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.0579639551875304e-07,
+      "loss": 0.0005,
+      "step": 8024
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.0577204091573304e-07,
+      "loss": 0.0005,
+      "step": 8025
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.057476863127131e-07,
+      "loss": 0.0004,
+      "step": 8026
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.057233317096931e-07,
+      "loss": 0.0005,
+      "step": 8027
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.0569897710667314e-07,
+      "loss": 0.0006,
+      "step": 8028
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.0567462250365314e-07,
+      "loss": 0.0005,
+      "step": 8029
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.056502679006332e-07,
+      "loss": 0.0005,
+      "step": 8030
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 3.0562591329761325e-07,
+      "loss": 0.0006,
+      "step": 8031
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.0560155869459325e-07,
+      "loss": 0.0029,
+      "step": 8032
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.055772040915733e-07,
+      "loss": 0.0007,
+      "step": 8033
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.055528494885533e-07,
+      "loss": 0.0005,
+      "step": 8034
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.0552849488553335e-07,
+      "loss": 0.0528,
+      "step": 8035
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.0550414028251335e-07,
+      "loss": 0.0006,
+      "step": 8036
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.054797856794934e-07,
+      "loss": 0.0007,
+      "step": 8037
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.0545543107647345e-07,
+      "loss": 0.0006,
+      "step": 8038
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.0543107647345345e-07,
+      "loss": 0.0005,
+      "step": 8039
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.054067218704335e-07,
+      "loss": 0.0005,
+      "step": 8040
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 3.053823672674135e-07,
+      "loss": 0.0007,
+      "step": 8041
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.0535801266439355e-07,
+      "loss": 0.0005,
+      "step": 8042
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.0533365806137355e-07,
+      "loss": 0.0007,
+      "step": 8043
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.053093034583536e-07,
+      "loss": 0.0012,
+      "step": 8044
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.052849488553336e-07,
+      "loss": 0.0008,
+      "step": 8045
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.0526059425231365e-07,
+      "loss": 0.0004,
+      "step": 8046
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.052362396492937e-07,
+      "loss": 0.0005,
+      "step": 8047
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.052118850462737e-07,
+      "loss": 0.8404,
+      "step": 8048
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.0518753044325375e-07,
+      "loss": 0.0006,
+      "step": 8049
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.0516317584023375e-07,
+      "loss": 0.0004,
+      "step": 8050
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 3.051388212372138e-07,
+      "loss": 0.0006,
+      "step": 8051
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.051144666341938e-07,
+      "loss": 0.0006,
+      "step": 8052
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.0509011203117385e-07,
+      "loss": 0.0008,
+      "step": 8053
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.050657574281539e-07,
+      "loss": 0.0006,
+      "step": 8054
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.050414028251339e-07,
+      "loss": 0.0007,
+      "step": 8055
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.05017048222114e-07,
+      "loss": 0.0004,
+      "step": 8056
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.04992693619094e-07,
+      "loss": 0.0008,
+      "step": 8057
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.0496833901607406e-07,
+      "loss": 0.0005,
+      "step": 8058
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.0494398441305406e-07,
+      "loss": 0.0006,
+      "step": 8059
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.049196298100341e-07,
+      "loss": 0.0005,
+      "step": 8060
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.0489527520701416e-07,
+      "loss": 0.2207,
+      "step": 8061
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 3.0487092060399416e-07,
+      "loss": 0.0006,
+      "step": 8062
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.048465660009742e-07,
+      "loss": 0.0005,
+      "step": 8063
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.048222113979542e-07,
+      "loss": 0.0046,
+      "step": 8064
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.0479785679493426e-07,
+      "loss": 0.0006,
+      "step": 8065
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.0477350219191426e-07,
+      "loss": 0.0007,
+      "step": 8066
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.047491475888943e-07,
+      "loss": 0.0007,
+      "step": 8067
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.047247929858743e-07,
+      "loss": 0.0007,
+      "step": 8068
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.0470043838285436e-07,
+      "loss": 0.0005,
+      "step": 8069
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.046760837798344e-07,
+      "loss": 0.0005,
+      "step": 8070
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.046517291768144e-07,
+      "loss": 0.0005,
+      "step": 8071
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 3.0462737457379446e-07,
+      "loss": 0.0004,
+      "step": 8072
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.0460301997077446e-07,
+      "loss": 0.0007,
+      "step": 8073
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.045786653677545e-07,
+      "loss": 0.0004,
+      "step": 8074
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.045543107647345e-07,
+      "loss": 0.0006,
+      "step": 8075
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.0452995616171456e-07,
+      "loss": 0.0005,
+      "step": 8076
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.045056015586946e-07,
+      "loss": 0.0004,
+      "step": 8077
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.044812469556746e-07,
+      "loss": 0.0007,
+      "step": 8078
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.0445689235265467e-07,
+      "loss": 0.0005,
+      "step": 8079
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.0443253774963467e-07,
+      "loss": 0.0006,
+      "step": 8080
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.044081831466147e-07,
+      "loss": 0.0006,
+      "step": 8081
+    },
+    {
+      "epoch": 7.85,
+      "learning_rate": 3.043838285435947e-07,
+      "loss": 0.0005,
+      "step": 8082
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.0435947394057477e-07,
+      "loss": 0.0006,
+      "step": 8083
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.0433511933755477e-07,
+      "loss": 0.0005,
+      "step": 8084
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.043107647345348e-07,
+      "loss": 0.0006,
+      "step": 8085
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.0428641013151487e-07,
+      "loss": 0.0033,
+      "step": 8086
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.0426205552849487e-07,
+      "loss": 0.0007,
+      "step": 8087
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.042377009254749e-07,
+      "loss": 0.0006,
+      "step": 8088
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.042133463224549e-07,
+      "loss": 0.1576,
+      "step": 8089
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.0418899171943497e-07,
+      "loss": 0.0005,
+      "step": 8090
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.0416463711641497e-07,
+      "loss": 0.0006,
+      "step": 8091
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.04140282513395e-07,
+      "loss": 0.0004,
+      "step": 8092
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 3.0411592791037507e-07,
+      "loss": 0.0005,
+      "step": 8093
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.0409157330735507e-07,
+      "loss": 0.0005,
+      "step": 8094
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.040672187043351e-07,
+      "loss": 0.0006,
+      "step": 8095
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.040428641013151e-07,
+      "loss": 0.0006,
+      "step": 8096
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.0401850949829517e-07,
+      "loss": 0.0007,
+      "step": 8097
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.0399415489527517e-07,
+      "loss": 0.0007,
+      "step": 8098
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.039698002922552e-07,
+      "loss": 0.0005,
+      "step": 8099
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.039454456892352e-07,
+      "loss": 0.0005,
+      "step": 8100
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.0392109108621527e-07,
+      "loss": 0.2482,
+      "step": 8101
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.038967364831953e-07,
+      "loss": 0.0006,
+      "step": 8102
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 3.038723818801753e-07,
+      "loss": 0.0015,
+      "step": 8103
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.038480272771554e-07,
+      "loss": 0.0005,
+      "step": 8104
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.038236726741354e-07,
+      "loss": 0.0006,
+      "step": 8105
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.037993180711154e-07,
+      "loss": 0.0006,
+      "step": 8106
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.037749634680954e-07,
+      "loss": 0.0004,
+      "step": 8107
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.037506088650755e-07,
+      "loss": 0.0004,
+      "step": 8108
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.0372625426205553e-07,
+      "loss": 0.0006,
+      "step": 8109
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.037018996590355e-07,
+      "loss": 0.0005,
+      "step": 8110
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.036775450560156e-07,
+      "loss": 0.0004,
+      "step": 8111
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.036531904529956e-07,
+      "loss": 0.0006,
+      "step": 8112
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 3.0362883584997563e-07,
+      "loss": 0.0005,
+      "step": 8113
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.0360448124695563e-07,
+      "loss": 0.0006,
+      "step": 8114
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.035801266439357e-07,
+      "loss": 0.0006,
+      "step": 8115
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.035557720409157e-07,
+      "loss": 0.0005,
+      "step": 8116
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.0353141743789573e-07,
+      "loss": 0.0007,
+      "step": 8117
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.035070628348758e-07,
+      "loss": 0.0006,
+      "step": 8118
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.034827082318558e-07,
+      "loss": 0.0006,
+      "step": 8119
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.0345835362883583e-07,
+      "loss": 0.0006,
+      "step": 8120
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.0343399902581583e-07,
+      "loss": 0.0007,
+      "step": 8121
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.034096444227959e-07,
+      "loss": 0.0006,
+      "step": 8122
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 3.033852898197759e-07,
+      "loss": 0.0006,
+      "step": 8123
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.0336093521675593e-07,
+      "loss": 0.0006,
+      "step": 8124
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.03336580613736e-07,
+      "loss": 0.0004,
+      "step": 8125
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.03312226010716e-07,
+      "loss": 0.0004,
+      "step": 8126
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.0328787140769603e-07,
+      "loss": 0.0005,
+      "step": 8127
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.0326351680467603e-07,
+      "loss": 0.0004,
+      "step": 8128
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.032391622016561e-07,
+      "loss": 0.0004,
+      "step": 8129
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.032148075986361e-07,
+      "loss": 0.0006,
+      "step": 8130
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.0319045299561613e-07,
+      "loss": 0.0006,
+      "step": 8131
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.0316609839259624e-07,
+      "loss": 0.0007,
+      "step": 8132
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.0314174378957624e-07,
+      "loss": 0.0006,
+      "step": 8133
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 3.031173891865563e-07,
+      "loss": 0.0005,
+      "step": 8134
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.030930345835363e-07,
+      "loss": 0.0006,
+      "step": 8135
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.0306867998051634e-07,
+      "loss": 0.0005,
+      "step": 8136
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.0304432537749634e-07,
+      "loss": 0.0006,
+      "step": 8137
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.030199707744764e-07,
+      "loss": 0.0005,
+      "step": 8138
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.029956161714564e-07,
+      "loss": 0.0006,
+      "step": 8139
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.0297126156843644e-07,
+      "loss": 0.0005,
+      "step": 8140
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.029469069654165e-07,
+      "loss": 0.0006,
+      "step": 8141
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.029225523623965e-07,
+      "loss": 0.0006,
+      "step": 8142
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.0289819775937654e-07,
+      "loss": 0.0005,
+      "step": 8143
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 3.0287384315635654e-07,
+      "loss": 0.0007,
+      "step": 8144
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.028494885533366e-07,
+      "loss": 0.0008,
+      "step": 8145
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.028251339503166e-07,
+      "loss": 0.0005,
+      "step": 8146
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.0280077934729664e-07,
+      "loss": 0.0004,
+      "step": 8147
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.027764247442767e-07,
+      "loss": 0.0007,
+      "step": 8148
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.027520701412567e-07,
+      "loss": 0.0005,
+      "step": 8149
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.0272771553823674e-07,
+      "loss": 0.9711,
+      "step": 8150
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.0270336093521674e-07,
+      "loss": 0.0006,
+      "step": 8151
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.026790063321968e-07,
+      "loss": 1.6395,
+      "step": 8152
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.026546517291768e-07,
+      "loss": 0.0005,
+      "step": 8153
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 3.0263029712615685e-07,
+      "loss": 0.0004,
+      "step": 8154
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.0260594252313684e-07,
+      "loss": 0.001,
+      "step": 8155
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.025815879201169e-07,
+      "loss": 0.0008,
+      "step": 8156
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.0255723331709695e-07,
+      "loss": 0.0013,
+      "step": 8157
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.0253287871407695e-07,
+      "loss": 0.0005,
+      "step": 8158
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.02508524111057e-07,
+      "loss": 0.0005,
+      "step": 8159
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.02484169508037e-07,
+      "loss": 0.0006,
+      "step": 8160
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.0245981490501705e-07,
+      "loss": 0.0068,
+      "step": 8161
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.0243546030199705e-07,
+      "loss": 0.0006,
+      "step": 8162
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.024111056989771e-07,
+      "loss": 0.0005,
+      "step": 8163
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.0238675109595715e-07,
+      "loss": 0.0006,
+      "step": 8164
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 3.0236239649293715e-07,
+      "loss": 0.0005,
+      "step": 8165
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.023380418899172e-07,
+      "loss": 0.0006,
+      "step": 8166
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.023136872868972e-07,
+      "loss": 0.0005,
+      "step": 8167
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.0228933268387725e-07,
+      "loss": 0.0005,
+      "step": 8168
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.0226497808085725e-07,
+      "loss": 0.0007,
+      "step": 8169
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.022406234778373e-07,
+      "loss": 0.0005,
+      "step": 8170
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.0221626887481735e-07,
+      "loss": 0.0012,
+      "step": 8171
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.0219191427179735e-07,
+      "loss": 0.0004,
+      "step": 8172
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.021675596687774e-07,
+      "loss": 0.0006,
+      "step": 8173
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.021432050657574e-07,
+      "loss": 0.0008,
+      "step": 8174
+    },
+    {
+      "epoch": 7.94,
+      "learning_rate": 3.0211885046273745e-07,
+      "loss": 0.0005,
+      "step": 8175
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.0209449585971745e-07,
+      "loss": 0.0007,
+      "step": 8176
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.020701412566975e-07,
+      "loss": 0.0005,
+      "step": 8177
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.020457866536775e-07,
+      "loss": 0.0005,
+      "step": 8178
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.0202143205065755e-07,
+      "loss": 0.0008,
+      "step": 8179
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.019970774476376e-07,
+      "loss": 0.0004,
+      "step": 8180
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.019727228446176e-07,
+      "loss": 0.0005,
+      "step": 8181
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.0194836824159766e-07,
+      "loss": 0.0006,
+      "step": 8182
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.0192401363857765e-07,
+      "loss": 0.0005,
+      "step": 8183
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.018996590355577e-07,
+      "loss": 0.0004,
+      "step": 8184
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 3.018753044325377e-07,
+      "loss": 0.0004,
+      "step": 8185
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.0185094982951776e-07,
+      "loss": 0.0005,
+      "step": 8186
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.018265952264978e-07,
+      "loss": 0.0007,
+      "step": 8187
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.018022406234778e-07,
+      "loss": 0.0006,
+      "step": 8188
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.0177788602045786e-07,
+      "loss": 0.0005,
+      "step": 8189
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.0175353141743786e-07,
+      "loss": 0.0006,
+      "step": 8190
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.017291768144179e-07,
+      "loss": 0.0005,
+      "step": 8191
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.017048222113979e-07,
+      "loss": 0.0005,
+      "step": 8192
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.0168046760837796e-07,
+      "loss": 0.1234,
+      "step": 8193
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.0165611300535796e-07,
+      "loss": 0.0008,
+      "step": 8194
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 3.01631758402338e-07,
+      "loss": 0.0004,
+      "step": 8195
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.0160740379931806e-07,
+      "loss": 0.0005,
+      "step": 8196
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.0158304919629806e-07,
+      "loss": 0.0005,
+      "step": 8197
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.015586945932781e-07,
+      "loss": 0.0007,
+      "step": 8198
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.015343399902581e-07,
+      "loss": 0.0006,
+      "step": 8199
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.0150998538723816e-07,
+      "loss": 0.0005,
+      "step": 8200
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.0148563078421816e-07,
+      "loss": 0.0006,
+      "step": 8201
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.014612761811982e-07,
+      "loss": 0.0005,
+      "step": 8202
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.0143692157817826e-07,
+      "loss": 0.0005,
+      "step": 8203
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.0141256697515826e-07,
+      "loss": 0.0005,
+      "step": 8204
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.013882123721383e-07,
+      "loss": 0.0005,
+      "step": 8205
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 3.013638577691183e-07,
+      "loss": 0.0004,
+      "step": 8206
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.0133950316609836e-07,
+      "loss": 0.0005,
+      "step": 8207
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.0131514856307836e-07,
+      "loss": 0.0005,
+      "step": 8208
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.0129079396005847e-07,
+      "loss": 0.0006,
+      "step": 8209
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.012664393570385e-07,
+      "loss": 0.0004,
+      "step": 8210
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.012420847540185e-07,
+      "loss": 0.0005,
+      "step": 8211
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.0121773015099857e-07,
+      "loss": 0.0007,
+      "step": 8212
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.0119337554797857e-07,
+      "loss": 0.0007,
+      "step": 8213
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.011690209449586e-07,
+      "loss": 0.0005,
+      "step": 8214
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.011446663419386e-07,
+      "loss": 0.0006,
+      "step": 8215
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 3.0112031173891867e-07,
+      "loss": 0.0006,
+      "step": 8216
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.0109595713589867e-07,
+      "loss": 0.0005,
+      "step": 8217
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.010716025328787e-07,
+      "loss": 0.0006,
+      "step": 8218
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.0104724792985877e-07,
+      "loss": 0.0004,
+      "step": 8219
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.0102289332683877e-07,
+      "loss": 0.0005,
+      "step": 8220
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.009985387238188e-07,
+      "loss": 0.0004,
+      "step": 8221
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.009741841207988e-07,
+      "loss": 0.0005,
+      "step": 8222
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.0094982951777887e-07,
+      "loss": 0.0005,
+      "step": 8223
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.0092547491475887e-07,
+      "loss": 0.0006,
+      "step": 8224
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.009011203117389e-07,
+      "loss": 0.0006,
+      "step": 8225
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 3.00876765708719e-07,
+      "loss": 0.0004,
+      "step": 8226
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.0085241110569897e-07,
+      "loss": 0.0005,
+      "step": 8227
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.00828056502679e-07,
+      "loss": 0.0004,
+      "step": 8228
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.00803701899659e-07,
+      "loss": 0.0005,
+      "step": 8229
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.007793472966391e-07,
+      "loss": 0.0006,
+      "step": 8230
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.007549926936191e-07,
+      "loss": 0.0005,
+      "step": 8231
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.007306380905991e-07,
+      "loss": 0.0004,
+      "step": 8232
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.9985415653864852,
+      "eval_loss": 0.01792530156672001,
+      "eval_runtime": 2.8066,
+      "eval_samples_per_second": 732.912,
+      "eval_steps_per_second": 11.758,
+      "step": 8232
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.007062834875791e-07,
+      "loss": 0.0004,
+      "step": 8233
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.006819288845592e-07,
+      "loss": 0.0006,
+      "step": 8234
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.0065757428153923e-07,
+      "loss": 0.0006,
+      "step": 8235
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.006332196785192e-07,
+      "loss": 0.0004,
+      "step": 8236
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 3.006088650754993e-07,
+      "loss": 0.0007,
+      "step": 8237
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.005845104724793e-07,
+      "loss": 0.0007,
+      "step": 8238
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.0056015586945933e-07,
+      "loss": 0.0012,
+      "step": 8239
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.0053580126643933e-07,
+      "loss": 0.0005,
+      "step": 8240
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.005114466634194e-07,
+      "loss": 0.0004,
+      "step": 8241
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.0048709206039943e-07,
+      "loss": 0.0005,
+      "step": 8242
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.0046273745737943e-07,
+      "loss": 0.0008,
+      "step": 8243
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.004383828543595e-07,
+      "loss": 0.0006,
+      "step": 8244
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.004140282513395e-07,
+      "loss": 0.0005,
+      "step": 8245
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.0038967364831953e-07,
+      "loss": 0.0004,
+      "step": 8246
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 3.0036531904529953e-07,
+      "loss": 0.0007,
+      "step": 8247
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.003409644422796e-07,
+      "loss": 0.0005,
+      "step": 8248
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.003166098392596e-07,
+      "loss": 0.0007,
+      "step": 8249
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.0029225523623963e-07,
+      "loss": 0.0004,
+      "step": 8250
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.002679006332197e-07,
+      "loss": 0.0005,
+      "step": 8251
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.002435460301997e-07,
+      "loss": 0.0005,
+      "step": 8252
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.0021919142717973e-07,
+      "loss": 0.0005,
+      "step": 8253
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.0019483682415973e-07,
+      "loss": 0.0006,
+      "step": 8254
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.001704822211398e-07,
+      "loss": 0.0005,
+      "step": 8255
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.001461276181198e-07,
+      "loss": 0.0004,
+      "step": 8256
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 3.0012177301509983e-07,
+      "loss": 0.0004,
+      "step": 8257
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 3.000974184120799e-07,
+      "loss": 0.0005,
+      "step": 8258
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 3.000730638090599e-07,
+      "loss": 0.0006,
+      "step": 8259
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 3.0004870920603994e-07,
+      "loss": 0.0006,
+      "step": 8260
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 3.0002435460301993e-07,
+      "loss": 0.0005,
+      "step": 8261
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 3e-07,
+      "loss": 0.001,
+      "step": 8262
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9997564539698e-07,
+      "loss": 0.0007,
+      "step": 8263
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9995129079396004e-07,
+      "loss": 0.0006,
+      "step": 8264
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9992693619094004e-07,
+      "loss": 0.0005,
+      "step": 8265
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.999025815879201e-07,
+      "loss": 0.0007,
+      "step": 8266
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9987822698490014e-07,
+      "loss": 0.0006,
+      "step": 8267
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 2.9985387238188014e-07,
+      "loss": 0.0006,
+      "step": 8268
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.998295177788602e-07,
+      "loss": 0.0005,
+      "step": 8269
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.998051631758402e-07,
+      "loss": 0.0006,
+      "step": 8270
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9978080857282024e-07,
+      "loss": 0.0005,
+      "step": 8271
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9975645396980024e-07,
+      "loss": 0.0004,
+      "step": 8272
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.997320993667803e-07,
+      "loss": 0.0007,
+      "step": 8273
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9970774476376034e-07,
+      "loss": 0.0005,
+      "step": 8274
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9968339016074034e-07,
+      "loss": 0.0005,
+      "step": 8275
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.996590355577204e-07,
+      "loss": 0.0006,
+      "step": 8276
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.996346809547004e-07,
+      "loss": 0.0006,
+      "step": 8277
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 2.9961032635168044e-07,
+      "loss": 0.0006,
+      "step": 8278
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.9958597174866044e-07,
+      "loss": 0.0005,
+      "step": 8279
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.995616171456405e-07,
+      "loss": 0.0007,
+      "step": 8280
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.995372625426205e-07,
+      "loss": 0.0005,
+      "step": 8281
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.9951290793960054e-07,
+      "loss": 0.0006,
+      "step": 8282
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.994885533365806e-07,
+      "loss": 0.0007,
+      "step": 8283
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.994641987335606e-07,
+      "loss": 0.0004,
+      "step": 8284
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.994398441305407e-07,
+      "loss": 0.0005,
+      "step": 8285
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.994154895275207e-07,
+      "loss": 0.0006,
+      "step": 8286
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.9939113492450075e-07,
+      "loss": 0.0005,
+      "step": 8287
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 2.9936678032148075e-07,
+      "loss": 0.0008,
+      "step": 8288
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.993424257184608e-07,
+      "loss": 0.0005,
+      "step": 8289
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9931807111544085e-07,
+      "loss": 0.0005,
+      "step": 8290
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9929371651242085e-07,
+      "loss": 0.0006,
+      "step": 8291
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.992693619094009e-07,
+      "loss": 0.0007,
+      "step": 8292
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.992450073063809e-07,
+      "loss": 0.0006,
+      "step": 8293
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9922065270336095e-07,
+      "loss": 0.0004,
+      "step": 8294
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9919629810034095e-07,
+      "loss": 0.0004,
+      "step": 8295
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.99171943497321e-07,
+      "loss": 0.0006,
+      "step": 8296
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9914758889430105e-07,
+      "loss": 0.0005,
+      "step": 8297
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 2.9912323429128105e-07,
+      "loss": 0.0006,
+      "step": 8298
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.990988796882611e-07,
+      "loss": 0.0006,
+      "step": 8299
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.990745250852411e-07,
+      "loss": 0.0004,
+      "step": 8300
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.9905017048222115e-07,
+      "loss": 0.0005,
+      "step": 8301
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.9902581587920115e-07,
+      "loss": 0.0007,
+      "step": 8302
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.990014612761812e-07,
+      "loss": 0.0005,
+      "step": 8303
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.989771066731612e-07,
+      "loss": 0.0004,
+      "step": 8304
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.9895275207014125e-07,
+      "loss": 0.0004,
+      "step": 8305
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.989283974671213e-07,
+      "loss": 0.0005,
+      "step": 8306
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.989040428641013e-07,
+      "loss": 0.0003,
+      "step": 8307
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.9887968826108136e-07,
+      "loss": 0.0006,
+      "step": 8308
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 2.9885533365806135e-07,
+      "loss": 0.0005,
+      "step": 8309
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.988309790550414e-07,
+      "loss": 0.1259,
+      "step": 8310
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.988066244520214e-07,
+      "loss": 0.0005,
+      "step": 8311
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.9878226984900146e-07,
+      "loss": 0.0007,
+      "step": 8312
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.987579152459815e-07,
+      "loss": 0.0005,
+      "step": 8313
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.987335606429615e-07,
+      "loss": 0.0005,
+      "step": 8314
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.9870920603994156e-07,
+      "loss": 0.0006,
+      "step": 8315
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.9868485143692156e-07,
+      "loss": 0.0007,
+      "step": 8316
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.986604968339016e-07,
+      "loss": 0.0008,
+      "step": 8317
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.986361422308816e-07,
+      "loss": 0.0006,
+      "step": 8318
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 2.9861178762786166e-07,
+      "loss": 0.0009,
+      "step": 8319
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9858743302484166e-07,
+      "loss": 0.0005,
+      "step": 8320
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.985630784218217e-07,
+      "loss": 0.0019,
+      "step": 8321
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9853872381880176e-07,
+      "loss": 0.0007,
+      "step": 8322
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9851436921578176e-07,
+      "loss": 0.0006,
+      "step": 8323
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.984900146127618e-07,
+      "loss": 0.0006,
+      "step": 8324
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.984656600097418e-07,
+      "loss": 0.0005,
+      "step": 8325
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9844130540672186e-07,
+      "loss": 0.0005,
+      "step": 8326
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9841695080370186e-07,
+      "loss": 0.0004,
+      "step": 8327
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.983925962006819e-07,
+      "loss": 0.0004,
+      "step": 8328
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 2.9836824159766196e-07,
+      "loss": 0.0011,
+      "step": 8329
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.9834388699464196e-07,
+      "loss": 0.0006,
+      "step": 8330
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.98319532391622e-07,
+      "loss": 0.0005,
+      "step": 8331
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.98295177788602e-07,
+      "loss": 0.0006,
+      "step": 8332
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.9827082318558206e-07,
+      "loss": 0.0006,
+      "step": 8333
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.9824646858256206e-07,
+      "loss": 0.0006,
+      "step": 8334
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.982221139795421e-07,
+      "loss": 0.0005,
+      "step": 8335
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.9819775937652217e-07,
+      "loss": 0.0004,
+      "step": 8336
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.9817340477350216e-07,
+      "loss": 0.0007,
+      "step": 8337
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.981490501704822e-07,
+      "loss": 0.0006,
+      "step": 8338
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.981246955674622e-07,
+      "loss": 0.2389,
+      "step": 8339
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 2.9810034096444227e-07,
+      "loss": 0.0007,
+      "step": 8340
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.9807598636142227e-07,
+      "loss": 0.0006,
+      "step": 8341
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.980516317584023e-07,
+      "loss": 0.0005,
+      "step": 8342
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.980272771553823e-07,
+      "loss": 0.0008,
+      "step": 8343
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.9800292255236237e-07,
+      "loss": 0.0005,
+      "step": 8344
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.979785679493424e-07,
+      "loss": 0.0004,
+      "step": 8345
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.979542133463224e-07,
+      "loss": 0.0006,
+      "step": 8346
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.9792985874330247e-07,
+      "loss": 0.0007,
+      "step": 8347
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.9790550414028247e-07,
+      "loss": 0.0005,
+      "step": 8348
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.978811495372625e-07,
+      "loss": 0.0005,
+      "step": 8349
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 2.978567949342425e-07,
+      "loss": 0.0005,
+      "step": 8350
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.9783244033122257e-07,
+      "loss": 0.0007,
+      "step": 8351
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.978080857282026e-07,
+      "loss": 0.0004,
+      "step": 8352
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.977837311251826e-07,
+      "loss": 0.0006,
+      "step": 8353
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.9775937652216267e-07,
+      "loss": 0.0005,
+      "step": 8354
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.9773502191914267e-07,
+      "loss": 0.0005,
+      "step": 8355
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.977106673161227e-07,
+      "loss": 0.0005,
+      "step": 8356
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.976863127131027e-07,
+      "loss": 0.0005,
+      "step": 8357
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.9766195811008277e-07,
+      "loss": 0.0008,
+      "step": 8358
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.9763760350706277e-07,
+      "loss": 0.0005,
+      "step": 8359
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 2.976132489040428e-07,
+      "loss": 0.0004,
+      "step": 8360
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.975888943010229e-07,
+      "loss": 0.0005,
+      "step": 8361
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.975645396980029e-07,
+      "loss": 0.0006,
+      "step": 8362
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.97540185094983e-07,
+      "loss": 0.0005,
+      "step": 8363
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.97515830491963e-07,
+      "loss": 0.0005,
+      "step": 8364
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.9749147588894303e-07,
+      "loss": 0.0006,
+      "step": 8365
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.9746712128592303e-07,
+      "loss": 0.0023,
+      "step": 8366
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.974427666829031e-07,
+      "loss": 0.0006,
+      "step": 8367
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.9741841207988313e-07,
+      "loss": 0.0005,
+      "step": 8368
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.9739405747686313e-07,
+      "loss": 0.0006,
+      "step": 8369
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 2.973697028738432e-07,
+      "loss": 0.0006,
+      "step": 8370
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.973453482708232e-07,
+      "loss": 0.0005,
+      "step": 8371
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.9732099366780323e-07,
+      "loss": 0.0006,
+      "step": 8372
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.9729663906478323e-07,
+      "loss": 0.0008,
+      "step": 8373
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.972722844617633e-07,
+      "loss": 0.0004,
+      "step": 8374
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.9724792985874333e-07,
+      "loss": 0.0005,
+      "step": 8375
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.9722357525572333e-07,
+      "loss": 0.0011,
+      "step": 8376
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.971992206527034e-07,
+      "loss": 0.0007,
+      "step": 8377
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.971748660496834e-07,
+      "loss": 0.001,
+      "step": 8378
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.9715051144666343e-07,
+      "loss": 0.0005,
+      "step": 8379
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.9712615684364343e-07,
+      "loss": 1.9112,
+      "step": 8380
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 2.971018022406235e-07,
+      "loss": 0.0005,
+      "step": 8381
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.970774476376035e-07,
+      "loss": 0.0005,
+      "step": 8382
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.9705309303458353e-07,
+      "loss": 0.0006,
+      "step": 8383
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.970287384315636e-07,
+      "loss": 0.0003,
+      "step": 8384
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.970043838285436e-07,
+      "loss": 0.0006,
+      "step": 8385
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.9698002922552364e-07,
+      "loss": 0.0008,
+      "step": 8386
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.9695567462250364e-07,
+      "loss": 0.0004,
+      "step": 8387
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.969313200194837e-07,
+      "loss": 0.0005,
+      "step": 8388
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.969069654164637e-07,
+      "loss": 0.0004,
+      "step": 8389
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.9688261081344374e-07,
+      "loss": 0.0006,
+      "step": 8390
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 2.968582562104238e-07,
+      "loss": 0.0007,
+      "step": 8391
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.968339016074038e-07,
+      "loss": 0.0006,
+      "step": 8392
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9680954700438384e-07,
+      "loss": 0.0006,
+      "step": 8393
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9678519240136384e-07,
+      "loss": 0.0006,
+      "step": 8394
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.967608377983439e-07,
+      "loss": 0.0056,
+      "step": 8395
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.967364831953239e-07,
+      "loss": 0.0005,
+      "step": 8396
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9671212859230394e-07,
+      "loss": 0.0006,
+      "step": 8397
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9668777398928394e-07,
+      "loss": 0.0005,
+      "step": 8398
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.96663419386264e-07,
+      "loss": 0.0005,
+      "step": 8399
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9663906478324404e-07,
+      "loss": 0.0005,
+      "step": 8400
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 2.9661471018022404e-07,
+      "loss": 0.0005,
+      "step": 8401
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.965903555772041e-07,
+      "loss": 0.0006,
+      "step": 8402
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.965660009741841e-07,
+      "loss": 0.0005,
+      "step": 8403
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9654164637116414e-07,
+      "loss": 0.0006,
+      "step": 8404
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9651729176814414e-07,
+      "loss": 0.0005,
+      "step": 8405
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.964929371651242e-07,
+      "loss": 0.0004,
+      "step": 8406
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9646858256210424e-07,
+      "loss": 0.0004,
+      "step": 8407
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9644422795908424e-07,
+      "loss": 0.0006,
+      "step": 8408
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.964198733560643e-07,
+      "loss": 0.0007,
+      "step": 8409
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.963955187530443e-07,
+      "loss": 0.0005,
+      "step": 8410
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9637116415002434e-07,
+      "loss": 0.0009,
+      "step": 8411
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 2.9634680954700434e-07,
+      "loss": 0.0005,
+      "step": 8412
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.963224549439844e-07,
+      "loss": 0.0005,
+      "step": 8413
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.962981003409644e-07,
+      "loss": 0.0004,
+      "step": 8414
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.9627374573794445e-07,
+      "loss": 0.4308,
+      "step": 8415
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.962493911349245e-07,
+      "loss": 0.0004,
+      "step": 8416
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.962250365319045e-07,
+      "loss": 0.0004,
+      "step": 8417
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.9620068192888455e-07,
+      "loss": 0.0007,
+      "step": 8418
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.9617632732586455e-07,
+      "loss": 0.0004,
+      "step": 8419
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.961519727228446e-07,
+      "loss": 0.0004,
+      "step": 8420
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.961276181198246e-07,
+      "loss": 0.0008,
+      "step": 8421
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 2.9610326351680465e-07,
+      "loss": 0.0004,
+      "step": 8422
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.960789089137847e-07,
+      "loss": 0.0006,
+      "step": 8423
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.960545543107647e-07,
+      "loss": 0.0006,
+      "step": 8424
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9603019970774475e-07,
+      "loss": 0.0005,
+      "step": 8425
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9600584510472475e-07,
+      "loss": 0.0006,
+      "step": 8426
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.959814905017048e-07,
+      "loss": 0.0005,
+      "step": 8427
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.959571358986848e-07,
+      "loss": 0.0005,
+      "step": 8428
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9593278129566485e-07,
+      "loss": 0.0005,
+      "step": 8429
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9590842669264485e-07,
+      "loss": 0.0005,
+      "step": 8430
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.958840720896249e-07,
+      "loss": 0.0007,
+      "step": 8431
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 2.9585971748660495e-07,
+      "loss": 0.0007,
+      "step": 8432
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.9583536288358495e-07,
+      "loss": 0.0005,
+      "step": 8433
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.95811008280565e-07,
+      "loss": 0.0006,
+      "step": 8434
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.95786653677545e-07,
+      "loss": 0.0005,
+      "step": 8435
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.9576229907452505e-07,
+      "loss": 0.0005,
+      "step": 8436
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.9573794447150505e-07,
+      "loss": 0.0005,
+      "step": 8437
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.957135898684851e-07,
+      "loss": 0.0004,
+      "step": 8438
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.956892352654652e-07,
+      "loss": 0.0006,
+      "step": 8439
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.956648806624452e-07,
+      "loss": 0.0005,
+      "step": 8440
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.9564052605942526e-07,
+      "loss": 0.0006,
+      "step": 8441
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 2.9561617145640526e-07,
+      "loss": 0.0004,
+      "step": 8442
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.955918168533853e-07,
+      "loss": 0.0006,
+      "step": 8443
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.955674622503653e-07,
+      "loss": 0.0006,
+      "step": 8444
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9554310764734536e-07,
+      "loss": 0.0005,
+      "step": 8445
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.955187530443254e-07,
+      "loss": 0.0005,
+      "step": 8446
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.954943984413054e-07,
+      "loss": 0.0005,
+      "step": 8447
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9547004383828546e-07,
+      "loss": 0.0005,
+      "step": 8448
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9544568923526546e-07,
+      "loss": 0.0007,
+      "step": 8449
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.954213346322455e-07,
+      "loss": 0.8207,
+      "step": 8450
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.953969800292255e-07,
+      "loss": 0.0005,
+      "step": 8451
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9537262542620556e-07,
+      "loss": 0.0006,
+      "step": 8452
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 2.9534827082318556e-07,
+      "loss": 0.0139,
+      "step": 8453
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.953239162201656e-07,
+      "loss": 0.0005,
+      "step": 8454
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9529956161714566e-07,
+      "loss": 0.0005,
+      "step": 8455
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9527520701412566e-07,
+      "loss": 0.0012,
+      "step": 8456
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.952508524111057e-07,
+      "loss": 0.0004,
+      "step": 8457
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.952264978080857e-07,
+      "loss": 0.0004,
+      "step": 8458
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9520214320506576e-07,
+      "loss": 0.0004,
+      "step": 8459
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9517778860204576e-07,
+      "loss": 0.0004,
+      "step": 8460
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.951534339990258e-07,
+      "loss": 0.0006,
+      "step": 8461
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9512907939600587e-07,
+      "loss": 0.0005,
+      "step": 8462
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 2.9510472479298587e-07,
+      "loss": 0.0006,
+      "step": 8463
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.950803701899659e-07,
+      "loss": 0.001,
+      "step": 8464
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.950560155869459e-07,
+      "loss": 0.0005,
+      "step": 8465
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.9503166098392597e-07,
+      "loss": 0.0005,
+      "step": 8466
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.9500730638090597e-07,
+      "loss": 0.0005,
+      "step": 8467
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.94982951777886e-07,
+      "loss": 0.0005,
+      "step": 8468
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.94958597174866e-07,
+      "loss": 0.0004,
+      "step": 8469
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.9493424257184607e-07,
+      "loss": 0.0008,
+      "step": 8470
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.949098879688261e-07,
+      "loss": 0.0005,
+      "step": 8471
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.948855333658061e-07,
+      "loss": 0.0007,
+      "step": 8472
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 2.9486117876278617e-07,
+      "loss": 0.0005,
+      "step": 8473
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9483682415976617e-07,
+      "loss": 0.0006,
+      "step": 8474
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.948124695567462e-07,
+      "loss": 0.0004,
+      "step": 8475
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.947881149537262e-07,
+      "loss": 0.0004,
+      "step": 8476
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9476376035070627e-07,
+      "loss": 0.0005,
+      "step": 8477
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.947394057476863e-07,
+      "loss": 0.0007,
+      "step": 8478
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.947150511446663e-07,
+      "loss": 0.0004,
+      "step": 8479
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9469069654164637e-07,
+      "loss": 0.0006,
+      "step": 8480
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9466634193862637e-07,
+      "loss": 0.0006,
+      "step": 8481
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.946419873356064e-07,
+      "loss": 0.0007,
+      "step": 8482
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.946176327325864e-07,
+      "loss": 0.0005,
+      "step": 8483
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 2.9459327812956647e-07,
+      "loss": 0.0005,
+      "step": 8484
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.9456892352654647e-07,
+      "loss": 0.0833,
+      "step": 8485
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.945445689235265e-07,
+      "loss": 0.0005,
+      "step": 8486
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.945202143205066e-07,
+      "loss": 0.0004,
+      "step": 8487
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.944958597174866e-07,
+      "loss": 0.0005,
+      "step": 8488
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.944715051144666e-07,
+      "loss": 0.0006,
+      "step": 8489
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.944471505114466e-07,
+      "loss": 0.0005,
+      "step": 8490
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.944227959084267e-07,
+      "loss": 0.0004,
+      "step": 8491
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.943984413054067e-07,
+      "loss": 0.0067,
+      "step": 8492
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.943740867023867e-07,
+      "loss": 0.0004,
+      "step": 8493
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 2.943497320993668e-07,
+      "loss": 0.0006,
+      "step": 8494
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.943253774963468e-07,
+      "loss": 0.0004,
+      "step": 8495
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9430102289332683e-07,
+      "loss": 0.0005,
+      "step": 8496
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9427666829030683e-07,
+      "loss": 0.0004,
+      "step": 8497
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.942523136872869e-07,
+      "loss": 0.0006,
+      "step": 8498
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.942279590842669e-07,
+      "loss": 0.0005,
+      "step": 8499
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9420360448124693e-07,
+      "loss": 0.0007,
+      "step": 8500
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9417924987822693e-07,
+      "loss": 0.0007,
+      "step": 8501
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.94154895275207e-07,
+      "loss": 0.0003,
+      "step": 8502
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9413054067218703e-07,
+      "loss": 0.0005,
+      "step": 8503
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 2.9410618606916703e-07,
+      "loss": 0.0004,
+      "step": 8504
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.940818314661471e-07,
+      "loss": 0.0005,
+      "step": 8505
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.940574768631271e-07,
+      "loss": 0.001,
+      "step": 8506
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.9403312226010713e-07,
+      "loss": 0.0006,
+      "step": 8507
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.9400876765708713e-07,
+      "loss": 0.0004,
+      "step": 8508
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.939844130540672e-07,
+      "loss": 0.0007,
+      "step": 8509
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.9396005845104723e-07,
+      "loss": 0.0007,
+      "step": 8510
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.9393570384802723e-07,
+      "loss": 0.0005,
+      "step": 8511
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.939113492450073e-07,
+      "loss": 0.0006,
+      "step": 8512
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.938869946419873e-07,
+      "loss": 0.0005,
+      "step": 8513
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 2.9386264003896733e-07,
+      "loss": 0.0005,
+      "step": 8514
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.938382854359474e-07,
+      "loss": 0.0004,
+      "step": 8515
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.9381393083292744e-07,
+      "loss": 0.0005,
+      "step": 8516
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.937895762299075e-07,
+      "loss": 0.0004,
+      "step": 8517
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.937652216268875e-07,
+      "loss": 0.0006,
+      "step": 8518
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.9374086702386754e-07,
+      "loss": 0.0006,
+      "step": 8519
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.9371651242084754e-07,
+      "loss": 0.0006,
+      "step": 8520
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.936921578178276e-07,
+      "loss": 0.0004,
+      "step": 8521
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.936678032148076e-07,
+      "loss": 0.0005,
+      "step": 8522
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.9364344861178764e-07,
+      "loss": 0.0005,
+      "step": 8523
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.9361909400876764e-07,
+      "loss": 0.0004,
+      "step": 8524
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 2.935947394057477e-07,
+      "loss": 0.0005,
+      "step": 8525
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9357038480272774e-07,
+      "loss": 0.0004,
+      "step": 8526
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9354603019970774e-07,
+      "loss": 0.0004,
+      "step": 8527
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.935216755966878e-07,
+      "loss": 0.0004,
+      "step": 8528
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.934973209936678e-07,
+      "loss": 0.0005,
+      "step": 8529
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9347296639064784e-07,
+      "loss": 0.0006,
+      "step": 8530
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9344861178762784e-07,
+      "loss": 0.0004,
+      "step": 8531
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.934242571846079e-07,
+      "loss": 0.0005,
+      "step": 8532
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9339990258158794e-07,
+      "loss": 0.0007,
+      "step": 8533
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.9337554797856794e-07,
+      "loss": 1.6993,
+      "step": 8534
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 2.93351193375548e-07,
+      "loss": 0.0012,
+      "step": 8535
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.93326838772528e-07,
+      "loss": 0.0005,
+      "step": 8536
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.9330248416950805e-07,
+      "loss": 0.0006,
+      "step": 8537
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.9327812956648804e-07,
+      "loss": 0.0009,
+      "step": 8538
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.932537749634681e-07,
+      "loss": 0.0004,
+      "step": 8539
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.9322942036044815e-07,
+      "loss": 0.0004,
+      "step": 8540
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.9320506575742815e-07,
+      "loss": 0.0005,
+      "step": 8541
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.931807111544082e-07,
+      "loss": 0.0005,
+      "step": 8542
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.931563565513882e-07,
+      "loss": 0.0009,
+      "step": 8543
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.9313200194836825e-07,
+      "loss": 0.0007,
+      "step": 8544
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 2.9310764734534825e-07,
+      "loss": 0.0005,
+      "step": 8545
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.930832927423283e-07,
+      "loss": 0.0004,
+      "step": 8546
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.930589381393083e-07,
+      "loss": 1.4946,
+      "step": 8547
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.9303458353628835e-07,
+      "loss": 0.0007,
+      "step": 8548
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.930102289332684e-07,
+      "loss": 0.0006,
+      "step": 8549
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.929858743302484e-07,
+      "loss": 0.0007,
+      "step": 8550
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.9296151972722845e-07,
+      "loss": 0.0005,
+      "step": 8551
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.9293716512420845e-07,
+      "loss": 0.0005,
+      "step": 8552
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.929128105211885e-07,
+      "loss": 0.0004,
+      "step": 8553
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.928884559181685e-07,
+      "loss": 0.0005,
+      "step": 8554
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.9286410131514855e-07,
+      "loss": 0.1217,
+      "step": 8555
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 2.928397467121286e-07,
+      "loss": 0.0005,
+      "step": 8556
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.928153921091086e-07,
+      "loss": 0.0005,
+      "step": 8557
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9279103750608865e-07,
+      "loss": 0.0004,
+      "step": 8558
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9276668290306865e-07,
+      "loss": 0.0005,
+      "step": 8559
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.927423283000487e-07,
+      "loss": 0.0004,
+      "step": 8560
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.927179736970287e-07,
+      "loss": 0.0005,
+      "step": 8561
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9269361909400875e-07,
+      "loss": 0.0005,
+      "step": 8562
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9266926449098875e-07,
+      "loss": 0.0003,
+      "step": 8563
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.926449098879688e-07,
+      "loss": 0.0004,
+      "step": 8564
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9262055528494886e-07,
+      "loss": 0.0004,
+      "step": 8565
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 2.9259620068192885e-07,
+      "loss": 0.0005,
+      "step": 8566
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.925718460789089e-07,
+      "loss": 0.0004,
+      "step": 8567
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.925474914758889e-07,
+      "loss": 0.0006,
+      "step": 8568
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.9252313687286896e-07,
+      "loss": 0.0004,
+      "step": 8569
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.9249878226984895e-07,
+      "loss": 0.0005,
+      "step": 8570
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.92474427666829e-07,
+      "loss": 0.0008,
+      "step": 8571
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.9245007306380906e-07,
+      "loss": 0.0005,
+      "step": 8572
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.9242571846078906e-07,
+      "loss": 0.0005,
+      "step": 8573
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.924013638577691e-07,
+      "loss": 0.0005,
+      "step": 8574
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.923770092547491e-07,
+      "loss": 0.0005,
+      "step": 8575
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 2.9235265465172916e-07,
+      "loss": 0.0005,
+      "step": 8576
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9232830004870916e-07,
+      "loss": 0.0005,
+      "step": 8577
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.923039454456892e-07,
+      "loss": 0.0004,
+      "step": 8578
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.922795908426692e-07,
+      "loss": 0.0008,
+      "step": 8579
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9225523623964926e-07,
+      "loss": 0.0008,
+      "step": 8580
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.922308816366293e-07,
+      "loss": 0.0005,
+      "step": 8581
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.922065270336093e-07,
+      "loss": 0.0006,
+      "step": 8582
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9218217243058936e-07,
+      "loss": 0.0006,
+      "step": 8583
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9215781782756936e-07,
+      "loss": 0.0005,
+      "step": 8584
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.921334632245494e-07,
+      "loss": 0.0005,
+      "step": 8585
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.921091086215294e-07,
+      "loss": 0.0003,
+      "step": 8586
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 2.9208475401850946e-07,
+      "loss": 0.0006,
+      "step": 8587
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.920603994154895e-07,
+      "loss": 0.0005,
+      "step": 8588
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.920360448124695e-07,
+      "loss": 0.0009,
+      "step": 8589
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.9201169020944956e-07,
+      "loss": 0.0006,
+      "step": 8590
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.9198733560642956e-07,
+      "loss": 0.8672,
+      "step": 8591
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.9196298100340967e-07,
+      "loss": 0.0006,
+      "step": 8592
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.9193862640038967e-07,
+      "loss": 0.0003,
+      "step": 8593
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.919142717973697e-07,
+      "loss": 0.0007,
+      "step": 8594
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.9188991719434977e-07,
+      "loss": 0.0004,
+      "step": 8595
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.9186556259132977e-07,
+      "loss": 0.0006,
+      "step": 8596
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 2.918412079883098e-07,
+      "loss": 0.0006,
+      "step": 8597
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.918168533852898e-07,
+      "loss": 0.0008,
+      "step": 8598
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9179249878226987e-07,
+      "loss": 0.0005,
+      "step": 8599
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9176814417924987e-07,
+      "loss": 0.0006,
+      "step": 8600
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.917437895762299e-07,
+      "loss": 0.0007,
+      "step": 8601
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.917194349732099e-07,
+      "loss": 0.0005,
+      "step": 8602
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9169508037018997e-07,
+      "loss": 0.0006,
+      "step": 8603
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9167072576717e-07,
+      "loss": 0.0006,
+      "step": 8604
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9164637116415e-07,
+      "loss": 0.0005,
+      "step": 8605
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9162201656113007e-07,
+      "loss": 0.0004,
+      "step": 8606
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 2.9159766195811007e-07,
+      "loss": 0.0006,
+      "step": 8607
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.915733073550901e-07,
+      "loss": 0.0004,
+      "step": 8608
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.915489527520701e-07,
+      "loss": 0.0004,
+      "step": 8609
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.9152459814905017e-07,
+      "loss": 0.0005,
+      "step": 8610
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.915002435460302e-07,
+      "loss": 0.0005,
+      "step": 8611
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.914758889430102e-07,
+      "loss": 0.0003,
+      "step": 8612
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.914515343399903e-07,
+      "loss": 0.0007,
+      "step": 8613
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.914271797369703e-07,
+      "loss": 0.0004,
+      "step": 8614
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.914028251339503e-07,
+      "loss": 0.0004,
+      "step": 8615
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.913784705309303e-07,
+      "loss": 0.0004,
+      "step": 8616
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 2.913541159279104e-07,
+      "loss": 0.0004,
+      "step": 8617
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.913297613248904e-07,
+      "loss": 0.0007,
+      "step": 8618
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.913054067218704e-07,
+      "loss": 0.0005,
+      "step": 8619
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.912810521188505e-07,
+      "loss": 0.0003,
+      "step": 8620
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.912566975158305e-07,
+      "loss": 0.0005,
+      "step": 8621
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.9123234291281053e-07,
+      "loss": 0.0005,
+      "step": 8622
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.9120798830979053e-07,
+      "loss": 0.0011,
+      "step": 8623
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.911836337067706e-07,
+      "loss": 0.0004,
+      "step": 8624
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.911592791037506e-07,
+      "loss": 0.0007,
+      "step": 8625
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.9113492450073063e-07,
+      "loss": 0.0004,
+      "step": 8626
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.911105698977107e-07,
+      "loss": 0.0006,
+      "step": 8627
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 2.910862152946907e-07,
+      "loss": 0.0007,
+      "step": 8628
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.9106186069167073e-07,
+      "loss": 0.0006,
+      "step": 8629
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.9103750608865073e-07,
+      "loss": 0.0006,
+      "step": 8630
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.910131514856308e-07,
+      "loss": 0.0004,
+      "step": 8631
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.909887968826108e-07,
+      "loss": 0.0005,
+      "step": 8632
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.9096444227959083e-07,
+      "loss": 0.0008,
+      "step": 8633
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.9094008767657083e-07,
+      "loss": 0.0005,
+      "step": 8634
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.909157330735509e-07,
+      "loss": 0.0005,
+      "step": 8635
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.9089137847053093e-07,
+      "loss": 0.0005,
+      "step": 8636
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.9086702386751093e-07,
+      "loss": 0.0004,
+      "step": 8637
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 2.90842669264491e-07,
+      "loss": 0.0004,
+      "step": 8638
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.90818314661471e-07,
+      "loss": 0.0005,
+      "step": 8639
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.9079396005845103e-07,
+      "loss": 0.0005,
+      "step": 8640
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.9076960545543103e-07,
+      "loss": 0.0005,
+      "step": 8641
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.907452508524111e-07,
+      "loss": 0.0004,
+      "step": 8642
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.9072089624939114e-07,
+      "loss": 0.0005,
+      "step": 8643
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.9069654164637113e-07,
+      "loss": 0.0004,
+      "step": 8644
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.906721870433512e-07,
+      "loss": 0.0005,
+      "step": 8645
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.906478324403312e-07,
+      "loss": 0.0005,
+      "step": 8646
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.9062347783731124e-07,
+      "loss": 0.0007,
+      "step": 8647
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 2.9059912323429124e-07,
+      "loss": 0.0004,
+      "step": 8648
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.905747686312713e-07,
+      "loss": 0.0005,
+      "step": 8649
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.905504140282513e-07,
+      "loss": 0.0005,
+      "step": 8650
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.9052605942523134e-07,
+      "loss": 0.0005,
+      "step": 8651
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.905017048222114e-07,
+      "loss": 0.0005,
+      "step": 8652
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.904773502191914e-07,
+      "loss": 0.0006,
+      "step": 8653
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.9045299561617144e-07,
+      "loss": 0.0095,
+      "step": 8654
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.9042864101315144e-07,
+      "loss": 0.0006,
+      "step": 8655
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.904042864101315e-07,
+      "loss": 0.0005,
+      "step": 8656
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.903799318071115e-07,
+      "loss": 0.0005,
+      "step": 8657
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.9035557720409154e-07,
+      "loss": 0.0006,
+      "step": 8658
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 2.903312226010716e-07,
+      "loss": 0.0004,
+      "step": 8659
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.903068679980516e-07,
+      "loss": 0.0005,
+      "step": 8660
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.9028251339503164e-07,
+      "loss": 0.0004,
+      "step": 8661
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.9025815879201164e-07,
+      "loss": 0.0006,
+      "step": 8662
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.902338041889917e-07,
+      "loss": 0.0007,
+      "step": 8663
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.902094495859717e-07,
+      "loss": 0.0005,
+      "step": 8664
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.9018509498295174e-07,
+      "loss": 0.0004,
+      "step": 8665
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.9016074037993174e-07,
+      "loss": 0.0009,
+      "step": 8666
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.901363857769118e-07,
+      "loss": 0.0005,
+      "step": 8667
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.9011203117389184e-07,
+      "loss": 0.0005,
+      "step": 8668
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 2.900876765708719e-07,
+      "loss": 0.0007,
+      "step": 8669
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.9006332196785195e-07,
+      "loss": 0.0004,
+      "step": 8670
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.9003896736483195e-07,
+      "loss": 0.0004,
+      "step": 8671
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.90014612761812e-07,
+      "loss": 0.0005,
+      "step": 8672
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.89990258158792e-07,
+      "loss": 0.0004,
+      "step": 8673
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.8996590355577205e-07,
+      "loss": 0.0004,
+      "step": 8674
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.899415489527521e-07,
+      "loss": 0.0013,
+      "step": 8675
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.899171943497321e-07,
+      "loss": 0.0005,
+      "step": 8676
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.8989283974671215e-07,
+      "loss": 0.0006,
+      "step": 8677
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.8986848514369215e-07,
+      "loss": 0.0005,
+      "step": 8678
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 2.898441305406722e-07,
+      "loss": 0.0006,
+      "step": 8679
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.898197759376522e-07,
+      "loss": 0.0006,
+      "step": 8680
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8979542133463225e-07,
+      "loss": 0.0005,
+      "step": 8681
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.897710667316123e-07,
+      "loss": 0.0005,
+      "step": 8682
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.897467121285923e-07,
+      "loss": 0.0007,
+      "step": 8683
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8972235752557235e-07,
+      "loss": 0.0005,
+      "step": 8684
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8969800292255235e-07,
+      "loss": 0.0005,
+      "step": 8685
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.896736483195324e-07,
+      "loss": 0.0007,
+      "step": 8686
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.896492937165124e-07,
+      "loss": 0.0003,
+      "step": 8687
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8962493911349245e-07,
+      "loss": 0.0006,
+      "step": 8688
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 2.8960058451047245e-07,
+      "loss": 0.0004,
+      "step": 8689
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.895762299074525e-07,
+      "loss": 0.0004,
+      "step": 8690
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8955187530443256e-07,
+      "loss": 0.0008,
+      "step": 8691
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8952752070141255e-07,
+      "loss": 0.0006,
+      "step": 8692
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.895031660983926e-07,
+      "loss": 0.0007,
+      "step": 8693
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.894788114953726e-07,
+      "loss": 0.0004,
+      "step": 8694
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8945445689235266e-07,
+      "loss": 0.0006,
+      "step": 8695
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8943010228933266e-07,
+      "loss": 0.0005,
+      "step": 8696
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.894057476863127e-07,
+      "loss": 0.0005,
+      "step": 8697
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8938139308329276e-07,
+      "loss": 0.0005,
+      "step": 8698
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.8935703848027276e-07,
+      "loss": 0.0007,
+      "step": 8699
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 2.893326838772528e-07,
+      "loss": 0.0004,
+      "step": 8700
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.893083292742328e-07,
+      "loss": 0.0004,
+      "step": 8701
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8928397467121286e-07,
+      "loss": 0.0006,
+      "step": 8702
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8925962006819286e-07,
+      "loss": 0.0006,
+      "step": 8703
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.892352654651729e-07,
+      "loss": 0.0007,
+      "step": 8704
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8921091086215296e-07,
+      "loss": 0.0006,
+      "step": 8705
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8918655625913296e-07,
+      "loss": 0.0006,
+      "step": 8706
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.89162201656113e-07,
+      "loss": 0.0006,
+      "step": 8707
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.89137847053093e-07,
+      "loss": 0.0006,
+      "step": 8708
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8911349245007306e-07,
+      "loss": 0.0007,
+      "step": 8709
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 2.8908913784705306e-07,
+      "loss": 0.0004,
+      "step": 8710
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.890647832440331e-07,
+      "loss": 0.0005,
+      "step": 8711
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.890404286410131e-07,
+      "loss": 0.0005,
+      "step": 8712
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.8901607403799316e-07,
+      "loss": 0.014,
+      "step": 8713
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.889917194349732e-07,
+      "loss": 0.0005,
+      "step": 8714
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.889673648319532e-07,
+      "loss": 0.0005,
+      "step": 8715
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.8894301022893326e-07,
+      "loss": 0.0004,
+      "step": 8716
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.8891865562591326e-07,
+      "loss": 0.0006,
+      "step": 8717
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.888943010228933e-07,
+      "loss": 0.0004,
+      "step": 8718
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.888699464198733e-07,
+      "loss": 0.0006,
+      "step": 8719
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 2.8884559181685336e-07,
+      "loss": 0.0004,
+      "step": 8720
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.888212372138334e-07,
+      "loss": 0.0006,
+      "step": 8721
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.887968826108134e-07,
+      "loss": 0.0007,
+      "step": 8722
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.8877252800779347e-07,
+      "loss": 0.0004,
+      "step": 8723
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.8874817340477347e-07,
+      "loss": 0.0005,
+      "step": 8724
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.887238188017535e-07,
+      "loss": 0.0004,
+      "step": 8725
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.886994641987335e-07,
+      "loss": 0.0005,
+      "step": 8726
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.8867510959571357e-07,
+      "loss": 0.0007,
+      "step": 8727
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.8865075499269357e-07,
+      "loss": 0.0004,
+      "step": 8728
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.886264003896736e-07,
+      "loss": 0.0006,
+      "step": 8729
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.8860204578665367e-07,
+      "loss": 0.0004,
+      "step": 8730
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 2.8857769118363367e-07,
+      "loss": 0.0005,
+      "step": 8731
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.885533365806137e-07,
+      "loss": 0.0005,
+      "step": 8732
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.885289819775937e-07,
+      "loss": 0.0003,
+      "step": 8733
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8850462737457377e-07,
+      "loss": 0.0004,
+      "step": 8734
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8848027277155377e-07,
+      "loss": 0.0005,
+      "step": 8735
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.884559181685338e-07,
+      "loss": 0.0008,
+      "step": 8736
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8843156356551387e-07,
+      "loss": 0.0004,
+      "step": 8737
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8840720896249387e-07,
+      "loss": 0.0006,
+      "step": 8738
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.883828543594739e-07,
+      "loss": 0.0004,
+      "step": 8739
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.883584997564539e-07,
+      "loss": 1.5067,
+      "step": 8740
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 2.8833414515343397e-07,
+      "loss": 0.0006,
+      "step": 8741
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.8830979055041397e-07,
+      "loss": 1.2635,
+      "step": 8742
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.88285435947394e-07,
+      "loss": 0.0007,
+      "step": 8743
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.88261081344374e-07,
+      "loss": 0.0005,
+      "step": 8744
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.8823672674135413e-07,
+      "loss": 0.0009,
+      "step": 8745
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.882123721383342e-07,
+      "loss": 0.0004,
+      "step": 8746
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.881880175353142e-07,
+      "loss": 0.0006,
+      "step": 8747
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.8816366293229423e-07,
+      "loss": 0.0004,
+      "step": 8748
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.8813930832927423e-07,
+      "loss": 0.0005,
+      "step": 8749
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.881149537262543e-07,
+      "loss": 0.0008,
+      "step": 8750
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 2.880905991232343e-07,
+      "loss": 0.0003,
+      "step": 8751
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.8806624452021433e-07,
+      "loss": 0.0004,
+      "step": 8752
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.880418899171944e-07,
+      "loss": 0.0004,
+      "step": 8753
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.880175353141744e-07,
+      "loss": 0.0004,
+      "step": 8754
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.8799318071115443e-07,
+      "loss": 0.0005,
+      "step": 8755
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.8796882610813443e-07,
+      "loss": 0.0005,
+      "step": 8756
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.879444715051145e-07,
+      "loss": 0.0005,
+      "step": 8757
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.879201169020945e-07,
+      "loss": 0.0003,
+      "step": 8758
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.8789576229907453e-07,
+      "loss": 0.0006,
+      "step": 8759
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.878714076960546e-07,
+      "loss": 0.0018,
+      "step": 8760
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 2.878470530930346e-07,
+      "loss": 0.0005,
+      "step": 8761
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8782269849001463e-07,
+      "loss": 0.0006,
+      "step": 8762
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8779834388699463e-07,
+      "loss": 0.0008,
+      "step": 8763
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.877739892839747e-07,
+      "loss": 0.0101,
+      "step": 8764
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.877496346809547e-07,
+      "loss": 0.0005,
+      "step": 8765
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8772528007793473e-07,
+      "loss": 0.001,
+      "step": 8766
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8770092547491473e-07,
+      "loss": 0.0007,
+      "step": 8767
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.876765708718948e-07,
+      "loss": 0.0005,
+      "step": 8768
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8765221626887484e-07,
+      "loss": 0.0005,
+      "step": 8769
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.8762786166585484e-07,
+      "loss": 0.0005,
+      "step": 8770
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.876035070628349e-07,
+      "loss": 0.0004,
+      "step": 8771
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 2.875791524598149e-07,
+      "loss": 0.0006,
+      "step": 8772
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.8755479785679494e-07,
+      "loss": 0.0005,
+      "step": 8773
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.8753044325377494e-07,
+      "loss": 0.0006,
+      "step": 8774
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.87506088650755e-07,
+      "loss": 0.0004,
+      "step": 8775
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.8748173404773504e-07,
+      "loss": 0.0005,
+      "step": 8776
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.8745737944471504e-07,
+      "loss": 0.0005,
+      "step": 8777
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.874330248416951e-07,
+      "loss": 0.0005,
+      "step": 8778
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.874086702386751e-07,
+      "loss": 0.0004,
+      "step": 8779
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.8738431563565514e-07,
+      "loss": 0.0005,
+      "step": 8780
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.8735996103263514e-07,
+      "loss": 0.0006,
+      "step": 8781
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 2.873356064296152e-07,
+      "loss": 0.0005,
+      "step": 8782
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.873112518265952e-07,
+      "loss": 0.0005,
+      "step": 8783
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.8728689722357524e-07,
+      "loss": 0.0007,
+      "step": 8784
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.872625426205553e-07,
+      "loss": 0.0005,
+      "step": 8785
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.872381880175353e-07,
+      "loss": 0.0004,
+      "step": 8786
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.8721383341451534e-07,
+      "loss": 0.0007,
+      "step": 8787
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.8718947881149534e-07,
+      "loss": 0.0004,
+      "step": 8788
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.871651242084754e-07,
+      "loss": 0.0004,
+      "step": 8789
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.871407696054554e-07,
+      "loss": 0.0006,
+      "step": 8790
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.8711641500243544e-07,
+      "loss": 0.0005,
+      "step": 8791
+    },
+    {
+      "epoch": 8.54,
+      "learning_rate": 2.870920603994155e-07,
+      "loss": 0.0005,
+      "step": 8792
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.870677057963955e-07,
+      "loss": 0.0006,
+      "step": 8793
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8704335119337554e-07,
+      "loss": 0.0004,
+      "step": 8794
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8701899659035554e-07,
+      "loss": 0.0004,
+      "step": 8795
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.869946419873356e-07,
+      "loss": 0.0004,
+      "step": 8796
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.869702873843156e-07,
+      "loss": 0.0005,
+      "step": 8797
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8694593278129565e-07,
+      "loss": 0.0004,
+      "step": 8798
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8692157817827564e-07,
+      "loss": 0.0006,
+      "step": 8799
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.868972235752557e-07,
+      "loss": 0.0007,
+      "step": 8800
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8687286897223575e-07,
+      "loss": 0.0004,
+      "step": 8801
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.8684851436921575e-07,
+      "loss": 0.0007,
+      "step": 8802
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 2.868241597661958e-07,
+      "loss": 0.0005,
+      "step": 8803
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.867998051631758e-07,
+      "loss": 0.0006,
+      "step": 8804
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8677545056015585e-07,
+      "loss": 0.0004,
+      "step": 8805
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8675109595713585e-07,
+      "loss": 0.0005,
+      "step": 8806
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.867267413541159e-07,
+      "loss": 0.1749,
+      "step": 8807
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8670238675109595e-07,
+      "loss": 0.0004,
+      "step": 8808
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8667803214807595e-07,
+      "loss": 0.0005,
+      "step": 8809
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.86653677545056e-07,
+      "loss": 0.0003,
+      "step": 8810
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.86629322942036e-07,
+      "loss": 0.0005,
+      "step": 8811
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8660496833901605e-07,
+      "loss": 0.0004,
+      "step": 8812
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 2.8658061373599605e-07,
+      "loss": 0.0005,
+      "step": 8813
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.865562591329761e-07,
+      "loss": 0.0012,
+      "step": 8814
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.865319045299561e-07,
+      "loss": 0.0005,
+      "step": 8815
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.8650754992693615e-07,
+      "loss": 0.0005,
+      "step": 8816
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.864831953239162e-07,
+      "loss": 0.0004,
+      "step": 8817
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.864588407208962e-07,
+      "loss": 0.0004,
+      "step": 8818
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.8643448611787625e-07,
+      "loss": 0.0006,
+      "step": 8819
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.8641013151485625e-07,
+      "loss": 0.0006,
+      "step": 8820
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.863857769118363e-07,
+      "loss": 0.0006,
+      "step": 8821
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.8636142230881636e-07,
+      "loss": 0.0004,
+      "step": 8822
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 2.863370677057964e-07,
+      "loss": 0.0007,
+      "step": 8823
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8631271310277646e-07,
+      "loss": 0.0005,
+      "step": 8824
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8628835849975646e-07,
+      "loss": 0.0006,
+      "step": 8825
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.862640038967365e-07,
+      "loss": 0.0004,
+      "step": 8826
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.862396492937165e-07,
+      "loss": 0.0006,
+      "step": 8827
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8621529469069656e-07,
+      "loss": 0.0005,
+      "step": 8828
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8619094008767656e-07,
+      "loss": 0.0004,
+      "step": 8829
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.861665854846566e-07,
+      "loss": 0.0007,
+      "step": 8830
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8614223088163666e-07,
+      "loss": 0.0006,
+      "step": 8831
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.8611787627861666e-07,
+      "loss": 0.0021,
+      "step": 8832
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 2.860935216755967e-07,
+      "loss": 0.0004,
+      "step": 8833
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.860691670725767e-07,
+      "loss": 0.0005,
+      "step": 8834
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8604481246955676e-07,
+      "loss": 0.0004,
+      "step": 8835
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8602045786653676e-07,
+      "loss": 0.0011,
+      "step": 8836
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.859961032635168e-07,
+      "loss": 0.0005,
+      "step": 8837
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.859717486604968e-07,
+      "loss": 0.0004,
+      "step": 8838
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8594739405747686e-07,
+      "loss": 0.0004,
+      "step": 8839
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.859230394544569e-07,
+      "loss": 0.0003,
+      "step": 8840
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.858986848514369e-07,
+      "loss": 0.0005,
+      "step": 8841
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8587433024841696e-07,
+      "loss": 0.0005,
+      "step": 8842
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.8584997564539696e-07,
+      "loss": 0.0004,
+      "step": 8843
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 2.85825621042377e-07,
+      "loss": 0.0005,
+      "step": 8844
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.85801266439357e-07,
+      "loss": 0.0006,
+      "step": 8845
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.8577691183633707e-07,
+      "loss": 0.0005,
+      "step": 8846
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.857525572333171e-07,
+      "loss": 0.0011,
+      "step": 8847
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.857282026302971e-07,
+      "loss": 0.0007,
+      "step": 8848
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.8570384802727717e-07,
+      "loss": 0.0005,
+      "step": 8849
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.8567949342425717e-07,
+      "loss": 0.0004,
+      "step": 8850
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.856551388212372e-07,
+      "loss": 0.0004,
+      "step": 8851
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.856307842182172e-07,
+      "loss": 0.0004,
+      "step": 8852
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.8560642961519727e-07,
+      "loss": 0.0005,
+      "step": 8853
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 2.8558207501217727e-07,
+      "loss": 0.0004,
+      "step": 8854
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.855577204091573e-07,
+      "loss": 0.0008,
+      "step": 8855
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8553336580613737e-07,
+      "loss": 0.0005,
+      "step": 8856
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8550901120311737e-07,
+      "loss": 0.0003,
+      "step": 8857
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.854846566000974e-07,
+      "loss": 0.0004,
+      "step": 8858
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.854603019970774e-07,
+      "loss": 0.0004,
+      "step": 8859
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8543594739405747e-07,
+      "loss": 0.0005,
+      "step": 8860
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8541159279103747e-07,
+      "loss": 0.0005,
+      "step": 8861
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.853872381880175e-07,
+      "loss": 0.0004,
+      "step": 8862
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8536288358499757e-07,
+      "loss": 0.0037,
+      "step": 8863
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 2.8533852898197757e-07,
+      "loss": 0.0005,
+      "step": 8864
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.853141743789576e-07,
+      "loss": 0.0003,
+      "step": 8865
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.852898197759376e-07,
+      "loss": 0.0089,
+      "step": 8866
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.8526546517291767e-07,
+      "loss": 0.0007,
+      "step": 8867
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.8524111056989767e-07,
+      "loss": 0.0004,
+      "step": 8868
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.852167559668777e-07,
+      "loss": 0.0007,
+      "step": 8869
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.851924013638578e-07,
+      "loss": 0.0005,
+      "step": 8870
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.851680467608378e-07,
+      "loss": 0.0006,
+      "step": 8871
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.851436921578178e-07,
+      "loss": 0.0004,
+      "step": 8872
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.851193375547978e-07,
+      "loss": 0.0006,
+      "step": 8873
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.850949829517779e-07,
+      "loss": 0.0006,
+      "step": 8874
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 2.850706283487579e-07,
+      "loss": 0.0005,
+      "step": 8875
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.850462737457379e-07,
+      "loss": 0.8158,
+      "step": 8876
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.850219191427179e-07,
+      "loss": 0.0004,
+      "step": 8877
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.84997564539698e-07,
+      "loss": 0.001,
+      "step": 8878
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.8497320993667803e-07,
+      "loss": 0.0006,
+      "step": 8879
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.8494885533365803e-07,
+      "loss": 0.0003,
+      "step": 8880
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.849245007306381e-07,
+      "loss": 0.001,
+      "step": 8881
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.849001461276181e-07,
+      "loss": 0.0003,
+      "step": 8882
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.8487579152459813e-07,
+      "loss": 0.0007,
+      "step": 8883
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.8485143692157813e-07,
+      "loss": 0.0006,
+      "step": 8884
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 2.848270823185582e-07,
+      "loss": 0.0006,
+      "step": 8885
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.8480272771553823e-07,
+      "loss": 0.0004,
+      "step": 8886
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.8477837311251823e-07,
+      "loss": 0.0007,
+      "step": 8887
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.847540185094983e-07,
+      "loss": 0.0005,
+      "step": 8888
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.847296639064783e-07,
+      "loss": 0.0004,
+      "step": 8889
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.8470530930345833e-07,
+      "loss": 0.0005,
+      "step": 8890
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.8468095470043833e-07,
+      "loss": 0.0005,
+      "step": 8891
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.846566000974184e-07,
+      "loss": 0.0004,
+      "step": 8892
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.846322454943984e-07,
+      "loss": 0.0005,
+      "step": 8893
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.8460789089137843e-07,
+      "loss": 0.0005,
+      "step": 8894
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 2.845835362883585e-07,
+      "loss": 0.0004,
+      "step": 8895
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.845591816853385e-07,
+      "loss": 0.0005,
+      "step": 8896
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.8453482708231853e-07,
+      "loss": 0.0004,
+      "step": 8897
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.845104724792986e-07,
+      "loss": 0.0006,
+      "step": 8898
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.8448611787627864e-07,
+      "loss": 0.0006,
+      "step": 8899
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.8446176327325864e-07,
+      "loss": 0.0004,
+      "step": 8900
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.844374086702387e-07,
+      "loss": 0.0005,
+      "step": 8901
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.8441305406721874e-07,
+      "loss": 0.0004,
+      "step": 8902
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.8438869946419874e-07,
+      "loss": 0.0005,
+      "step": 8903
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.843643448611788e-07,
+      "loss": 0.0005,
+      "step": 8904
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 2.843399902581588e-07,
+      "loss": 0.0004,
+      "step": 8905
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8431563565513884e-07,
+      "loss": 0.0004,
+      "step": 8906
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8429128105211884e-07,
+      "loss": 0.0005,
+      "step": 8907
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.842669264490989e-07,
+      "loss": 0.0005,
+      "step": 8908
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8424257184607894e-07,
+      "loss": 0.0004,
+      "step": 8909
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8421821724305894e-07,
+      "loss": 0.0005,
+      "step": 8910
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.84193862640039e-07,
+      "loss": 0.0004,
+      "step": 8911
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.84169508037019e-07,
+      "loss": 0.0005,
+      "step": 8912
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8414515343399904e-07,
+      "loss": 0.0006,
+      "step": 8913
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.8412079883097904e-07,
+      "loss": 0.0003,
+      "step": 8914
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.840964442279591e-07,
+      "loss": 0.0006,
+      "step": 8915
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 2.840720896249391e-07,
+      "loss": 0.0004,
+      "step": 8916
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.8404773502191914e-07,
+      "loss": 0.0005,
+      "step": 8917
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.840233804188992e-07,
+      "loss": 0.0129,
+      "step": 8918
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.839990258158792e-07,
+      "loss": 0.0005,
+      "step": 8919
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.8397467121285925e-07,
+      "loss": 0.0036,
+      "step": 8920
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.8395031660983924e-07,
+      "loss": 0.0005,
+      "step": 8921
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.839259620068193e-07,
+      "loss": 0.0005,
+      "step": 8922
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.839016074037993e-07,
+      "loss": 0.0003,
+      "step": 8923
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.8387725280077935e-07,
+      "loss": 0.0004,
+      "step": 8924
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.838528981977594e-07,
+      "loss": 0.0003,
+      "step": 8925
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 2.838285435947394e-07,
+      "loss": 0.0005,
+      "step": 8926
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.8380418899171945e-07,
+      "loss": 0.0006,
+      "step": 8927
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.8377983438869945e-07,
+      "loss": 0.0005,
+      "step": 8928
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.837554797856795e-07,
+      "loss": 0.0004,
+      "step": 8929
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.837311251826595e-07,
+      "loss": 0.0003,
+      "step": 8930
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.8370677057963955e-07,
+      "loss": 0.0013,
+      "step": 8931
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.8368241597661955e-07,
+      "loss": 0.0003,
+      "step": 8932
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.836580613735996e-07,
+      "loss": 0.0005,
+      "step": 8933
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.8363370677057965e-07,
+      "loss": 0.0005,
+      "step": 8934
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.8360935216755965e-07,
+      "loss": 0.0004,
+      "step": 8935
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 2.835849975645397e-07,
+      "loss": 0.0004,
+      "step": 8936
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.835606429615197e-07,
+      "loss": 0.0005,
+      "step": 8937
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8353628835849975e-07,
+      "loss": 0.0003,
+      "step": 8938
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8351193375547975e-07,
+      "loss": 0.0004,
+      "step": 8939
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.834875791524598e-07,
+      "loss": 0.0005,
+      "step": 8940
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8346322454943985e-07,
+      "loss": 0.0005,
+      "step": 8941
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8343886994641985e-07,
+      "loss": 0.0005,
+      "step": 8942
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.834145153433999e-07,
+      "loss": 0.0004,
+      "step": 8943
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.833901607403799e-07,
+      "loss": 0.0004,
+      "step": 8944
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8336580613735995e-07,
+      "loss": 0.0004,
+      "step": 8945
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8334145153433995e-07,
+      "loss": 0.0004,
+      "step": 8946
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 2.8331709693132e-07,
+      "loss": 0.0005,
+      "step": 8947
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.832927423283e-07,
+      "loss": 0.0005,
+      "step": 8948
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.8326838772528005e-07,
+      "loss": 0.0006,
+      "step": 8949
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.832440331222601e-07,
+      "loss": 0.0013,
+      "step": 8950
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.832196785192401e-07,
+      "loss": 0.0005,
+      "step": 8951
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.8319532391622016e-07,
+      "loss": 0.0004,
+      "step": 8952
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.8317096931320015e-07,
+      "loss": 0.0005,
+      "step": 8953
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.831466147101802e-07,
+      "loss": 0.0006,
+      "step": 8954
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.831222601071602e-07,
+      "loss": 0.0005,
+      "step": 8955
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.8309790550414026e-07,
+      "loss": 0.0003,
+      "step": 8956
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 2.830735509011203e-07,
+      "loss": 0.0005,
+      "step": 8957
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.830491962981003e-07,
+      "loss": 0.0006,
+      "step": 8958
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8302484169508036e-07,
+      "loss": 0.0006,
+      "step": 8959
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8300048709206036e-07,
+      "loss": 0.0004,
+      "step": 8960
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.829761324890404e-07,
+      "loss": 0.0005,
+      "step": 8961
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.829517778860204e-07,
+      "loss": 0.0006,
+      "step": 8962
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8292742328300046e-07,
+      "loss": 0.0026,
+      "step": 8963
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8290306867998046e-07,
+      "loss": 0.0005,
+      "step": 8964
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.828787140769605e-07,
+      "loss": 0.0004,
+      "step": 8965
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8285435947394056e-07,
+      "loss": 0.0007,
+      "step": 8966
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 2.8283000487092056e-07,
+      "loss": 0.0004,
+      "step": 8967
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.828056502679006e-07,
+      "loss": 0.0013,
+      "step": 8968
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.827812956648806e-07,
+      "loss": 0.0004,
+      "step": 8969
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8275694106186066e-07,
+      "loss": 0.0004,
+      "step": 8970
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8273258645884066e-07,
+      "loss": 0.0004,
+      "step": 8971
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.827082318558207e-07,
+      "loss": 0.0028,
+      "step": 8972
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8268387725280076e-07,
+      "loss": 0.0005,
+      "step": 8973
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8265952264978076e-07,
+      "loss": 0.0006,
+      "step": 8974
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8263516804676087e-07,
+      "loss": 0.0005,
+      "step": 8975
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.8261081344374087e-07,
+      "loss": 0.0005,
+      "step": 8976
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.825864588407209e-07,
+      "loss": 0.0004,
+      "step": 8977
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 2.825621042377009e-07,
+      "loss": 0.0005,
+      "step": 8978
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.8253774963468097e-07,
+      "loss": 0.0006,
+      "step": 8979
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.82513395031661e-07,
+      "loss": 0.0004,
+      "step": 8980
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.82489040428641e-07,
+      "loss": 0.0005,
+      "step": 8981
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.8246468582562107e-07,
+      "loss": 0.0006,
+      "step": 8982
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.8244033122260107e-07,
+      "loss": 0.0005,
+      "step": 8983
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.824159766195811e-07,
+      "loss": 0.0005,
+      "step": 8984
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.823916220165611e-07,
+      "loss": 0.0004,
+      "step": 8985
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.8236726741354117e-07,
+      "loss": 0.0006,
+      "step": 8986
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.8234291281052117e-07,
+      "loss": 0.0005,
+      "step": 8987
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 2.823185582075012e-07,
+      "loss": 0.0005,
+      "step": 8988
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.8229420360448127e-07,
+      "loss": 0.0004,
+      "step": 8989
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.8226984900146127e-07,
+      "loss": 0.0004,
+      "step": 8990
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.822454943984413e-07,
+      "loss": 0.0004,
+      "step": 8991
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.822211397954213e-07,
+      "loss": 0.0007,
+      "step": 8992
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.8219678519240137e-07,
+      "loss": 0.0004,
+      "step": 8993
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.8217243058938137e-07,
+      "loss": 0.0007,
+      "step": 8994
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.821480759863614e-07,
+      "loss": 0.0004,
+      "step": 8995
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.821237213833415e-07,
+      "loss": 0.0007,
+      "step": 8996
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.820993667803215e-07,
+      "loss": 0.0006,
+      "step": 8997
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 2.820750121773015e-07,
+      "loss": 0.0006,
+      "step": 8998
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.820506575742815e-07,
+      "loss": 0.0006,
+      "step": 8999
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.820263029712616e-07,
+      "loss": 0.0005,
+      "step": 9000
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.820019483682416e-07,
+      "loss": 0.0005,
+      "step": 9001
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.819775937652216e-07,
+      "loss": 0.0004,
+      "step": 9002
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.819532391622016e-07,
+      "loss": 2.3363,
+      "step": 9003
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.819288845591817e-07,
+      "loss": 0.001,
+      "step": 9004
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.8190452995616173e-07,
+      "loss": 0.0005,
+      "step": 9005
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.8188017535314173e-07,
+      "loss": 0.0006,
+      "step": 9006
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.818558207501218e-07,
+      "loss": 0.0005,
+      "step": 9007
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 2.818314661471018e-07,
+      "loss": 0.0005,
+      "step": 9008
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8180711154408183e-07,
+      "loss": 0.0004,
+      "step": 9009
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8178275694106183e-07,
+      "loss": 0.0004,
+      "step": 9010
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.817584023380419e-07,
+      "loss": 0.0004,
+      "step": 9011
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8173404773502193e-07,
+      "loss": 0.0005,
+      "step": 9012
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8170969313200193e-07,
+      "loss": 0.0006,
+      "step": 9013
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.81685338528982e-07,
+      "loss": 0.0005,
+      "step": 9014
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.81660983925962e-07,
+      "loss": 0.0005,
+      "step": 9015
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8163662932294203e-07,
+      "loss": 0.0004,
+      "step": 9016
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.8161227471992203e-07,
+      "loss": 0.0004,
+      "step": 9017
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.815879201169021e-07,
+      "loss": 0.0004,
+      "step": 9018
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 2.815635655138821e-07,
+      "loss": 0.0005,
+      "step": 9019
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.8153921091086213e-07,
+      "loss": 0.0005,
+      "step": 9020
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.815148563078422e-07,
+      "loss": 0.0004,
+      "step": 9021
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.814905017048222e-07,
+      "loss": 0.0003,
+      "step": 9022
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.8146614710180223e-07,
+      "loss": 0.0006,
+      "step": 9023
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.8144179249878223e-07,
+      "loss": 0.0004,
+      "step": 9024
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.814174378957623e-07,
+      "loss": 0.0006,
+      "step": 9025
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.813930832927423e-07,
+      "loss": 0.0004,
+      "step": 9026
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.8136872868972233e-07,
+      "loss": 0.0005,
+      "step": 9027
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.813443740867024e-07,
+      "loss": 0.0003,
+      "step": 9028
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 2.813200194836824e-07,
+      "loss": 0.0004,
+      "step": 9029
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.8129566488066244e-07,
+      "loss": 0.0005,
+      "step": 9030
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.8127131027764244e-07,
+      "loss": 0.0003,
+      "step": 9031
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.812469556746225e-07,
+      "loss": 0.0005,
+      "step": 9032
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.812226010716025e-07,
+      "loss": 0.0005,
+      "step": 9033
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.8119824646858254e-07,
+      "loss": 0.0005,
+      "step": 9034
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.8117389186556254e-07,
+      "loss": 0.0004,
+      "step": 9035
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.811495372625426e-07,
+      "loss": 0.0005,
+      "step": 9036
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.8112518265952264e-07,
+      "loss": 0.0004,
+      "step": 9037
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.8110082805650264e-07,
+      "loss": 0.0003,
+      "step": 9038
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 2.810764734534827e-07,
+      "loss": 0.0004,
+      "step": 9039
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.810521188504627e-07,
+      "loss": 0.0005,
+      "step": 9040
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8102776424744274e-07,
+      "loss": 0.0135,
+      "step": 9041
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8100340964442274e-07,
+      "loss": 0.0008,
+      "step": 9042
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.809790550414028e-07,
+      "loss": 0.0006,
+      "step": 9043
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8095470043838284e-07,
+      "loss": 0.0005,
+      "step": 9044
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8093034583536284e-07,
+      "loss": 0.0003,
+      "step": 9045
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.809059912323429e-07,
+      "loss": 0.0004,
+      "step": 9046
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.808816366293229e-07,
+      "loss": 0.0003,
+      "step": 9047
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8085728202630294e-07,
+      "loss": 0.0004,
+      "step": 9048
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.8083292742328294e-07,
+      "loss": 0.0007,
+      "step": 9049
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 2.80808572820263e-07,
+      "loss": 0.0004,
+      "step": 9050
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.8078421821724304e-07,
+      "loss": 0.0005,
+      "step": 9051
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.807598636142231e-07,
+      "loss": 0.0004,
+      "step": 9052
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.8073550901120315e-07,
+      "loss": 0.0004,
+      "step": 9053
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.8071115440818315e-07,
+      "loss": 0.0004,
+      "step": 9054
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.806867998051632e-07,
+      "loss": 0.0005,
+      "step": 9055
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.806624452021432e-07,
+      "loss": 0.0005,
+      "step": 9056
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.8063809059912325e-07,
+      "loss": 0.0004,
+      "step": 9057
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.8061373599610325e-07,
+      "loss": 0.0005,
+      "step": 9058
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.805893813930833e-07,
+      "loss": 0.0004,
+      "step": 9059
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 2.8056502679006335e-07,
+      "loss": 0.0004,
+      "step": 9060
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.8054067218704335e-07,
+      "loss": 0.0003,
+      "step": 9061
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.805163175840234e-07,
+      "loss": 0.0005,
+      "step": 9062
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.804919629810034e-07,
+      "loss": 0.0006,
+      "step": 9063
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.8046760837798345e-07,
+      "loss": 0.0005,
+      "step": 9064
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.8044325377496345e-07,
+      "loss": 0.0006,
+      "step": 9065
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.804188991719435e-07,
+      "loss": 0.0004,
+      "step": 9066
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.8039454456892355e-07,
+      "loss": 0.0103,
+      "step": 9067
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.8037018996590355e-07,
+      "loss": 0.0003,
+      "step": 9068
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.803458353628836e-07,
+      "loss": 0.0004,
+      "step": 9069
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 2.803214807598636e-07,
+      "loss": 0.0004,
+      "step": 9070
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.8029712615684365e-07,
+      "loss": 0.0003,
+      "step": 9071
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.8027277155382365e-07,
+      "loss": 0.0005,
+      "step": 9072
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.802484169508037e-07,
+      "loss": 0.0005,
+      "step": 9073
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.8022406234778376e-07,
+      "loss": 0.0006,
+      "step": 9074
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.8019970774476375e-07,
+      "loss": 0.0006,
+      "step": 9075
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.801753531417438e-07,
+      "loss": 0.0004,
+      "step": 9076
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.801509985387238e-07,
+      "loss": 0.0005,
+      "step": 9077
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.8012664393570386e-07,
+      "loss": 0.0006,
+      "step": 9078
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.8010228933268386e-07,
+      "loss": 0.0005,
+      "step": 9079
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 2.800779347296639e-07,
+      "loss": 0.0006,
+      "step": 9080
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.800535801266439e-07,
+      "loss": 0.0005,
+      "step": 9081
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.8002922552362396e-07,
+      "loss": 0.0004,
+      "step": 9082
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.80004870920604e-07,
+      "loss": 0.142,
+      "step": 9083
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.79980516317584e-07,
+      "loss": 0.0005,
+      "step": 9084
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.7995616171456406e-07,
+      "loss": 0.0006,
+      "step": 9085
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.7993180711154406e-07,
+      "loss": 0.0005,
+      "step": 9086
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.799074525085241e-07,
+      "loss": 0.0004,
+      "step": 9087
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.798830979055041e-07,
+      "loss": 0.0005,
+      "step": 9088
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.7985874330248416e-07,
+      "loss": 0.0004,
+      "step": 9089
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.798343886994642e-07,
+      "loss": 0.0005,
+      "step": 9090
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 2.798100340964442e-07,
+      "loss": 0.0004,
+      "step": 9091
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.7978567949342426e-07,
+      "loss": 0.0004,
+      "step": 9092
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.7976132489040426e-07,
+      "loss": 0.0004,
+      "step": 9093
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.797369702873843e-07,
+      "loss": 0.0005,
+      "step": 9094
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.797126156843643e-07,
+      "loss": 0.0004,
+      "step": 9095
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.7968826108134436e-07,
+      "loss": 0.0003,
+      "step": 9096
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.7966390647832436e-07,
+      "loss": 0.0006,
+      "step": 9097
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.796395518753044e-07,
+      "loss": 0.0005,
+      "step": 9098
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.7961519727228446e-07,
+      "loss": 0.0004,
+      "step": 9099
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.7959084266926446e-07,
+      "loss": 0.0004,
+      "step": 9100
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 2.795664880662445e-07,
+      "loss": 0.0005,
+      "step": 9101
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.795421334632245e-07,
+      "loss": 0.0007,
+      "step": 9102
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.7951777886020457e-07,
+      "loss": 1.4724,
+      "step": 9103
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.7949342425718456e-07,
+      "loss": 0.0005,
+      "step": 9104
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.794690696541646e-07,
+      "loss": 0.0005,
+      "step": 9105
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.7944471505114467e-07,
+      "loss": 0.0004,
+      "step": 9106
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.7942036044812467e-07,
+      "loss": 0.0004,
+      "step": 9107
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.793960058451047e-07,
+      "loss": 0.0006,
+      "step": 9108
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.793716512420847e-07,
+      "loss": 0.0005,
+      "step": 9109
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.7934729663906477e-07,
+      "loss": 0.0007,
+      "step": 9110
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 2.7932294203604477e-07,
+      "loss": 0.0004,
+      "step": 9111
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.792985874330248e-07,
+      "loss": 0.0005,
+      "step": 9112
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.792742328300048e-07,
+      "loss": 0.0005,
+      "step": 9113
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.7924987822698487e-07,
+      "loss": 0.0003,
+      "step": 9114
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.792255236239649e-07,
+      "loss": 0.0006,
+      "step": 9115
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.792011690209449e-07,
+      "loss": 0.0005,
+      "step": 9116
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.7917681441792497e-07,
+      "loss": 0.0004,
+      "step": 9117
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.7915245981490497e-07,
+      "loss": 0.0006,
+      "step": 9118
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.79128105211885e-07,
+      "loss": 0.0004,
+      "step": 9119
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.79103750608865e-07,
+      "loss": 0.0003,
+      "step": 9120
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.7907939600584507e-07,
+      "loss": 0.0006,
+      "step": 9121
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 2.790550414028251e-07,
+      "loss": 0.0004,
+      "step": 9122
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.790306867998051e-07,
+      "loss": 0.0004,
+      "step": 9123
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.7900633219678517e-07,
+      "loss": 0.0004,
+      "step": 9124
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.7898197759376517e-07,
+      "loss": 0.0004,
+      "step": 9125
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.789576229907452e-07,
+      "loss": 0.0004,
+      "step": 9126
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.789332683877252e-07,
+      "loss": 0.0004,
+      "step": 9127
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.7890891378470533e-07,
+      "loss": 1.5085,
+      "step": 9128
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.788845591816854e-07,
+      "loss": 0.0005,
+      "step": 9129
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.788602045786654e-07,
+      "loss": 0.0005,
+      "step": 9130
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.7883584997564543e-07,
+      "loss": 0.0006,
+      "step": 9131
+    },
+    {
+      "epoch": 8.87,
+      "learning_rate": 2.7881149537262543e-07,
+      "loss": 0.0346,
+      "step": 9132
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.787871407696055e-07,
+      "loss": 0.0005,
+      "step": 9133
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.787627861665855e-07,
+      "loss": 0.0007,
+      "step": 9134
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.7873843156356553e-07,
+      "loss": 0.0005,
+      "step": 9135
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.7871407696054553e-07,
+      "loss": 0.0004,
+      "step": 9136
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.786897223575256e-07,
+      "loss": 0.0004,
+      "step": 9137
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.7866536775450563e-07,
+      "loss": 0.0011,
+      "step": 9138
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.7864101315148563e-07,
+      "loss": 0.0007,
+      "step": 9139
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.786166585484657e-07,
+      "loss": 0.0004,
+      "step": 9140
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.785923039454457e-07,
+      "loss": 0.0004,
+      "step": 9141
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 2.7856794934242573e-07,
+      "loss": 0.0005,
+      "step": 9142
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.7854359473940573e-07,
+      "loss": 0.0003,
+      "step": 9143
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.785192401363858e-07,
+      "loss": 0.0005,
+      "step": 9144
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.7849488553336583e-07,
+      "loss": 0.0006,
+      "step": 9145
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.7847053093034583e-07,
+      "loss": 0.0004,
+      "step": 9146
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.784461763273259e-07,
+      "loss": 0.0006,
+      "step": 9147
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.784218217243059e-07,
+      "loss": 0.0003,
+      "step": 9148
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.7839746712128593e-07,
+      "loss": 0.0003,
+      "step": 9149
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.7837311251826593e-07,
+      "loss": 0.0005,
+      "step": 9150
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.78348757915246e-07,
+      "loss": 0.0005,
+      "step": 9151
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 2.78324403312226e-07,
+      "loss": 0.0003,
+      "step": 9152
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.7830004870920604e-07,
+      "loss": 0.0006,
+      "step": 9153
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.782756941061861e-07,
+      "loss": 0.0005,
+      "step": 9154
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.782513395031661e-07,
+      "loss": 0.0006,
+      "step": 9155
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.7822698490014614e-07,
+      "loss": 0.0006,
+      "step": 9156
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.7820263029712614e-07,
+      "loss": 0.0004,
+      "step": 9157
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.781782756941062e-07,
+      "loss": 0.0005,
+      "step": 9158
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.781539210910862e-07,
+      "loss": 0.0006,
+      "step": 9159
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.7812956648806624e-07,
+      "loss": 0.0006,
+      "step": 9160
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.781052118850463e-07,
+      "loss": 0.0004,
+      "step": 9161
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.780808572820263e-07,
+      "loss": 0.0005,
+      "step": 9162
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 2.7805650267900634e-07,
+      "loss": 0.0003,
+      "step": 9163
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.7803214807598634e-07,
+      "loss": 0.0004,
+      "step": 9164
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.780077934729664e-07,
+      "loss": 0.0004,
+      "step": 9165
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.779834388699464e-07,
+      "loss": 0.0004,
+      "step": 9166
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.7795908426692644e-07,
+      "loss": 0.0004,
+      "step": 9167
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.7793472966390644e-07,
+      "loss": 0.0004,
+      "step": 9168
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.779103750608865e-07,
+      "loss": 0.0004,
+      "step": 9169
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.7788602045786654e-07,
+      "loss": 0.0004,
+      "step": 9170
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.7786166585484654e-07,
+      "loss": 0.0007,
+      "step": 9171
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.778373112518266e-07,
+      "loss": 0.0005,
+      "step": 9172
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 2.778129566488066e-07,
+      "loss": 0.0004,
+      "step": 9173
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.7778860204578664e-07,
+      "loss": 0.001,
+      "step": 9174
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.7776424744276664e-07,
+      "loss": 0.0005,
+      "step": 9175
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.777398928397467e-07,
+      "loss": 0.0005,
+      "step": 9176
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.7771553823672674e-07,
+      "loss": 0.0004,
+      "step": 9177
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.7769118363370674e-07,
+      "loss": 0.0003,
+      "step": 9178
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.776668290306868e-07,
+      "loss": 0.0004,
+      "step": 9179
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.776424744276668e-07,
+      "loss": 0.0006,
+      "step": 9180
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.7761811982464685e-07,
+      "loss": 0.0006,
+      "step": 9181
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.7759376522162684e-07,
+      "loss": 0.0005,
+      "step": 9182
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 2.775694106186069e-07,
+      "loss": 0.0027,
+      "step": 9183
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.775450560155869e-07,
+      "loss": 0.0007,
+      "step": 9184
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.7752070141256695e-07,
+      "loss": 0.0004,
+      "step": 9185
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.77496346809547e-07,
+      "loss": 0.0004,
+      "step": 9186
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.77471992206527e-07,
+      "loss": 0.0004,
+      "step": 9187
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.7744763760350705e-07,
+      "loss": 0.0005,
+      "step": 9188
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.7742328300048705e-07,
+      "loss": 0.0005,
+      "step": 9189
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.773989283974671e-07,
+      "loss": 0.0005,
+      "step": 9190
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.773745737944471e-07,
+      "loss": 0.0005,
+      "step": 9191
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.7735021919142715e-07,
+      "loss": 0.0005,
+      "step": 9192
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.773258645884072e-07,
+      "loss": 0.0005,
+      "step": 9193
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 2.773015099853872e-07,
+      "loss": 0.0003,
+      "step": 9194
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.7727715538236725e-07,
+      "loss": 0.0004,
+      "step": 9195
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.7725280077934725e-07,
+      "loss": 0.0005,
+      "step": 9196
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.772284461763273e-07,
+      "loss": 0.0005,
+      "step": 9197
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.772040915733073e-07,
+      "loss": 0.0005,
+      "step": 9198
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.7717973697028735e-07,
+      "loss": 0.0003,
+      "step": 9199
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.7715538236726735e-07,
+      "loss": 0.0004,
+      "step": 9200
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.771310277642474e-07,
+      "loss": 0.0005,
+      "step": 9201
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.7710667316122745e-07,
+      "loss": 0.0004,
+      "step": 9202
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.7708231855820745e-07,
+      "loss": 0.0006,
+      "step": 9203
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 2.770579639551875e-07,
+      "loss": 0.0005,
+      "step": 9204
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.7703360935216756e-07,
+      "loss": 0.0005,
+      "step": 9205
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.770092547491476e-07,
+      "loss": 0.0005,
+      "step": 9206
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.769849001461276e-07,
+      "loss": 0.0007,
+      "step": 9207
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.7696054554310766e-07,
+      "loss": 0.0005,
+      "step": 9208
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.769361909400877e-07,
+      "loss": 0.0006,
+      "step": 9209
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.769118363370677e-07,
+      "loss": 0.0004,
+      "step": 9210
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.7688748173404776e-07,
+      "loss": 0.0003,
+      "step": 9211
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.7686312713102776e-07,
+      "loss": 0.0005,
+      "step": 9212
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.768387725280078e-07,
+      "loss": 0.0005,
+      "step": 9213
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 2.768144179249878e-07,
+      "loss": 0.0006,
+      "step": 9214
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.7679006332196786e-07,
+      "loss": 0.0003,
+      "step": 9215
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.767657087189479e-07,
+      "loss": 0.0004,
+      "step": 9216
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.767413541159279e-07,
+      "loss": 0.0004,
+      "step": 9217
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.7671699951290796e-07,
+      "loss": 0.001,
+      "step": 9218
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.7669264490988796e-07,
+      "loss": 0.0003,
+      "step": 9219
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.76668290306868e-07,
+      "loss": 0.0003,
+      "step": 9220
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.76643935703848e-07,
+      "loss": 0.0005,
+      "step": 9221
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.7661958110082806e-07,
+      "loss": 1.4015,
+      "step": 9222
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.7659522649780806e-07,
+      "loss": 0.0004,
+      "step": 9223
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 2.765708718947881e-07,
+      "loss": 0.0004,
+      "step": 9224
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.7654651729176816e-07,
+      "loss": 0.0005,
+      "step": 9225
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.7652216268874816e-07,
+      "loss": 0.0004,
+      "step": 9226
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.764978080857282e-07,
+      "loss": 0.0005,
+      "step": 9227
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.764734534827082e-07,
+      "loss": 0.0004,
+      "step": 9228
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.7644909887968827e-07,
+      "loss": 0.0005,
+      "step": 9229
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.7642474427666826e-07,
+      "loss": 0.0006,
+      "step": 9230
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.764003896736483e-07,
+      "loss": 0.0004,
+      "step": 9231
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.7637603507062837e-07,
+      "loss": 0.0005,
+      "step": 9232
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.7635168046760837e-07,
+      "loss": 0.0107,
+      "step": 9233
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.763273258645884e-07,
+      "loss": 0.0005,
+      "step": 9234
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 2.763029712615684e-07,
+      "loss": 0.0004,
+      "step": 9235
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.7627861665854847e-07,
+      "loss": 0.0014,
+      "step": 9236
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.7625426205552847e-07,
+      "loss": 0.0004,
+      "step": 9237
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.762299074525085e-07,
+      "loss": 0.0007,
+      "step": 9238
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.7620555284948857e-07,
+      "loss": 0.0003,
+      "step": 9239
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.7618119824646857e-07,
+      "loss": 0.0012,
+      "step": 9240
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.761568436434486e-07,
+      "loss": 0.0006,
+      "step": 9241
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.761324890404286e-07,
+      "loss": 0.0003,
+      "step": 9242
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.7610813443740867e-07,
+      "loss": 0.0004,
+      "step": 9243
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.7608377983438867e-07,
+      "loss": 0.0006,
+      "step": 9244
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 2.760594252313687e-07,
+      "loss": 0.007,
+      "step": 9245
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.760350706283487e-07,
+      "loss": 0.0005,
+      "step": 9246
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.7601071602532877e-07,
+      "loss": 0.0004,
+      "step": 9247
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.759863614223088e-07,
+      "loss": 0.0004,
+      "step": 9248
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.759620068192888e-07,
+      "loss": 0.0006,
+      "step": 9249
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.7593765221626887e-07,
+      "loss": 0.0006,
+      "step": 9250
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.7591329761324887e-07,
+      "loss": 0.0005,
+      "step": 9251
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.758889430102289e-07,
+      "loss": 0.0004,
+      "step": 9252
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.758645884072089e-07,
+      "loss": 0.0007,
+      "step": 9253
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.75840233804189e-07,
+      "loss": 0.0005,
+      "step": 9254
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 2.75815879201169e-07,
+      "loss": 0.0006,
+      "step": 9255
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.75791524598149e-07,
+      "loss": 0.0004,
+      "step": 9256
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.757671699951291e-07,
+      "loss": 0.0005,
+      "step": 9257
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.757428153921091e-07,
+      "loss": 0.0005,
+      "step": 9258
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.757184607890891e-07,
+      "loss": 0.0006,
+      "step": 9259
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.756941061860691e-07,
+      "loss": 0.0005,
+      "step": 9260
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.756697515830492e-07,
+      "loss": 0.0009,
+      "step": 9261
+    },
+    {
+      "epoch": 9.0,
+      "eval_accuracy": 0.9990277102576568,
+      "eval_loss": 0.014952875673770905,
+      "eval_runtime": 2.8155,
+      "eval_samples_per_second": 730.61,
+      "eval_steps_per_second": 11.721,
+      "step": 9261
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.756453969800292e-07,
+      "loss": 0.0006,
+      "step": 9262
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.7562104237700923e-07,
+      "loss": 0.0004,
+      "step": 9263
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.755966877739893e-07,
+      "loss": 0.0004,
+      "step": 9264
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.755723331709693e-07,
+      "loss": 0.0004,
+      "step": 9265
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 2.7554797856794933e-07,
+      "loss": 0.0004,
+      "step": 9266
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.7552362396492933e-07,
+      "loss": 0.0004,
+      "step": 9267
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.754992693619094e-07,
+      "loss": 0.0006,
+      "step": 9268
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.754749147588894e-07,
+      "loss": 0.0006,
+      "step": 9269
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.7545056015586943e-07,
+      "loss": 0.0006,
+      "step": 9270
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.754262055528495e-07,
+      "loss": 0.0005,
+      "step": 9271
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.754018509498295e-07,
+      "loss": 0.026,
+      "step": 9272
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.7537749634680953e-07,
+      "loss": 0.0006,
+      "step": 9273
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.7535314174378953e-07,
+      "loss": 0.0005,
+      "step": 9274
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.753287871407696e-07,
+      "loss": 0.0004,
+      "step": 9275
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 2.753044325377496e-07,
+      "loss": 0.0004,
+      "step": 9276
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.7528007793472963e-07,
+      "loss": 0.0004,
+      "step": 9277
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.7525572333170963e-07,
+      "loss": 0.0006,
+      "step": 9278
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.752313687286897e-07,
+      "loss": 0.0005,
+      "step": 9279
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.7520701412566973e-07,
+      "loss": 0.0004,
+      "step": 9280
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.751826595226498e-07,
+      "loss": 0.0004,
+      "step": 9281
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.7515830491962984e-07,
+      "loss": 0.0004,
+      "step": 9282
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.7513395031660984e-07,
+      "loss": 0.0003,
+      "step": 9283
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.751095957135899e-07,
+      "loss": 0.0006,
+      "step": 9284
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.750852411105699e-07,
+      "loss": 0.0008,
+      "step": 9285
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 2.7506088650754994e-07,
+      "loss": 0.0003,
+      "step": 9286
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.7503653190453e-07,
+      "loss": 0.0006,
+      "step": 9287
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.7501217730151e-07,
+      "loss": 0.0004,
+      "step": 9288
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.7498782269849004e-07,
+      "loss": 0.0005,
+      "step": 9289
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.7496346809547004e-07,
+      "loss": 0.2588,
+      "step": 9290
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.749391134924501e-07,
+      "loss": 0.0004,
+      "step": 9291
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.749147588894301e-07,
+      "loss": 0.1992,
+      "step": 9292
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.7489040428641014e-07,
+      "loss": 0.0005,
+      "step": 9293
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.748660496833902e-07,
+      "loss": 0.0003,
+      "step": 9294
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.748416950803702e-07,
+      "loss": 0.0005,
+      "step": 9295
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.7481734047735024e-07,
+      "loss": 0.0005,
+      "step": 9296
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 2.7479298587433024e-07,
+      "loss": 0.0005,
+      "step": 9297
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.747686312713103e-07,
+      "loss": 1.8574,
+      "step": 9298
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.747442766682903e-07,
+      "loss": 0.0004,
+      "step": 9299
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.7471992206527034e-07,
+      "loss": 0.0006,
+      "step": 9300
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.7469556746225034e-07,
+      "loss": 0.0006,
+      "step": 9301
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.746712128592304e-07,
+      "loss": 0.0004,
+      "step": 9302
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.7464685825621045e-07,
+      "loss": 0.0005,
+      "step": 9303
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.7462250365319044e-07,
+      "loss": 0.0005,
+      "step": 9304
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.745981490501705e-07,
+      "loss": 0.0005,
+      "step": 9305
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.745737944471505e-07,
+      "loss": 0.0004,
+      "step": 9306
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 2.7454943984413055e-07,
+      "loss": 0.0006,
+      "step": 9307
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.7452508524111054e-07,
+      "loss": 0.0004,
+      "step": 9308
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.745007306380906e-07,
+      "loss": 0.0005,
+      "step": 9309
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.7447637603507065e-07,
+      "loss": 0.0004,
+      "step": 9310
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.7445202143205065e-07,
+      "loss": 0.5137,
+      "step": 9311
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.744276668290307e-07,
+      "loss": 0.0005,
+      "step": 9312
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.744033122260107e-07,
+      "loss": 0.0003,
+      "step": 9313
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.7437895762299075e-07,
+      "loss": 0.0006,
+      "step": 9314
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.7435460301997075e-07,
+      "loss": 0.0004,
+      "step": 9315
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.743302484169508e-07,
+      "loss": 0.0037,
+      "step": 9316
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 2.743058938139308e-07,
+      "loss": 0.0006,
+      "step": 9317
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.7428153921091085e-07,
+      "loss": 0.0003,
+      "step": 9318
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.742571846078909e-07,
+      "loss": 0.0005,
+      "step": 9319
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.742328300048709e-07,
+      "loss": 0.0004,
+      "step": 9320
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.7420847540185095e-07,
+      "loss": 0.0005,
+      "step": 9321
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.7418412079883095e-07,
+      "loss": 0.0004,
+      "step": 9322
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.74159766195811e-07,
+      "loss": 0.0004,
+      "step": 9323
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.74135411592791e-07,
+      "loss": 0.0008,
+      "step": 9324
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.7411105698977105e-07,
+      "loss": 0.0006,
+      "step": 9325
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.740867023867511e-07,
+      "loss": 0.0004,
+      "step": 9326
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 2.740623477837311e-07,
+      "loss": 0.0003,
+      "step": 9327
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7403799318071115e-07,
+      "loss": 0.0005,
+      "step": 9328
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7401363857769115e-07,
+      "loss": 0.0005,
+      "step": 9329
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.739892839746712e-07,
+      "loss": 0.0004,
+      "step": 9330
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.739649293716512e-07,
+      "loss": 0.0005,
+      "step": 9331
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7394057476863125e-07,
+      "loss": 0.0004,
+      "step": 9332
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7391622016561125e-07,
+      "loss": 0.0005,
+      "step": 9333
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.738918655625913e-07,
+      "loss": 0.0005,
+      "step": 9334
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7386751095957136e-07,
+      "loss": 0.0004,
+      "step": 9335
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.7384315635655135e-07,
+      "loss": 0.0003,
+      "step": 9336
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.738188017535314e-07,
+      "loss": 0.0005,
+      "step": 9337
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 2.737944471505114e-07,
+      "loss": 0.0005,
+      "step": 9338
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7377009254749146e-07,
+      "loss": 0.0004,
+      "step": 9339
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7374573794447146e-07,
+      "loss": 2.3684,
+      "step": 9340
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.737213833414515e-07,
+      "loss": 0.0005,
+      "step": 9341
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7369702873843156e-07,
+      "loss": 0.0004,
+      "step": 9342
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7367267413541156e-07,
+      "loss": 0.0008,
+      "step": 9343
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.736483195323916e-07,
+      "loss": 0.0008,
+      "step": 9344
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.736239649293716e-07,
+      "loss": 0.0004,
+      "step": 9345
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7359961032635166e-07,
+      "loss": 0.0007,
+      "step": 9346
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.7357525572333166e-07,
+      "loss": 0.0005,
+      "step": 9347
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 2.735509011203117e-07,
+      "loss": 0.0004,
+      "step": 9348
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.735265465172917e-07,
+      "loss": 0.0004,
+      "step": 9349
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7350219191427176e-07,
+      "loss": 0.0005,
+      "step": 9350
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.734778373112518e-07,
+      "loss": 0.0008,
+      "step": 9351
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.734534827082318e-07,
+      "loss": 0.0006,
+      "step": 9352
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7342912810521186e-07,
+      "loss": 0.0008,
+      "step": 9353
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7340477350219186e-07,
+      "loss": 0.0008,
+      "step": 9354
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.733804188991719e-07,
+      "loss": 0.0005,
+      "step": 9355
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.733560642961519e-07,
+      "loss": 0.0004,
+      "step": 9356
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7333170969313196e-07,
+      "loss": 0.0004,
+      "step": 9357
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 2.7330735509011207e-07,
+      "loss": 0.0004,
+      "step": 9358
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7328300048709207e-07,
+      "loss": 0.0006,
+      "step": 9359
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.732586458840721e-07,
+      "loss": 0.0006,
+      "step": 9360
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.732342912810521e-07,
+      "loss": 0.0006,
+      "step": 9361
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7320993667803217e-07,
+      "loss": 0.0004,
+      "step": 9362
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7318558207501217e-07,
+      "loss": 0.0007,
+      "step": 9363
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.731612274719922e-07,
+      "loss": 0.0005,
+      "step": 9364
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7313687286897227e-07,
+      "loss": 0.0005,
+      "step": 9365
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7311251826595227e-07,
+      "loss": 0.0005,
+      "step": 9366
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.730881636629323e-07,
+      "loss": 0.0004,
+      "step": 9367
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.730638090599123e-07,
+      "loss": 0.0005,
+      "step": 9368
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 2.7303945445689237e-07,
+      "loss": 0.0005,
+      "step": 9369
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.7301509985387237e-07,
+      "loss": 0.0004,
+      "step": 9370
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.729907452508524e-07,
+      "loss": 0.0006,
+      "step": 9371
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.729663906478324e-07,
+      "loss": 0.0005,
+      "step": 9372
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.7294203604481247e-07,
+      "loss": 0.0004,
+      "step": 9373
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.729176814417925e-07,
+      "loss": 0.0004,
+      "step": 9374
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.728933268387725e-07,
+      "loss": 0.0006,
+      "step": 9375
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.7286897223575257e-07,
+      "loss": 0.0005,
+      "step": 9376
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.7284461763273257e-07,
+      "loss": 0.0004,
+      "step": 9377
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.728202630297126e-07,
+      "loss": 0.0006,
+      "step": 9378
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 2.727959084266926e-07,
+      "loss": 0.0006,
+      "step": 9379
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.727715538236727e-07,
+      "loss": 0.0004,
+      "step": 9380
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.727471992206527e-07,
+      "loss": 0.0004,
+      "step": 9381
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.727228446176327e-07,
+      "loss": 0.0005,
+      "step": 9382
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.726984900146128e-07,
+      "loss": 0.0005,
+      "step": 9383
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.726741354115928e-07,
+      "loss": 0.0005,
+      "step": 9384
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.726497808085728e-07,
+      "loss": 0.0005,
+      "step": 9385
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.726254262055528e-07,
+      "loss": 0.0005,
+      "step": 9386
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.726010716025329e-07,
+      "loss": 0.0006,
+      "step": 9387
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.725767169995129e-07,
+      "loss": 0.0004,
+      "step": 9388
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 2.7255236239649293e-07,
+      "loss": 0.0005,
+      "step": 9389
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.72528007793473e-07,
+      "loss": 0.0004,
+      "step": 9390
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.72503653190453e-07,
+      "loss": 0.0006,
+      "step": 9391
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.7247929858743303e-07,
+      "loss": 0.0005,
+      "step": 9392
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.7245494398441303e-07,
+      "loss": 0.0006,
+      "step": 9393
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.724305893813931e-07,
+      "loss": 0.0005,
+      "step": 9394
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.724062347783731e-07,
+      "loss": 0.0006,
+      "step": 9395
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.7238188017535313e-07,
+      "loss": 0.0004,
+      "step": 9396
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.723575255723332e-07,
+      "loss": 0.0004,
+      "step": 9397
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.723331709693132e-07,
+      "loss": 0.0004,
+      "step": 9398
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 2.7230881636629323e-07,
+      "loss": 0.0007,
+      "step": 9399
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7228446176327323e-07,
+      "loss": 0.0004,
+      "step": 9400
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.722601071602533e-07,
+      "loss": 0.0004,
+      "step": 9401
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.722357525572333e-07,
+      "loss": 0.0004,
+      "step": 9402
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7221139795421333e-07,
+      "loss": 0.0008,
+      "step": 9403
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7218704335119333e-07,
+      "loss": 0.0005,
+      "step": 9404
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.721626887481734e-07,
+      "loss": 0.0005,
+      "step": 9405
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7213833414515343e-07,
+      "loss": 0.0004,
+      "step": 9406
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7211397954213343e-07,
+      "loss": 0.0004,
+      "step": 9407
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.720896249391135e-07,
+      "loss": 0.0003,
+      "step": 9408
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.720652703360935e-07,
+      "loss": 0.0004,
+      "step": 9409
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 2.7204091573307353e-07,
+      "loss": 0.0002,
+      "step": 9410
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7201656113005353e-07,
+      "loss": 0.0004,
+      "step": 9411
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.719922065270336e-07,
+      "loss": 0.0006,
+      "step": 9412
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7196785192401364e-07,
+      "loss": 0.0004,
+      "step": 9413
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7194349732099364e-07,
+      "loss": 0.0007,
+      "step": 9414
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.719191427179737e-07,
+      "loss": 0.0008,
+      "step": 9415
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.718947881149537e-07,
+      "loss": 0.0004,
+      "step": 9416
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7187043351193374e-07,
+      "loss": 0.0003,
+      "step": 9417
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7184607890891374e-07,
+      "loss": 0.0003,
+      "step": 9418
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.718217243058938e-07,
+      "loss": 0.0004,
+      "step": 9419
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 2.7179736970287384e-07,
+      "loss": 0.0005,
+      "step": 9420
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.7177301509985384e-07,
+      "loss": 0.0007,
+      "step": 9421
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.717486604968339e-07,
+      "loss": 0.0005,
+      "step": 9422
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.717243058938139e-07,
+      "loss": 0.0003,
+      "step": 9423
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.7169995129079394e-07,
+      "loss": 0.0004,
+      "step": 9424
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.7167559668777394e-07,
+      "loss": 0.0005,
+      "step": 9425
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.71651242084754e-07,
+      "loss": 0.0006,
+      "step": 9426
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.71626887481734e-07,
+      "loss": 0.0005,
+      "step": 9427
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.7160253287871404e-07,
+      "loss": 0.0005,
+      "step": 9428
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.715781782756941e-07,
+      "loss": 0.0004,
+      "step": 9429
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 2.715538236726741e-07,
+      "loss": 0.0004,
+      "step": 9430
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7152946906965414e-07,
+      "loss": 0.0003,
+      "step": 9431
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7150511446663414e-07,
+      "loss": 0.0005,
+      "step": 9432
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.714807598636142e-07,
+      "loss": 0.0004,
+      "step": 9433
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.714564052605942e-07,
+      "loss": 0.0006,
+      "step": 9434
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.714320506575743e-07,
+      "loss": 0.0005,
+      "step": 9435
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7140769605455435e-07,
+      "loss": 0.0003,
+      "step": 9436
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7138334145153435e-07,
+      "loss": 0.0006,
+      "step": 9437
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.713589868485144e-07,
+      "loss": 0.0006,
+      "step": 9438
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.713346322454944e-07,
+      "loss": 0.0005,
+      "step": 9439
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7131027764247445e-07,
+      "loss": 0.0005,
+      "step": 9440
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 2.7128592303945445e-07,
+      "loss": 0.0007,
+      "step": 9441
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.712615684364345e-07,
+      "loss": 0.0004,
+      "step": 9442
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.7123721383341455e-07,
+      "loss": 0.0005,
+      "step": 9443
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.7121285923039455e-07,
+      "loss": 0.0005,
+      "step": 9444
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.711885046273746e-07,
+      "loss": 0.0003,
+      "step": 9445
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.711641500243546e-07,
+      "loss": 0.0003,
+      "step": 9446
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.7113979542133465e-07,
+      "loss": 0.0006,
+      "step": 9447
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.7111544081831465e-07,
+      "loss": 0.0004,
+      "step": 9448
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.710910862152947e-07,
+      "loss": 0.0006,
+      "step": 9449
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.710667316122747e-07,
+      "loss": 0.0007,
+      "step": 9450
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 2.7104237700925475e-07,
+      "loss": 0.0005,
+      "step": 9451
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.710180224062348e-07,
+      "loss": 0.0005,
+      "step": 9452
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.709936678032148e-07,
+      "loss": 0.0005,
+      "step": 9453
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.7096931320019485e-07,
+      "loss": 0.0007,
+      "step": 9454
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.7094495859717485e-07,
+      "loss": 0.0005,
+      "step": 9455
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.709206039941549e-07,
+      "loss": 0.0012,
+      "step": 9456
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.708962493911349e-07,
+      "loss": 0.0004,
+      "step": 9457
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.7087189478811495e-07,
+      "loss": 0.0005,
+      "step": 9458
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.70847540185095e-07,
+      "loss": 0.0005,
+      "step": 9459
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.70823185582075e-07,
+      "loss": 0.0007,
+      "step": 9460
+    },
+    {
+      "epoch": 9.19,
+      "learning_rate": 2.7079883097905506e-07,
+      "loss": 0.0004,
+      "step": 9461
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.7077447637603506e-07,
+      "loss": 0.0004,
+      "step": 9462
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.707501217730151e-07,
+      "loss": 0.0004,
+      "step": 9463
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.707257671699951e-07,
+      "loss": 0.0004,
+      "step": 9464
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.7070141256697516e-07,
+      "loss": 0.0007,
+      "step": 9465
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.7067705796395516e-07,
+      "loss": 0.0009,
+      "step": 9466
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.706527033609352e-07,
+      "loss": 0.0005,
+      "step": 9467
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.7062834875791526e-07,
+      "loss": 0.0008,
+      "step": 9468
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.7060399415489526e-07,
+      "loss": 0.0004,
+      "step": 9469
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.705796395518753e-07,
+      "loss": 0.0006,
+      "step": 9470
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 2.705552849488553e-07,
+      "loss": 0.0004,
+      "step": 9471
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7053093034583536e-07,
+      "loss": 0.0003,
+      "step": 9472
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7050657574281536e-07,
+      "loss": 0.0012,
+      "step": 9473
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.704822211397954e-07,
+      "loss": 0.0006,
+      "step": 9474
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7045786653677546e-07,
+      "loss": 0.0006,
+      "step": 9475
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7043351193375546e-07,
+      "loss": 0.0148,
+      "step": 9476
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.704091573307355e-07,
+      "loss": 0.0005,
+      "step": 9477
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.703848027277155e-07,
+      "loss": 0.0004,
+      "step": 9478
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7036044812469556e-07,
+      "loss": 0.0005,
+      "step": 9479
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.7033609352167556e-07,
+      "loss": 0.0005,
+      "step": 9480
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.703117389186556e-07,
+      "loss": 0.0003,
+      "step": 9481
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 2.702873843156356e-07,
+      "loss": 0.1715,
+      "step": 9482
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.7026302971261566e-07,
+      "loss": 0.0003,
+      "step": 9483
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.702386751095957e-07,
+      "loss": 0.0004,
+      "step": 9484
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.702143205065757e-07,
+      "loss": 0.0003,
+      "step": 9485
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.7018996590355577e-07,
+      "loss": 0.001,
+      "step": 9486
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.7016561130053576e-07,
+      "loss": 0.0005,
+      "step": 9487
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.701412566975158e-07,
+      "loss": 0.0003,
+      "step": 9488
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.701169020944958e-07,
+      "loss": 0.0006,
+      "step": 9489
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.7009254749147587e-07,
+      "loss": 0.0006,
+      "step": 9490
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.700681928884559e-07,
+      "loss": 0.0007,
+      "step": 9491
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 2.700438382854359e-07,
+      "loss": 0.0005,
+      "step": 9492
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.7001948368241597e-07,
+      "loss": 0.0078,
+      "step": 9493
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.6999512907939597e-07,
+      "loss": 0.0005,
+      "step": 9494
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.69970774476376e-07,
+      "loss": 0.0006,
+      "step": 9495
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.69946419873356e-07,
+      "loss": 0.0003,
+      "step": 9496
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.6992206527033607e-07,
+      "loss": 0.0006,
+      "step": 9497
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.6989771066731607e-07,
+      "loss": 0.0004,
+      "step": 9498
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.698733560642961e-07,
+      "loss": 0.0003,
+      "step": 9499
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.6984900146127617e-07,
+      "loss": 0.0003,
+      "step": 9500
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.6982464685825617e-07,
+      "loss": 0.0004,
+      "step": 9501
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 2.698002922552362e-07,
+      "loss": 0.0004,
+      "step": 9502
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.697759376522162e-07,
+      "loss": 0.0004,
+      "step": 9503
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.6975158304919627e-07,
+      "loss": 0.0004,
+      "step": 9504
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.6972722844617627e-07,
+      "loss": 0.0005,
+      "step": 9505
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.697028738431563e-07,
+      "loss": 0.0006,
+      "step": 9506
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.6967851924013637e-07,
+      "loss": 0.0003,
+      "step": 9507
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.6965416463711637e-07,
+      "loss": 0.0006,
+      "step": 9508
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.696298100340964e-07,
+      "loss": 0.0005,
+      "step": 9509
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.696054554310764e-07,
+      "loss": 0.0004,
+      "step": 9510
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.6958110082805653e-07,
+      "loss": 0.0005,
+      "step": 9511
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.695567462250365e-07,
+      "loss": 0.0005,
+      "step": 9512
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 2.695323916220166e-07,
+      "loss": 0.0005,
+      "step": 9513
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.6950803701899663e-07,
+      "loss": 0.0003,
+      "step": 9514
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.6948368241597663e-07,
+      "loss": 0.0006,
+      "step": 9515
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.694593278129567e-07,
+      "loss": 0.0004,
+      "step": 9516
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.694349732099367e-07,
+      "loss": 0.0004,
+      "step": 9517
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.6941061860691673e-07,
+      "loss": 0.0004,
+      "step": 9518
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.6938626400389673e-07,
+      "loss": 0.0007,
+      "step": 9519
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.693619094008768e-07,
+      "loss": 0.0003,
+      "step": 9520
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.693375547978568e-07,
+      "loss": 0.0004,
+      "step": 9521
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.6931320019483683e-07,
+      "loss": 0.0005,
+      "step": 9522
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 2.692888455918169e-07,
+      "loss": 0.0005,
+      "step": 9523
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.692644909887969e-07,
+      "loss": 0.0005,
+      "step": 9524
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.6924013638577693e-07,
+      "loss": 0.0475,
+      "step": 9525
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.6921578178275693e-07,
+      "loss": 0.0004,
+      "step": 9526
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.69191427179737e-07,
+      "loss": 0.0005,
+      "step": 9527
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.69167072576717e-07,
+      "loss": 0.0004,
+      "step": 9528
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.6914271797369703e-07,
+      "loss": 0.0005,
+      "step": 9529
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.691183633706771e-07,
+      "loss": 0.0005,
+      "step": 9530
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.690940087676571e-07,
+      "loss": 0.0004,
+      "step": 9531
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.6906965416463713e-07,
+      "loss": 0.0004,
+      "step": 9532
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 2.6904529956161713e-07,
+      "loss": 0.0005,
+      "step": 9533
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.690209449585972e-07,
+      "loss": 0.0004,
+      "step": 9534
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.689965903555772e-07,
+      "loss": 0.0004,
+      "step": 9535
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.6897223575255724e-07,
+      "loss": 0.0004,
+      "step": 9536
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.6894788114953723e-07,
+      "loss": 0.0004,
+      "step": 9537
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.689235265465173e-07,
+      "loss": 0.0004,
+      "step": 9538
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.6889917194349734e-07,
+      "loss": 0.0005,
+      "step": 9539
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.6887481734047734e-07,
+      "loss": 0.0003,
+      "step": 9540
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.688504627374574e-07,
+      "loss": 0.0006,
+      "step": 9541
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.688261081344374e-07,
+      "loss": 0.0005,
+      "step": 9542
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 2.6880175353141744e-07,
+      "loss": 0.0004,
+      "step": 9543
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.6877739892839744e-07,
+      "loss": 0.0004,
+      "step": 9544
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.687530443253775e-07,
+      "loss": 0.0004,
+      "step": 9545
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.6872868972235754e-07,
+      "loss": 0.0006,
+      "step": 9546
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.6870433511933754e-07,
+      "loss": 0.0003,
+      "step": 9547
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.686799805163176e-07,
+      "loss": 0.0006,
+      "step": 9548
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.686556259132976e-07,
+      "loss": 0.0005,
+      "step": 9549
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.6863127131027764e-07,
+      "loss": 0.0002,
+      "step": 9550
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.6860691670725764e-07,
+      "loss": 0.0007,
+      "step": 9551
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.685825621042377e-07,
+      "loss": 0.0006,
+      "step": 9552
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.685582075012177e-07,
+      "loss": 0.0003,
+      "step": 9553
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 2.6853385289819774e-07,
+      "loss": 0.0005,
+      "step": 9554
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.685094982951778e-07,
+      "loss": 0.0004,
+      "step": 9555
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.684851436921578e-07,
+      "loss": 0.0004,
+      "step": 9556
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.6846078908913784e-07,
+      "loss": 0.0004,
+      "step": 9557
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.6843643448611784e-07,
+      "loss": 0.0003,
+      "step": 9558
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.684120798830979e-07,
+      "loss": 0.0003,
+      "step": 9559
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.683877252800779e-07,
+      "loss": 0.0004,
+      "step": 9560
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.6836337067705794e-07,
+      "loss": 0.0004,
+      "step": 9561
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.68339016074038e-07,
+      "loss": 0.0004,
+      "step": 9562
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.68314661471018e-07,
+      "loss": 0.0006,
+      "step": 9563
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 2.6829030686799805e-07,
+      "loss": 0.0006,
+      "step": 9564
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.6826595226497804e-07,
+      "loss": 0.0004,
+      "step": 9565
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.682415976619581e-07,
+      "loss": 0.0004,
+      "step": 9566
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.682172430589381e-07,
+      "loss": 0.0005,
+      "step": 9567
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.6819288845591815e-07,
+      "loss": 0.0005,
+      "step": 9568
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.6816853385289814e-07,
+      "loss": 0.0004,
+      "step": 9569
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.681441792498782e-07,
+      "loss": 0.0003,
+      "step": 9570
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.6811982464685825e-07,
+      "loss": 0.0004,
+      "step": 9571
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.6809547004383825e-07,
+      "loss": 0.0004,
+      "step": 9572
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.680711154408183e-07,
+      "loss": 0.0004,
+      "step": 9573
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 2.680467608377983e-07,
+      "loss": 0.0004,
+      "step": 9574
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.6802240623477835e-07,
+      "loss": 0.0005,
+      "step": 9575
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.6799805163175835e-07,
+      "loss": 0.0003,
+      "step": 9576
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.679736970287384e-07,
+      "loss": 0.0005,
+      "step": 9577
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.6794934242571845e-07,
+      "loss": 0.0008,
+      "step": 9578
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.6792498782269845e-07,
+      "loss": 0.0003,
+      "step": 9579
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.679006332196785e-07,
+      "loss": 0.0003,
+      "step": 9580
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.678762786166585e-07,
+      "loss": 0.0004,
+      "step": 9581
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.6785192401363855e-07,
+      "loss": 0.0003,
+      "step": 9582
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.6782756941061855e-07,
+      "loss": 0.0007,
+      "step": 9583
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.678032148075986e-07,
+      "loss": 0.0006,
+      "step": 9584
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 2.6777886020457865e-07,
+      "loss": 0.0006,
+      "step": 9585
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.6775450560155865e-07,
+      "loss": 0.0005,
+      "step": 9586
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.677301509985387e-07,
+      "loss": 0.0013,
+      "step": 9587
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.6770579639551876e-07,
+      "loss": 0.0005,
+      "step": 9588
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.676814417924988e-07,
+      "loss": 0.0005,
+      "step": 9589
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.676570871894788e-07,
+      "loss": 0.0004,
+      "step": 9590
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.6763273258645886e-07,
+      "loss": 0.0003,
+      "step": 9591
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.6760837798343886e-07,
+      "loss": 0.0005,
+      "step": 9592
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.675840233804189e-07,
+      "loss": 0.9698,
+      "step": 9593
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.6755966877739896e-07,
+      "loss": 0.0004,
+      "step": 9594
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 2.6753531417437896e-07,
+      "loss": 0.0004,
+      "step": 9595
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.67510959571359e-07,
+      "loss": 0.0004,
+      "step": 9596
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.67486604968339e-07,
+      "loss": 0.0005,
+      "step": 9597
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.6746225036531906e-07,
+      "loss": 0.0003,
+      "step": 9598
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.6743789576229906e-07,
+      "loss": 0.0004,
+      "step": 9599
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.674135411592791e-07,
+      "loss": 0.0005,
+      "step": 9600
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.6738918655625916e-07,
+      "loss": 0.0004,
+      "step": 9601
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.6736483195323916e-07,
+      "loss": 0.0004,
+      "step": 9602
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.673404773502192e-07,
+      "loss": 0.0004,
+      "step": 9603
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.673161227471992e-07,
+      "loss": 0.0007,
+      "step": 9604
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 2.6729176814417926e-07,
+      "loss": 0.0005,
+      "step": 9605
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.6726741354115926e-07,
+      "loss": 0.0036,
+      "step": 9606
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.672430589381393e-07,
+      "loss": 0.0004,
+      "step": 9607
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.6721870433511936e-07,
+      "loss": 0.0006,
+      "step": 9608
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.6719434973209936e-07,
+      "loss": 0.0004,
+      "step": 9609
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.671699951290794e-07,
+      "loss": 0.0005,
+      "step": 9610
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.671456405260594e-07,
+      "loss": 0.0004,
+      "step": 9611
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.6712128592303947e-07,
+      "loss": 0.0004,
+      "step": 9612
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.6709693132001946e-07,
+      "loss": 0.0003,
+      "step": 9613
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.670725767169995e-07,
+      "loss": 0.0004,
+      "step": 9614
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.670482221139795e-07,
+      "loss": 0.0005,
+      "step": 9615
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 2.6702386751095957e-07,
+      "loss": 0.0005,
+      "step": 9616
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.669995129079396e-07,
+      "loss": 0.0004,
+      "step": 9617
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.669751583049196e-07,
+      "loss": 0.0005,
+      "step": 9618
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.6695080370189967e-07,
+      "loss": 0.0005,
+      "step": 9619
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.6692644909887967e-07,
+      "loss": 0.0003,
+      "step": 9620
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.669020944958597e-07,
+      "loss": 0.0004,
+      "step": 9621
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.668777398928397e-07,
+      "loss": 0.0004,
+      "step": 9622
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.6685338528981977e-07,
+      "loss": 0.0006,
+      "step": 9623
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.668290306867998e-07,
+      "loss": 0.0004,
+      "step": 9624
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.668046760837798e-07,
+      "loss": 0.0004,
+      "step": 9625
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 2.6678032148075987e-07,
+      "loss": 0.0002,
+      "step": 9626
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.6675596687773987e-07,
+      "loss": 0.0004,
+      "step": 9627
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.667316122747199e-07,
+      "loss": 0.0007,
+      "step": 9628
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.667072576716999e-07,
+      "loss": 0.0003,
+      "step": 9629
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.6668290306867997e-07,
+      "loss": 0.0008,
+      "step": 9630
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.6665854846565997e-07,
+      "loss": 0.0006,
+      "step": 9631
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.6663419386264e-07,
+      "loss": 0.0007,
+      "step": 9632
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.6660983925962007e-07,
+      "loss": 0.0004,
+      "step": 9633
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.6658548465660007e-07,
+      "loss": 0.0004,
+      "step": 9634
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.665611300535801e-07,
+      "loss": 0.0004,
+      "step": 9635
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 2.665367754505601e-07,
+      "loss": 0.0004,
+      "step": 9636
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.665124208475402e-07,
+      "loss": 0.0005,
+      "step": 9637
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.6648806624452017e-07,
+      "loss": 0.0003,
+      "step": 9638
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.664637116415002e-07,
+      "loss": 0.0007,
+      "step": 9639
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.664393570384803e-07,
+      "loss": 0.0002,
+      "step": 9640
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.664150024354603e-07,
+      "loss": 0.0007,
+      "step": 9641
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.663906478324403e-07,
+      "loss": 0.0004,
+      "step": 9642
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.663662932294203e-07,
+      "loss": 0.0005,
+      "step": 9643
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.663419386264004e-07,
+      "loss": 0.0003,
+      "step": 9644
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.663175840233804e-07,
+      "loss": 0.0006,
+      "step": 9645
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 2.6629322942036043e-07,
+      "loss": 0.0005,
+      "step": 9646
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.662688748173404e-07,
+      "loss": 0.0005,
+      "step": 9647
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.662445202143205e-07,
+      "loss": 1.3187,
+      "step": 9648
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.6622016561130053e-07,
+      "loss": 0.0004,
+      "step": 9649
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.6619581100828053e-07,
+      "loss": 0.0006,
+      "step": 9650
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.661714564052606e-07,
+      "loss": 0.0005,
+      "step": 9651
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.661471018022406e-07,
+      "loss": 0.0005,
+      "step": 9652
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.6612274719922063e-07,
+      "loss": 0.0005,
+      "step": 9653
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.6609839259620063e-07,
+      "loss": 0.0004,
+      "step": 9654
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.660740379931807e-07,
+      "loss": 0.0005,
+      "step": 9655
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.6604968339016073e-07,
+      "loss": 0.0004,
+      "step": 9656
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 2.6602532878714073e-07,
+      "loss": 0.0015,
+      "step": 9657
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.660009741841208e-07,
+      "loss": 0.0005,
+      "step": 9658
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.659766195811008e-07,
+      "loss": 0.0004,
+      "step": 9659
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.6595226497808083e-07,
+      "loss": 0.0003,
+      "step": 9660
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.6592791037506083e-07,
+      "loss": 0.0006,
+      "step": 9661
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.659035557720409e-07,
+      "loss": 0.0005,
+      "step": 9662
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.658792011690209e-07,
+      "loss": 0.0005,
+      "step": 9663
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.65854846566001e-07,
+      "loss": 0.0004,
+      "step": 9664
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.6583049196298104e-07,
+      "loss": 0.0005,
+      "step": 9665
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.6580613735996104e-07,
+      "loss": 0.0004,
+      "step": 9666
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 2.657817827569411e-07,
+      "loss": 0.0006,
+      "step": 9667
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.657574281539211e-07,
+      "loss": 0.0003,
+      "step": 9668
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.6573307355090114e-07,
+      "loss": 2.0985,
+      "step": 9669
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.6570871894788114e-07,
+      "loss": 0.0004,
+      "step": 9670
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.656843643448612e-07,
+      "loss": 0.0003,
+      "step": 9671
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.6566000974184124e-07,
+      "loss": 0.0005,
+      "step": 9672
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.6563565513882124e-07,
+      "loss": 0.0198,
+      "step": 9673
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.656113005358013e-07,
+      "loss": 0.0005,
+      "step": 9674
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.655869459327813e-07,
+      "loss": 0.0004,
+      "step": 9675
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.6556259132976134e-07,
+      "loss": 0.0003,
+      "step": 9676
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 2.6553823672674134e-07,
+      "loss": 0.0003,
+      "step": 9677
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.655138821237214e-07,
+      "loss": 0.0004,
+      "step": 9678
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.6548952752070144e-07,
+      "loss": 0.0006,
+      "step": 9679
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.6546517291768144e-07,
+      "loss": 0.0006,
+      "step": 9680
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.654408183146615e-07,
+      "loss": 0.0004,
+      "step": 9681
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.654164637116415e-07,
+      "loss": 0.0005,
+      "step": 9682
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.6539210910862154e-07,
+      "loss": 0.0005,
+      "step": 9683
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.6536775450560154e-07,
+      "loss": 0.0004,
+      "step": 9684
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.653433999025816e-07,
+      "loss": 0.0003,
+      "step": 9685
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.653190452995616e-07,
+      "loss": 0.0005,
+      "step": 9686
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.6529469069654164e-07,
+      "loss": 0.0004,
+      "step": 9687
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 2.652703360935217e-07,
+      "loss": 0.0005,
+      "step": 9688
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.652459814905017e-07,
+      "loss": 0.0005,
+      "step": 9689
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.6522162688748175e-07,
+      "loss": 0.0004,
+      "step": 9690
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.6519727228446174e-07,
+      "loss": 0.0003,
+      "step": 9691
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.651729176814418e-07,
+      "loss": 0.0004,
+      "step": 9692
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.651485630784218e-07,
+      "loss": 0.0003,
+      "step": 9693
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.6512420847540185e-07,
+      "loss": 0.0005,
+      "step": 9694
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.650998538723819e-07,
+      "loss": 0.0007,
+      "step": 9695
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.650754992693619e-07,
+      "loss": 0.0008,
+      "step": 9696
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.6505114466634195e-07,
+      "loss": 0.0004,
+      "step": 9697
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 2.6502679006332195e-07,
+      "loss": 0.0005,
+      "step": 9698
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.65002435460302e-07,
+      "loss": 0.0004,
+      "step": 9699
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.64978080857282e-07,
+      "loss": 0.0004,
+      "step": 9700
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.6495372625426205e-07,
+      "loss": 0.0004,
+      "step": 9701
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.6492937165124205e-07,
+      "loss": 0.002,
+      "step": 9702
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.649050170482221e-07,
+      "loss": 0.0004,
+      "step": 9703
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.6488066244520215e-07,
+      "loss": 0.0006,
+      "step": 9704
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.6485630784218215e-07,
+      "loss": 0.0005,
+      "step": 9705
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.648319532391622e-07,
+      "loss": 0.0005,
+      "step": 9706
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.648075986361422e-07,
+      "loss": 0.0004,
+      "step": 9707
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 2.6478324403312225e-07,
+      "loss": 0.0004,
+      "step": 9708
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.6475888943010225e-07,
+      "loss": 0.0004,
+      "step": 9709
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.647345348270823e-07,
+      "loss": 0.0004,
+      "step": 9710
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.6471018022406235e-07,
+      "loss": 0.0003,
+      "step": 9711
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.6468582562104235e-07,
+      "loss": 0.0006,
+      "step": 9712
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.646614710180224e-07,
+      "loss": 0.0006,
+      "step": 9713
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.646371164150024e-07,
+      "loss": 0.0004,
+      "step": 9714
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.6461276181198245e-07,
+      "loss": 0.0004,
+      "step": 9715
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.6458840720896245e-07,
+      "loss": 0.0004,
+      "step": 9716
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.645640526059425e-07,
+      "loss": 0.0004,
+      "step": 9717
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 2.645396980029225e-07,
+      "loss": 0.0004,
+      "step": 9718
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.6451534339990255e-07,
+      "loss": 0.0005,
+      "step": 9719
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.644909887968826e-07,
+      "loss": 0.0003,
+      "step": 9720
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.644666341938626e-07,
+      "loss": 0.0003,
+      "step": 9721
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.6444227959084266e-07,
+      "loss": 0.0005,
+      "step": 9722
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.6441792498782266e-07,
+      "loss": 0.0003,
+      "step": 9723
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.643935703848027e-07,
+      "loss": 0.0005,
+      "step": 9724
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.643692157817827e-07,
+      "loss": 0.0003,
+      "step": 9725
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.6434486117876276e-07,
+      "loss": 0.0004,
+      "step": 9726
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.643205065757428e-07,
+      "loss": 0.0005,
+      "step": 9727
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.642961519727228e-07,
+      "loss": 0.0006,
+      "step": 9728
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 2.6427179736970286e-07,
+      "loss": 0.0006,
+      "step": 9729
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.6424744276668286e-07,
+      "loss": 0.0003,
+      "step": 9730
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.642230881636629e-07,
+      "loss": 0.0004,
+      "step": 9731
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.641987335606429e-07,
+      "loss": 0.0003,
+      "step": 9732
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.6417437895762296e-07,
+      "loss": 0.0005,
+      "step": 9733
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.6415002435460296e-07,
+      "loss": 0.0004,
+      "step": 9734
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.64125669751583e-07,
+      "loss": 0.0008,
+      "step": 9735
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.6410131514856306e-07,
+      "loss": 0.0003,
+      "step": 9736
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.6407696054554306e-07,
+      "loss": 0.0003,
+      "step": 9737
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.640526059425231e-07,
+      "loss": 0.0005,
+      "step": 9738
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 2.640282513395031e-07,
+      "loss": 0.0005,
+      "step": 9739
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.6400389673648316e-07,
+      "loss": 0.0003,
+      "step": 9740
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.639795421334632e-07,
+      "loss": 0.0008,
+      "step": 9741
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.6395518753044327e-07,
+      "loss": 0.0006,
+      "step": 9742
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.639308329274233e-07,
+      "loss": 0.0006,
+      "step": 9743
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.639064783244033e-07,
+      "loss": 0.0004,
+      "step": 9744
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.6388212372138337e-07,
+      "loss": 0.0004,
+      "step": 9745
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.6385776911836337e-07,
+      "loss": 0.0005,
+      "step": 9746
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.638334145153434e-07,
+      "loss": 0.0007,
+      "step": 9747
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.638090599123234e-07,
+      "loss": 0.0004,
+      "step": 9748
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 2.6378470530930347e-07,
+      "loss": 0.0005,
+      "step": 9749
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.637603507062835e-07,
+      "loss": 0.0003,
+      "step": 9750
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.637359961032635e-07,
+      "loss": 0.0002,
+      "step": 9751
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.6371164150024357e-07,
+      "loss": 0.0004,
+      "step": 9752
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.6368728689722357e-07,
+      "loss": 0.0005,
+      "step": 9753
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.636629322942036e-07,
+      "loss": 0.0004,
+      "step": 9754
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.636385776911836e-07,
+      "loss": 0.0004,
+      "step": 9755
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.6361422308816367e-07,
+      "loss": 0.0003,
+      "step": 9756
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.6358986848514367e-07,
+      "loss": 0.0006,
+      "step": 9757
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.635655138821237e-07,
+      "loss": 0.0004,
+      "step": 9758
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.6354115927910377e-07,
+      "loss": 0.0004,
+      "step": 9759
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 2.6351680467608377e-07,
+      "loss": 0.0005,
+      "step": 9760
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.634924500730638e-07,
+      "loss": 0.0004,
+      "step": 9761
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.634680954700438e-07,
+      "loss": 0.0005,
+      "step": 9762
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.634437408670239e-07,
+      "loss": 0.0004,
+      "step": 9763
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.6341938626400387e-07,
+      "loss": 0.0005,
+      "step": 9764
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.633950316609839e-07,
+      "loss": 0.0005,
+      "step": 9765
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.63370677057964e-07,
+      "loss": 0.0005,
+      "step": 9766
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.63346322454944e-07,
+      "loss": 0.0005,
+      "step": 9767
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.63321967851924e-07,
+      "loss": 0.0004,
+      "step": 9768
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.63297613248904e-07,
+      "loss": 0.0003,
+      "step": 9769
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 2.632732586458841e-07,
+      "loss": 0.0004,
+      "step": 9770
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.632489040428641e-07,
+      "loss": 0.0004,
+      "step": 9771
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.6322454943984413e-07,
+      "loss": 0.0002,
+      "step": 9772
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.632001948368241e-07,
+      "loss": 0.0005,
+      "step": 9773
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.631758402338042e-07,
+      "loss": 0.001,
+      "step": 9774
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.6315148563078423e-07,
+      "loss": 0.0005,
+      "step": 9775
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.6312713102776423e-07,
+      "loss": 0.0004,
+      "step": 9776
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.631027764247443e-07,
+      "loss": 0.0004,
+      "step": 9777
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.630784218217243e-07,
+      "loss": 0.0003,
+      "step": 9778
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.6305406721870433e-07,
+      "loss": 0.0004,
+      "step": 9779
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 2.6302971261568433e-07,
+      "loss": 0.0004,
+      "step": 9780
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.630053580126644e-07,
+      "loss": 0.0018,
+      "step": 9781
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.6298100340964443e-07,
+      "loss": 0.0004,
+      "step": 9782
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.6295664880662443e-07,
+      "loss": 0.0003,
+      "step": 9783
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.629322942036045e-07,
+      "loss": 0.0005,
+      "step": 9784
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.629079396005845e-07,
+      "loss": 0.0005,
+      "step": 9785
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.6288358499756453e-07,
+      "loss": 0.0006,
+      "step": 9786
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.6285923039454453e-07,
+      "loss": 0.0004,
+      "step": 9787
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.628348757915246e-07,
+      "loss": 0.0004,
+      "step": 9788
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.6281052118850463e-07,
+      "loss": 0.0011,
+      "step": 9789
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 2.6278616658548463e-07,
+      "loss": 0.0003,
+      "step": 9790
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.627618119824647e-07,
+      "loss": 0.0006,
+      "step": 9791
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.627374573794447e-07,
+      "loss": 0.0004,
+      "step": 9792
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.6271310277642473e-07,
+      "loss": 0.0005,
+      "step": 9793
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.6268874817340473e-07,
+      "loss": 0.0004,
+      "step": 9794
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.626643935703848e-07,
+      "loss": 0.0004,
+      "step": 9795
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.626400389673648e-07,
+      "loss": 0.0003,
+      "step": 9796
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.6261568436434484e-07,
+      "loss": 0.0004,
+      "step": 9797
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.625913297613249e-07,
+      "loss": 0.0003,
+      "step": 9798
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.625669751583049e-07,
+      "loss": 0.0005,
+      "step": 9799
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.6254262055528494e-07,
+      "loss": 0.0006,
+      "step": 9800
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 2.6251826595226494e-07,
+      "loss": 0.001,
+      "step": 9801
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.62493911349245e-07,
+      "loss": 0.0003,
+      "step": 9802
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.62469556746225e-07,
+      "loss": 0.0005,
+      "step": 9803
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.6244520214320504e-07,
+      "loss": 0.0005,
+      "step": 9804
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.624208475401851e-07,
+      "loss": 0.0005,
+      "step": 9805
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.623964929371651e-07,
+      "loss": 0.0004,
+      "step": 9806
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.6237213833414514e-07,
+      "loss": 0.0005,
+      "step": 9807
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.6234778373112514e-07,
+      "loss": 0.0004,
+      "step": 9808
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.623234291281052e-07,
+      "loss": 0.0004,
+      "step": 9809
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.622990745250852e-07,
+      "loss": 0.0003,
+      "step": 9810
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 2.6227471992206524e-07,
+      "loss": 0.0005,
+      "step": 9811
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.6225036531904524e-07,
+      "loss": 0.0003,
+      "step": 9812
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.622260107160253e-07,
+      "loss": 0.0003,
+      "step": 9813
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.6220165611300534e-07,
+      "loss": 0.0003,
+      "step": 9814
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.6217730150998534e-07,
+      "loss": 0.0003,
+      "step": 9815
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.621529469069654e-07,
+      "loss": 0.0006,
+      "step": 9816
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.621285923039454e-07,
+      "loss": 0.0004,
+      "step": 9817
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.621042377009255e-07,
+      "loss": 0.0004,
+      "step": 9818
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.620798830979055e-07,
+      "loss": 0.0004,
+      "step": 9819
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.6205552849488555e-07,
+      "loss": 0.0004,
+      "step": 9820
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 2.620311738918656e-07,
+      "loss": 0.0006,
+      "step": 9821
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.620068192888456e-07,
+      "loss": 0.0005,
+      "step": 9822
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.6198246468582565e-07,
+      "loss": 0.0003,
+      "step": 9823
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.6195811008280565e-07,
+      "loss": 0.0003,
+      "step": 9824
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.619337554797857e-07,
+      "loss": 0.0004,
+      "step": 9825
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.619094008767657e-07,
+      "loss": 0.0003,
+      "step": 9826
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.6188504627374575e-07,
+      "loss": 0.0006,
+      "step": 9827
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.618606916707258e-07,
+      "loss": 0.0006,
+      "step": 9828
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.618363370677058e-07,
+      "loss": 0.0004,
+      "step": 9829
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.6181198246468585e-07,
+      "loss": 0.0006,
+      "step": 9830
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.6178762786166585e-07,
+      "loss": 0.0003,
+      "step": 9831
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 2.617632732586459e-07,
+      "loss": 0.0004,
+      "step": 9832
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.617389186556259e-07,
+      "loss": 0.0005,
+      "step": 9833
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.6171456405260595e-07,
+      "loss": 0.0005,
+      "step": 9834
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.6169020944958595e-07,
+      "loss": 0.0002,
+      "step": 9835
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.61665854846566e-07,
+      "loss": 0.0005,
+      "step": 9836
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.6164150024354605e-07,
+      "loss": 0.0006,
+      "step": 9837
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.6161714564052605e-07,
+      "loss": 0.0004,
+      "step": 9838
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.615927910375061e-07,
+      "loss": 0.0004,
+      "step": 9839
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.615684364344861e-07,
+      "loss": 0.0004,
+      "step": 9840
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.6154408183146615e-07,
+      "loss": 0.0004,
+      "step": 9841
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 2.6151972722844615e-07,
+      "loss": 0.0004,
+      "step": 9842
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.614953726254262e-07,
+      "loss": 0.0003,
+      "step": 9843
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.6147101802240626e-07,
+      "loss": 0.0004,
+      "step": 9844
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.6144666341938626e-07,
+      "loss": 0.0004,
+      "step": 9845
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.614223088163663e-07,
+      "loss": 0.0003,
+      "step": 9846
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.613979542133463e-07,
+      "loss": 0.0008,
+      "step": 9847
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.6137359961032636e-07,
+      "loss": 0.0004,
+      "step": 9848
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.6134924500730636e-07,
+      "loss": 0.0004,
+      "step": 9849
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.613248904042864e-07,
+      "loss": 0.0003,
+      "step": 9850
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.613005358012664e-07,
+      "loss": 0.0084,
+      "step": 9851
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 2.6127618119824646e-07,
+      "loss": 0.0005,
+      "step": 9852
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.612518265952265e-07,
+      "loss": 0.0005,
+      "step": 9853
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.612274719922065e-07,
+      "loss": 0.0005,
+      "step": 9854
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.6120311738918656e-07,
+      "loss": 0.0004,
+      "step": 9855
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.6117876278616656e-07,
+      "loss": 0.0005,
+      "step": 9856
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.611544081831466e-07,
+      "loss": 0.0006,
+      "step": 9857
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.611300535801266e-07,
+      "loss": 0.0005,
+      "step": 9858
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.6110569897710666e-07,
+      "loss": 0.0005,
+      "step": 9859
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.610813443740867e-07,
+      "loss": 0.0004,
+      "step": 9860
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.610569897710667e-07,
+      "loss": 0.0004,
+      "step": 9861
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 2.6103263516804676e-07,
+      "loss": 0.0004,
+      "step": 9862
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.6100828056502676e-07,
+      "loss": 0.0011,
+      "step": 9863
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.609839259620068e-07,
+      "loss": 0.0006,
+      "step": 9864
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.609595713589868e-07,
+      "loss": 0.0006,
+      "step": 9865
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.6093521675596686e-07,
+      "loss": 0.0004,
+      "step": 9866
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.6091086215294686e-07,
+      "loss": 0.0003,
+      "step": 9867
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.608865075499269e-07,
+      "loss": 0.0004,
+      "step": 9868
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.6086215294690697e-07,
+      "loss": 0.0005,
+      "step": 9869
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.6083779834388696e-07,
+      "loss": 2.3364,
+      "step": 9870
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.60813443740867e-07,
+      "loss": 0.0006,
+      "step": 9871
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.60789089137847e-07,
+      "loss": 0.0004,
+      "step": 9872
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 2.6076473453482707e-07,
+      "loss": 0.0004,
+      "step": 9873
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.6074037993180706e-07,
+      "loss": 0.0004,
+      "step": 9874
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.607160253287871e-07,
+      "loss": 0.0003,
+      "step": 9875
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.6069167072576717e-07,
+      "loss": 0.0003,
+      "step": 9876
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.6066731612274717e-07,
+      "loss": 0.0004,
+      "step": 9877
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.606429615197272e-07,
+      "loss": 0.0002,
+      "step": 9878
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.606186069167072e-07,
+      "loss": 0.0005,
+      "step": 9879
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.6059425231368727e-07,
+      "loss": 0.0005,
+      "step": 9880
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.6056989771066727e-07,
+      "loss": 0.0005,
+      "step": 9881
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.605455431076473e-07,
+      "loss": 0.0005,
+      "step": 9882
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 2.605211885046273e-07,
+      "loss": 0.0004,
+      "step": 9883
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.6049683390160737e-07,
+      "loss": 1.0827,
+      "step": 9884
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.604724792985874e-07,
+      "loss": 0.0004,
+      "step": 9885
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.604481246955674e-07,
+      "loss": 0.0004,
+      "step": 9886
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.6042377009254747e-07,
+      "loss": 0.0003,
+      "step": 9887
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.6039941548952747e-07,
+      "loss": 0.0004,
+      "step": 9888
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.603750608865075e-07,
+      "loss": 0.0003,
+      "step": 9889
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.603507062834875e-07,
+      "loss": 0.0004,
+      "step": 9890
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.6032635168046757e-07,
+      "loss": 0.0003,
+      "step": 9891
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.603019970774476e-07,
+      "loss": 0.0004,
+      "step": 9892
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 2.602776424744276e-07,
+      "loss": 0.0006,
+      "step": 9893
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.6025328787140773e-07,
+      "loss": 0.0005,
+      "step": 9894
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.602289332683877e-07,
+      "loss": 0.0003,
+      "step": 9895
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.602045786653678e-07,
+      "loss": 0.7504,
+      "step": 9896
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.601802240623478e-07,
+      "loss": 0.0004,
+      "step": 9897
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.6015586945932783e-07,
+      "loss": 0.0003,
+      "step": 9898
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.601315148563079e-07,
+      "loss": 0.0003,
+      "step": 9899
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.601071602532879e-07,
+      "loss": 0.0005,
+      "step": 9900
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.6008280565026793e-07,
+      "loss": 0.0003,
+      "step": 9901
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.6005845104724793e-07,
+      "loss": 0.0005,
+      "step": 9902
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.60034096444228e-07,
+      "loss": 0.0004,
+      "step": 9903
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 2.60009741841208e-07,
+      "loss": 0.0004,
+      "step": 9904
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.5998538723818803e-07,
+      "loss": 0.0006,
+      "step": 9905
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.5996103263516803e-07,
+      "loss": 0.0005,
+      "step": 9906
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.599366780321481e-07,
+      "loss": 0.0005,
+      "step": 9907
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.5991232342912813e-07,
+      "loss": 0.0003,
+      "step": 9908
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.5988796882610813e-07,
+      "loss": 0.0003,
+      "step": 9909
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.598636142230882e-07,
+      "loss": 0.0004,
+      "step": 9910
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.598392596200682e-07,
+      "loss": 0.0006,
+      "step": 9911
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.5981490501704823e-07,
+      "loss": 0.0005,
+      "step": 9912
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.5979055041402823e-07,
+      "loss": 0.0004,
+      "step": 9913
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 2.597661958110083e-07,
+      "loss": 0.0003,
+      "step": 9914
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.5974184120798833e-07,
+      "loss": 0.0004,
+      "step": 9915
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.5971748660496833e-07,
+      "loss": 0.0003,
+      "step": 9916
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.596931320019484e-07,
+      "loss": 0.0004,
+      "step": 9917
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.596687773989284e-07,
+      "loss": 0.0005,
+      "step": 9918
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.5964442279590844e-07,
+      "loss": 0.0003,
+      "step": 9919
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.5962006819288843e-07,
+      "loss": 0.0005,
+      "step": 9920
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.595957135898685e-07,
+      "loss": 0.0005,
+      "step": 9921
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.595713589868485e-07,
+      "loss": 0.0004,
+      "step": 9922
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.5954700438382854e-07,
+      "loss": 0.0004,
+      "step": 9923
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 2.595226497808086e-07,
+      "loss": 0.0004,
+      "step": 9924
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.594982951777886e-07,
+      "loss": 0.0006,
+      "step": 9925
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.5947394057476864e-07,
+      "loss": 0.0004,
+      "step": 9926
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.5944958597174864e-07,
+      "loss": 0.0005,
+      "step": 9927
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.594252313687287e-07,
+      "loss": 0.0006,
+      "step": 9928
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.594008767657087e-07,
+      "loss": 0.0003,
+      "step": 9929
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.5937652216268874e-07,
+      "loss": 0.0005,
+      "step": 9930
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.593521675596688e-07,
+      "loss": 0.0004,
+      "step": 9931
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.593278129566488e-07,
+      "loss": 0.0003,
+      "step": 9932
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.5930345835362884e-07,
+      "loss": 0.0004,
+      "step": 9933
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 2.5927910375060884e-07,
+      "loss": 0.0004,
+      "step": 9934
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.592547491475889e-07,
+      "loss": 0.0003,
+      "step": 9935
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.592303945445689e-07,
+      "loss": 0.0003,
+      "step": 9936
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.5920603994154894e-07,
+      "loss": 0.0004,
+      "step": 9937
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.5918168533852894e-07,
+      "loss": 0.0003,
+      "step": 9938
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.59157330735509e-07,
+      "loss": 0.0004,
+      "step": 9939
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.5913297613248904e-07,
+      "loss": 0.0004,
+      "step": 9940
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.5910862152946904e-07,
+      "loss": 0.0003,
+      "step": 9941
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.590842669264491e-07,
+      "loss": 0.0003,
+      "step": 9942
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.590599123234291e-07,
+      "loss": 0.0002,
+      "step": 9943
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.5903555772040914e-07,
+      "loss": 0.0004,
+      "step": 9944
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 2.5901120311738914e-07,
+      "loss": 0.0005,
+      "step": 9945
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.589868485143692e-07,
+      "loss": 0.0004,
+      "step": 9946
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.5896249391134925e-07,
+      "loss": 0.0003,
+      "step": 9947
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.5893813930832924e-07,
+      "loss": 0.0003,
+      "step": 9948
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.589137847053093e-07,
+      "loss": 0.0002,
+      "step": 9949
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.588894301022893e-07,
+      "loss": 0.0004,
+      "step": 9950
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.5886507549926935e-07,
+      "loss": 0.0004,
+      "step": 9951
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.5884072089624934e-07,
+      "loss": 0.0005,
+      "step": 9952
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.588163662932294e-07,
+      "loss": 0.0005,
+      "step": 9953
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.5879201169020945e-07,
+      "loss": 0.0005,
+      "step": 9954
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 2.5876765708718945e-07,
+      "loss": 0.0005,
+      "step": 9955
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.587433024841695e-07,
+      "loss": 0.0004,
+      "step": 9956
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.587189478811495e-07,
+      "loss": 0.0005,
+      "step": 9957
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.5869459327812955e-07,
+      "loss": 0.0006,
+      "step": 9958
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.5867023867510955e-07,
+      "loss": 0.0004,
+      "step": 9959
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.586458840720896e-07,
+      "loss": 0.0004,
+      "step": 9960
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.586215294690696e-07,
+      "loss": 0.0003,
+      "step": 9961
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.5859717486604965e-07,
+      "loss": 0.0004,
+      "step": 9962
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.585728202630297e-07,
+      "loss": 0.0005,
+      "step": 9963
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.585484656600097e-07,
+      "loss": 0.0005,
+      "step": 9964
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 2.5852411105698975e-07,
+      "loss": 0.0003,
+      "step": 9965
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.5849975645396975e-07,
+      "loss": 0.0004,
+      "step": 9966
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.584754018509498e-07,
+      "loss": 0.0003,
+      "step": 9967
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.584510472479298e-07,
+      "loss": 0.0003,
+      "step": 9968
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.5842669264490985e-07,
+      "loss": 0.0011,
+      "step": 9969
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.584023380418899e-07,
+      "loss": 0.0003,
+      "step": 9970
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.5837798343886996e-07,
+      "loss": 0.0004,
+      "step": 9971
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.5835362883585e-07,
+      "loss": 0.0003,
+      "step": 9972
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.5832927423283e-07,
+      "loss": 0.0009,
+      "step": 9973
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.5830491962981006e-07,
+      "loss": 0.0005,
+      "step": 9974
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.5828056502679006e-07,
+      "loss": 0.0004,
+      "step": 9975
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 2.582562104237701e-07,
+      "loss": 0.0003,
+      "step": 9976
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.5823185582075016e-07,
+      "loss": 0.0004,
+      "step": 9977
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.5820750121773016e-07,
+      "loss": 0.0003,
+      "step": 9978
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.581831466147102e-07,
+      "loss": 0.0003,
+      "step": 9979
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.581587920116902e-07,
+      "loss": 0.0003,
+      "step": 9980
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.5813443740867026e-07,
+      "loss": 0.0004,
+      "step": 9981
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.5811008280565026e-07,
+      "loss": 0.0004,
+      "step": 9982
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.580857282026303e-07,
+      "loss": 0.0003,
+      "step": 9983
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.580613735996103e-07,
+      "loss": 0.0004,
+      "step": 9984
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.5803701899659036e-07,
+      "loss": 0.0004,
+      "step": 9985
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 2.580126643935704e-07,
+      "loss": 0.0003,
+      "step": 9986
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.579883097905504e-07,
+      "loss": 0.0004,
+      "step": 9987
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.5796395518753046e-07,
+      "loss": 0.0004,
+      "step": 9988
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.5793960058451046e-07,
+      "loss": 0.0004,
+      "step": 9989
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.579152459814905e-07,
+      "loss": 0.0005,
+      "step": 9990
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.578908913784705e-07,
+      "loss": 0.0005,
+      "step": 9991
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.5786653677545056e-07,
+      "loss": 0.0005,
+      "step": 9992
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.578421821724306e-07,
+      "loss": 0.0007,
+      "step": 9993
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.578178275694106e-07,
+      "loss": 0.0004,
+      "step": 9994
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.5779347296639067e-07,
+      "loss": 0.0006,
+      "step": 9995
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 2.5776911836337066e-07,
+      "loss": 0.0003,
+      "step": 9996
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.577447637603507e-07,
+      "loss": 0.0004,
+      "step": 9997
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.577204091573307e-07,
+      "loss": 0.0003,
+      "step": 9998
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.5769605455431077e-07,
+      "loss": 0.0003,
+      "step": 9999
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.5767169995129076e-07,
+      "loss": 0.0004,
+      "step": 10000
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.576473453482708e-07,
+      "loss": 0.0004,
+      "step": 10001
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.5762299074525087e-07,
+      "loss": 0.0005,
+      "step": 10002
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.5759863614223087e-07,
+      "loss": 0.0004,
+      "step": 10003
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.575742815392109e-07,
+      "loss": 0.0004,
+      "step": 10004
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.575499269361909e-07,
+      "loss": 0.0004,
+      "step": 10005
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.5752557233317097e-07,
+      "loss": 0.1651,
+      "step": 10006
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 2.5750121773015097e-07,
+      "loss": 0.0003,
+      "step": 10007
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.57476863127131e-07,
+      "loss": 0.0003,
+      "step": 10008
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.5745250852411107e-07,
+      "loss": 0.0004,
+      "step": 10009
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.5742815392109107e-07,
+      "loss": 0.0003,
+      "step": 10010
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.574037993180711e-07,
+      "loss": 0.0003,
+      "step": 10011
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.573794447150511e-07,
+      "loss": 0.0003,
+      "step": 10012
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.5735509011203117e-07,
+      "loss": 0.0007,
+      "step": 10013
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.5733073550901117e-07,
+      "loss": 0.0004,
+      "step": 10014
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.573063809059912e-07,
+      "loss": 0.0004,
+      "step": 10015
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.572820263029712e-07,
+      "loss": 0.0003,
+      "step": 10016
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 2.5725767169995127e-07,
+      "loss": 0.0006,
+      "step": 10017
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.572333170969313e-07,
+      "loss": 0.0004,
+      "step": 10018
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.572089624939113e-07,
+      "loss": 0.0003,
+      "step": 10019
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.571846078908914e-07,
+      "loss": 0.0004,
+      "step": 10020
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.5716025328787137e-07,
+      "loss": 0.0003,
+      "step": 10021
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.571358986848514e-07,
+      "loss": 0.0004,
+      "step": 10022
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.571115440818314e-07,
+      "loss": 0.0004,
+      "step": 10023
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.570871894788115e-07,
+      "loss": 0.0006,
+      "step": 10024
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.570628348757915e-07,
+      "loss": 0.0003,
+      "step": 10025
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.570384802727715e-07,
+      "loss": 0.0004,
+      "step": 10026
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 2.570141256697516e-07,
+      "loss": 0.0004,
+      "step": 10027
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.569897710667316e-07,
+      "loss": 0.0005,
+      "step": 10028
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.5696541646371163e-07,
+      "loss": 0.0004,
+      "step": 10029
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.569410618606916e-07,
+      "loss": 0.0005,
+      "step": 10030
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.569167072576717e-07,
+      "loss": 0.0004,
+      "step": 10031
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.568923526546517e-07,
+      "loss": 0.6099,
+      "step": 10032
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.5686799805163173e-07,
+      "loss": 0.0003,
+      "step": 10033
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.568436434486118e-07,
+      "loss": 0.0004,
+      "step": 10034
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.568192888455918e-07,
+      "loss": 0.0002,
+      "step": 10035
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.5679493424257183e-07,
+      "loss": 0.0003,
+      "step": 10036
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 2.5677057963955183e-07,
+      "loss": 0.0005,
+      "step": 10037
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.567462250365319e-07,
+      "loss": 0.0005,
+      "step": 10038
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.567218704335119e-07,
+      "loss": 0.0004,
+      "step": 10039
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.5669751583049193e-07,
+      "loss": 0.0005,
+      "step": 10040
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.56673161227472e-07,
+      "loss": 0.0004,
+      "step": 10041
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.56648806624452e-07,
+      "loss": 0.0012,
+      "step": 10042
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.5662445202143203e-07,
+      "loss": 0.0004,
+      "step": 10043
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.5660009741841203e-07,
+      "loss": 0.0005,
+      "step": 10044
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.565757428153921e-07,
+      "loss": 0.0005,
+      "step": 10045
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.565513882123721e-07,
+      "loss": 0.0005,
+      "step": 10046
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.565270336093522e-07,
+      "loss": 0.0003,
+      "step": 10047
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 2.5650267900633224e-07,
+      "loss": 0.0003,
+      "step": 10048
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.5647832440331224e-07,
+      "loss": 0.0004,
+      "step": 10049
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.564539698002923e-07,
+      "loss": 0.0003,
+      "step": 10050
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.564296151972723e-07,
+      "loss": 0.0003,
+      "step": 10051
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.5640526059425234e-07,
+      "loss": 0.0005,
+      "step": 10052
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.5638090599123234e-07,
+      "loss": 0.0005,
+      "step": 10053
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.563565513882124e-07,
+      "loss": 0.0003,
+      "step": 10054
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.563321967851924e-07,
+      "loss": 0.0003,
+      "step": 10055
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.5630784218217244e-07,
+      "loss": 0.0005,
+      "step": 10056
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.562834875791525e-07,
+      "loss": 0.1515,
+      "step": 10057
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 2.562591329761325e-07,
+      "loss": 0.0004,
+      "step": 10058
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.5623477837311254e-07,
+      "loss": 0.0003,
+      "step": 10059
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.5621042377009254e-07,
+      "loss": 0.0004,
+      "step": 10060
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.561860691670726e-07,
+      "loss": 0.0004,
+      "step": 10061
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.561617145640526e-07,
+      "loss": 0.0033,
+      "step": 10062
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.5613735996103264e-07,
+      "loss": 0.0002,
+      "step": 10063
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.561130053580127e-07,
+      "loss": 0.0004,
+      "step": 10064
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.560886507549927e-07,
+      "loss": 0.0004,
+      "step": 10065
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.5606429615197274e-07,
+      "loss": 0.0003,
+      "step": 10066
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.5603994154895274e-07,
+      "loss": 0.0003,
+      "step": 10067
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 2.560155869459328e-07,
+      "loss": 0.0004,
+      "step": 10068
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.559912323429128e-07,
+      "loss": 0.0004,
+      "step": 10069
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.5596687773989284e-07,
+      "loss": 0.0713,
+      "step": 10070
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.5594252313687284e-07,
+      "loss": 2.6342,
+      "step": 10071
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.559181685338529e-07,
+      "loss": 0.0004,
+      "step": 10072
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.5589381393083295e-07,
+      "loss": 0.0005,
+      "step": 10073
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.5586945932781294e-07,
+      "loss": 0.0005,
+      "step": 10074
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.55845104724793e-07,
+      "loss": 0.0004,
+      "step": 10075
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.55820750121773e-07,
+      "loss": 0.0004,
+      "step": 10076
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.5579639551875305e-07,
+      "loss": 0.0005,
+      "step": 10077
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.5577204091573305e-07,
+      "loss": 0.0003,
+      "step": 10078
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 2.557476863127131e-07,
+      "loss": 0.0005,
+      "step": 10079
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.5572333170969315e-07,
+      "loss": 0.0005,
+      "step": 10080
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.5569897710667315e-07,
+      "loss": 0.0003,
+      "step": 10081
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.556746225036532e-07,
+      "loss": 0.0004,
+      "step": 10082
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.556502679006332e-07,
+      "loss": 0.0004,
+      "step": 10083
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.5562591329761325e-07,
+      "loss": 0.0028,
+      "step": 10084
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.5560155869459325e-07,
+      "loss": 0.0005,
+      "step": 10085
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.555772040915733e-07,
+      "loss": 0.0005,
+      "step": 10086
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.555528494885533e-07,
+      "loss": 0.0007,
+      "step": 10087
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.5552849488553335e-07,
+      "loss": 0.0005,
+      "step": 10088
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 2.555041402825134e-07,
+      "loss": 0.0005,
+      "step": 10089
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.554797856794934e-07,
+      "loss": 0.0004,
+      "step": 10090
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.5545543107647345e-07,
+      "loss": 0.0012,
+      "step": 10091
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.5543107647345345e-07,
+      "loss": 0.0003,
+      "step": 10092
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.554067218704335e-07,
+      "loss": 0.0003,
+      "step": 10093
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.553823672674135e-07,
+      "loss": 0.0004,
+      "step": 10094
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.5535801266439355e-07,
+      "loss": 0.0003,
+      "step": 10095
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.553336580613736e-07,
+      "loss": 0.0005,
+      "step": 10096
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.553093034583536e-07,
+      "loss": 0.0005,
+      "step": 10097
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.5528494885533365e-07,
+      "loss": 0.0003,
+      "step": 10098
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 2.5526059425231365e-07,
+      "loss": 0.0006,
+      "step": 10099
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.552362396492937e-07,
+      "loss": 0.0005,
+      "step": 10100
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.552118850462737e-07,
+      "loss": 0.0005,
+      "step": 10101
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.5518753044325375e-07,
+      "loss": 0.0004,
+      "step": 10102
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.5516317584023375e-07,
+      "loss": 0.0003,
+      "step": 10103
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.551388212372138e-07,
+      "loss": 0.0002,
+      "step": 10104
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.5511446663419386e-07,
+      "loss": 0.0003,
+      "step": 10105
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.5509011203117386e-07,
+      "loss": 0.0004,
+      "step": 10106
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.550657574281539e-07,
+      "loss": 0.0003,
+      "step": 10107
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.550414028251339e-07,
+      "loss": 0.0004,
+      "step": 10108
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 2.5501704822211396e-07,
+      "loss": 0.0004,
+      "step": 10109
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.5499269361909396e-07,
+      "loss": 0.0003,
+      "step": 10110
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.54968339016074e-07,
+      "loss": 0.0048,
+      "step": 10111
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.5494398441305406e-07,
+      "loss": 0.0004,
+      "step": 10112
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.5491962981003406e-07,
+      "loss": 0.0004,
+      "step": 10113
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.548952752070141e-07,
+      "loss": 0.0003,
+      "step": 10114
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.548709206039941e-07,
+      "loss": 0.0005,
+      "step": 10115
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.5484656600097416e-07,
+      "loss": 0.0004,
+      "step": 10116
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.5482221139795416e-07,
+      "loss": 0.0003,
+      "step": 10117
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.547978567949342e-07,
+      "loss": 0.0006,
+      "step": 10118
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.5477350219191426e-07,
+      "loss": 0.0003,
+      "step": 10119
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 2.5474914758889426e-07,
+      "loss": 0.0006,
+      "step": 10120
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.547247929858743e-07,
+      "loss": 0.0003,
+      "step": 10121
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.547004383828543e-07,
+      "loss": 0.0004,
+      "step": 10122
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.5467608377983436e-07,
+      "loss": 0.0005,
+      "step": 10123
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.546517291768144e-07,
+      "loss": 0.0003,
+      "step": 10124
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.5462737457379447e-07,
+      "loss": 0.0007,
+      "step": 10125
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.5460301997077446e-07,
+      "loss": 0.0006,
+      "step": 10126
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.545786653677545e-07,
+      "loss": 0.0003,
+      "step": 10127
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.5455431076473457e-07,
+      "loss": 0.0004,
+      "step": 10128
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.5452995616171457e-07,
+      "loss": 0.0004,
+      "step": 10129
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 2.545056015586946e-07,
+      "loss": 0.0003,
+      "step": 10130
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.544812469556746e-07,
+      "loss": 0.0003,
+      "step": 10131
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.5445689235265467e-07,
+      "loss": 0.0003,
+      "step": 10132
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.5443253774963467e-07,
+      "loss": 0.0003,
+      "step": 10133
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.544081831466147e-07,
+      "loss": 0.0004,
+      "step": 10134
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.5438382854359477e-07,
+      "loss": 0.0004,
+      "step": 10135
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.5435947394057477e-07,
+      "loss": 0.0004,
+      "step": 10136
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.543351193375548e-07,
+      "loss": 0.0005,
+      "step": 10137
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.543107647345348e-07,
+      "loss": 0.0004,
+      "step": 10138
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.5428641013151487e-07,
+      "loss": 0.0003,
+      "step": 10139
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 2.5426205552849487e-07,
+      "loss": 0.0004,
+      "step": 10140
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.542377009254749e-07,
+      "loss": 0.0004,
+      "step": 10141
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.542133463224549e-07,
+      "loss": 0.0007,
+      "step": 10142
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.5418899171943497e-07,
+      "loss": 0.0004,
+      "step": 10143
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.54164637116415e-07,
+      "loss": 0.0004,
+      "step": 10144
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.54140282513395e-07,
+      "loss": 0.0434,
+      "step": 10145
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.541159279103751e-07,
+      "loss": 0.0006,
+      "step": 10146
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.5409157330735507e-07,
+      "loss": 0.0004,
+      "step": 10147
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.540672187043351e-07,
+      "loss": 0.0004,
+      "step": 10148
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.540428641013151e-07,
+      "loss": 0.0003,
+      "step": 10149
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.540185094982952e-07,
+      "loss": 0.0004,
+      "step": 10150
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 2.539941548952752e-07,
+      "loss": 0.0003,
+      "step": 10151
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.539698002922552e-07,
+      "loss": 0.0005,
+      "step": 10152
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.539454456892353e-07,
+      "loss": 0.0007,
+      "step": 10153
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.539210910862153e-07,
+      "loss": 0.0006,
+      "step": 10154
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.5389673648319533e-07,
+      "loss": 0.0003,
+      "step": 10155
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.538723818801753e-07,
+      "loss": 0.0003,
+      "step": 10156
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.538480272771554e-07,
+      "loss": 0.0005,
+      "step": 10157
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.5382367267413543e-07,
+      "loss": 0.0004,
+      "step": 10158
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.5379931807111543e-07,
+      "loss": 0.0004,
+      "step": 10159
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.537749634680955e-07,
+      "loss": 0.0006,
+      "step": 10160
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 2.537506088650755e-07,
+      "loss": 0.0004,
+      "step": 10161
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.5372625426205553e-07,
+      "loss": 0.0006,
+      "step": 10162
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.5370189965903553e-07,
+      "loss": 0.0004,
+      "step": 10163
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.536775450560156e-07,
+      "loss": 0.0003,
+      "step": 10164
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.536531904529956e-07,
+      "loss": 0.0003,
+      "step": 10165
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.5362883584997563e-07,
+      "loss": 0.0005,
+      "step": 10166
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.536044812469557e-07,
+      "loss": 0.0004,
+      "step": 10167
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.535801266439357e-07,
+      "loss": 0.0008,
+      "step": 10168
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.5355577204091573e-07,
+      "loss": 0.0003,
+      "step": 10169
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.5353141743789573e-07,
+      "loss": 0.0004,
+      "step": 10170
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 2.535070628348758e-07,
+      "loss": 0.0007,
+      "step": 10171
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.534827082318558e-07,
+      "loss": 0.0003,
+      "step": 10172
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.5345835362883583e-07,
+      "loss": 0.0003,
+      "step": 10173
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.534339990258159e-07,
+      "loss": 0.0004,
+      "step": 10174
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.534096444227959e-07,
+      "loss": 0.0003,
+      "step": 10175
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.5338528981977593e-07,
+      "loss": 0.0004,
+      "step": 10176
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.5336093521675593e-07,
+      "loss": 0.0005,
+      "step": 10177
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.53336580613736e-07,
+      "loss": 0.0003,
+      "step": 10178
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.53312226010716e-07,
+      "loss": 0.0004,
+      "step": 10179
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.5328787140769604e-07,
+      "loss": 0.0003,
+      "step": 10180
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 2.5326351680467603e-07,
+      "loss": 0.0004,
+      "step": 10181
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.532391622016561e-07,
+      "loss": 0.0004,
+      "step": 10182
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.5321480759863614e-07,
+      "loss": 0.0003,
+      "step": 10183
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.5319045299561614e-07,
+      "loss": 0.0003,
+      "step": 10184
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.531660983925962e-07,
+      "loss": 0.0005,
+      "step": 10185
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.531417437895762e-07,
+      "loss": 0.0003,
+      "step": 10186
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.5311738918655624e-07,
+      "loss": 0.0005,
+      "step": 10187
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.5309303458353624e-07,
+      "loss": 0.0004,
+      "step": 10188
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.530686799805163e-07,
+      "loss": 0.0003,
+      "step": 10189
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.5304432537749634e-07,
+      "loss": 0.0003,
+      "step": 10190
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.5301997077447634e-07,
+      "loss": 0.0004,
+      "step": 10191
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 2.529956161714564e-07,
+      "loss": 0.0003,
+      "step": 10192
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.529712615684364e-07,
+      "loss": 0.0003,
+      "step": 10193
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5294690696541644e-07,
+      "loss": 0.0003,
+      "step": 10194
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5292255236239644e-07,
+      "loss": 0.0005,
+      "step": 10195
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.528981977593765e-07,
+      "loss": 0.0004,
+      "step": 10196
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.528738431563565e-07,
+      "loss": 0.0004,
+      "step": 10197
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.5284948855333654e-07,
+      "loss": 0.0007,
+      "step": 10198
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.528251339503166e-07,
+      "loss": 0.0004,
+      "step": 10199
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.528007793472966e-07,
+      "loss": 0.0003,
+      "step": 10200
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.527764247442767e-07,
+      "loss": 0.0004,
+      "step": 10201
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 2.527520701412567e-07,
+      "loss": 0.0005,
+      "step": 10202
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5272771553823675e-07,
+      "loss": 0.0003,
+      "step": 10203
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5270336093521675e-07,
+      "loss": 0.0004,
+      "step": 10204
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.526790063321968e-07,
+      "loss": 0.0005,
+      "step": 10205
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5265465172917685e-07,
+      "loss": 0.0004,
+      "step": 10206
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5263029712615685e-07,
+      "loss": 0.0005,
+      "step": 10207
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.526059425231369e-07,
+      "loss": 0.0003,
+      "step": 10208
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.525815879201169e-07,
+      "loss": 0.0003,
+      "step": 10209
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5255723331709695e-07,
+      "loss": 0.0005,
+      "step": 10210
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.5253287871407695e-07,
+      "loss": 0.0006,
+      "step": 10211
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 2.52508524111057e-07,
+      "loss": 0.0004,
+      "step": 10212
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.5248416950803705e-07,
+      "loss": 0.0004,
+      "step": 10213
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.5245981490501705e-07,
+      "loss": 0.0003,
+      "step": 10214
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.524354603019971e-07,
+      "loss": 0.0003,
+      "step": 10215
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.524111056989771e-07,
+      "loss": 0.0004,
+      "step": 10216
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.5238675109595715e-07,
+      "loss": 0.0003,
+      "step": 10217
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.5236239649293715e-07,
+      "loss": 0.0004,
+      "step": 10218
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.523380418899172e-07,
+      "loss": 0.0005,
+      "step": 10219
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.523136872868972e-07,
+      "loss": 0.0004,
+      "step": 10220
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.5228933268387725e-07,
+      "loss": 0.0005,
+      "step": 10221
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.522649780808573e-07,
+      "loss": 0.0004,
+      "step": 10222
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 2.522406234778373e-07,
+      "loss": 0.0004,
+      "step": 10223
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.5221626887481735e-07,
+      "loss": 0.0003,
+      "step": 10224
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.5219191427179735e-07,
+      "loss": 0.0004,
+      "step": 10225
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.521675596687774e-07,
+      "loss": 0.0003,
+      "step": 10226
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.521432050657574e-07,
+      "loss": 0.0003,
+      "step": 10227
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.5211885046273746e-07,
+      "loss": 0.0003,
+      "step": 10228
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.520944958597175e-07,
+      "loss": 0.0004,
+      "step": 10229
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.520701412566975e-07,
+      "loss": 0.0003,
+      "step": 10230
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.5204578665367756e-07,
+      "loss": 0.0003,
+      "step": 10231
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.5202143205065756e-07,
+      "loss": 0.0005,
+      "step": 10232
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 2.519970774476376e-07,
+      "loss": 0.0004,
+      "step": 10233
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.519727228446176e-07,
+      "loss": 0.0004,
+      "step": 10234
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.5194836824159766e-07,
+      "loss": 0.0004,
+      "step": 10235
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.5192401363857766e-07,
+      "loss": 0.0004,
+      "step": 10236
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.518996590355577e-07,
+      "loss": 0.0005,
+      "step": 10237
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.5187530443253776e-07,
+      "loss": 0.0003,
+      "step": 10238
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.5185094982951776e-07,
+      "loss": 0.0004,
+      "step": 10239
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.518265952264978e-07,
+      "loss": 0.0004,
+      "step": 10240
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.518022406234778e-07,
+      "loss": 0.0003,
+      "step": 10241
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.5177788602045786e-07,
+      "loss": 0.0005,
+      "step": 10242
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 2.5175353141743786e-07,
+      "loss": 0.0003,
+      "step": 10243
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.517291768144179e-07,
+      "loss": 0.0003,
+      "step": 10244
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.5170482221139796e-07,
+      "loss": 0.0002,
+      "step": 10245
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.5168046760837796e-07,
+      "loss": 0.0003,
+      "step": 10246
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.51656113005358e-07,
+      "loss": 0.0004,
+      "step": 10247
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.51631758402338e-07,
+      "loss": 0.0003,
+      "step": 10248
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.5160740379931806e-07,
+      "loss": 0.0003,
+      "step": 10249
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.5158304919629806e-07,
+      "loss": 0.0004,
+      "step": 10250
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.515586945932781e-07,
+      "loss": 0.0002,
+      "step": 10251
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.515343399902581e-07,
+      "loss": 0.0003,
+      "step": 10252
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 2.5150998538723816e-07,
+      "loss": 0.0005,
+      "step": 10253
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.514856307842182e-07,
+      "loss": 0.9749,
+      "step": 10254
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.514612761811982e-07,
+      "loss": 0.0003,
+      "step": 10255
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.5143692157817827e-07,
+      "loss": 0.0004,
+      "step": 10256
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.5141256697515826e-07,
+      "loss": 0.0002,
+      "step": 10257
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.513882123721383e-07,
+      "loss": 0.0004,
+      "step": 10258
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.513638577691183e-07,
+      "loss": 0.0004,
+      "step": 10259
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.5133950316609837e-07,
+      "loss": 0.0003,
+      "step": 10260
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.513151485630784e-07,
+      "loss": 0.0003,
+      "step": 10261
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.512907939600584e-07,
+      "loss": 0.0004,
+      "step": 10262
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.5126643935703847e-07,
+      "loss": 0.0004,
+      "step": 10263
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 2.5124208475401847e-07,
+      "loss": 0.0002,
+      "step": 10264
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.512177301509985e-07,
+      "loss": 0.0003,
+      "step": 10265
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.511933755479785e-07,
+      "loss": 0.0004,
+      "step": 10266
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.5116902094495857e-07,
+      "loss": 0.0005,
+      "step": 10267
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.5114466634193857e-07,
+      "loss": 0.0005,
+      "step": 10268
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.511203117389186e-07,
+      "loss": 0.0004,
+      "step": 10269
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.5109595713589867e-07,
+      "loss": 0.0005,
+      "step": 10270
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.5107160253287867e-07,
+      "loss": 0.0004,
+      "step": 10271
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.510472479298587e-07,
+      "loss": 0.0003,
+      "step": 10272
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.510228933268387e-07,
+      "loss": 0.0003,
+      "step": 10273
+    },
+    {
+      "epoch": 9.98,
+      "learning_rate": 2.5099853872381877e-07,
+      "loss": 0.0004,
+      "step": 10274
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.5097418412079877e-07,
+      "loss": 0.0005,
+      "step": 10275
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.509498295177788e-07,
+      "loss": 0.0005,
+      "step": 10276
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.5092547491475893e-07,
+      "loss": 0.0005,
+      "step": 10277
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.509011203117389e-07,
+      "loss": 0.0005,
+      "step": 10278
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.50876765708719e-07,
+      "loss": 0.0003,
+      "step": 10279
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.50852411105699e-07,
+      "loss": 0.0004,
+      "step": 10280
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.5082805650267903e-07,
+      "loss": 0.0003,
+      "step": 10281
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.50803701899659e-07,
+      "loss": 0.0005,
+      "step": 10282
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.507793472966391e-07,
+      "loss": 0.0004,
+      "step": 10283
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 2.5075499269361913e-07,
+      "loss": 0.0004,
+      "step": 10284
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.5073063809059913e-07,
+      "loss": 0.0004,
+      "step": 10285
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.507062834875792e-07,
+      "loss": 0.0005,
+      "step": 10286
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.506819288845592e-07,
+      "loss": 0.0004,
+      "step": 10287
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.5065757428153923e-07,
+      "loss": 0.0003,
+      "step": 10288
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.5063321967851923e-07,
+      "loss": 0.0003,
+      "step": 10289
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.506088650754993e-07,
+      "loss": 0.0005,
+      "step": 10290
+    },
+    {
+      "epoch": 10.0,
+      "eval_accuracy": 0.9985415653864852,
+      "eval_loss": 0.01590646617114544,
+      "eval_runtime": 2.8019,
+      "eval_samples_per_second": 734.139,
+      "eval_steps_per_second": 11.778,
+      "step": 10290
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.505845104724793e-07,
+      "loss": 0.0004,
+      "step": 10291
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.5056015586945933e-07,
+      "loss": 0.0003,
+      "step": 10292
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.505358012664394e-07,
+      "loss": 0.0004,
+      "step": 10293
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.505114466634194e-07,
+      "loss": 0.0005,
+      "step": 10294
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 2.5048709206039943e-07,
+      "loss": 0.0005,
+      "step": 10295
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.5046273745737943e-07,
+      "loss": 0.0004,
+      "step": 10296
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.504383828543595e-07,
+      "loss": 0.0003,
+      "step": 10297
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.504140282513395e-07,
+      "loss": 0.0003,
+      "step": 10298
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.5038967364831953e-07,
+      "loss": 0.0004,
+      "step": 10299
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.503653190452996e-07,
+      "loss": 0.0006,
+      "step": 10300
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.503409644422796e-07,
+      "loss": 0.0003,
+      "step": 10301
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.5031660983925964e-07,
+      "loss": 0.0003,
+      "step": 10302
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.5029225523623963e-07,
+      "loss": 0.0003,
+      "step": 10303
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.502679006332197e-07,
+      "loss": 0.0003,
+      "step": 10304
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 2.502435460301997e-07,
+      "loss": 0.0004,
+      "step": 10305
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.5021919142717974e-07,
+      "loss": 0.0003,
+      "step": 10306
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.5019483682415973e-07,
+      "loss": 0.0006,
+      "step": 10307
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.501704822211398e-07,
+      "loss": 0.0003,
+      "step": 10308
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.5014612761811984e-07,
+      "loss": 0.0005,
+      "step": 10309
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.5012177301509984e-07,
+      "loss": 0.0004,
+      "step": 10310
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.500974184120799e-07,
+      "loss": 0.0008,
+      "step": 10311
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.500730638090599e-07,
+      "loss": 0.0005,
+      "step": 10312
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.5004870920603994e-07,
+      "loss": 0.0005,
+      "step": 10313
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.5002435460301994e-07,
+      "loss": 0.0004,
+      "step": 10314
+    },
+    {
+      "epoch": 10.02,
+      "learning_rate": 2.5e-07,
+      "loss": 0.0005,
+      "step": 10315
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4997564539698004e-07,
+      "loss": 0.0004,
+      "step": 10316
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4995129079396004e-07,
+      "loss": 0.0006,
+      "step": 10317
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.499269361909401e-07,
+      "loss": 0.0004,
+      "step": 10318
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.499025815879201e-07,
+      "loss": 0.0005,
+      "step": 10319
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4987822698490014e-07,
+      "loss": 0.0003,
+      "step": 10320
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4985387238188014e-07,
+      "loss": 0.0002,
+      "step": 10321
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.498295177788602e-07,
+      "loss": 0.0003,
+      "step": 10322
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4980516317584024e-07,
+      "loss": 0.0003,
+      "step": 10323
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.4978080857282024e-07,
+      "loss": 0.0004,
+      "step": 10324
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.497564539698003e-07,
+      "loss": 0.0003,
+      "step": 10325
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 2.497320993667803e-07,
+      "loss": 0.0005,
+      "step": 10326
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.4970774476376034e-07,
+      "loss": 0.0002,
+      "step": 10327
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.4968339016074034e-07,
+      "loss": 0.0004,
+      "step": 10328
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.496590355577204e-07,
+      "loss": 0.0004,
+      "step": 10329
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.496346809547004e-07,
+      "loss": 0.0005,
+      "step": 10330
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.4961032635168044e-07,
+      "loss": 0.0004,
+      "step": 10331
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.495859717486605e-07,
+      "loss": 0.0009,
+      "step": 10332
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.495616171456405e-07,
+      "loss": 0.0005,
+      "step": 10333
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.4953726254262055e-07,
+      "loss": 0.0003,
+      "step": 10334
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.495129079396006e-07,
+      "loss": 0.0003,
+      "step": 10335
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 2.494885533365806e-07,
+      "loss": 0.0002,
+      "step": 10336
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4946419873356065e-07,
+      "loss": 0.0005,
+      "step": 10337
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4943984413054065e-07,
+      "loss": 0.0004,
+      "step": 10338
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.494154895275207e-07,
+      "loss": 0.0003,
+      "step": 10339
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4939113492450075e-07,
+      "loss": 0.0005,
+      "step": 10340
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4936678032148075e-07,
+      "loss": 0.0003,
+      "step": 10341
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.493424257184608e-07,
+      "loss": 0.0004,
+      "step": 10342
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.493180711154408e-07,
+      "loss": 0.0003,
+      "step": 10343
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4929371651242085e-07,
+      "loss": 0.0005,
+      "step": 10344
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.4926936190940085e-07,
+      "loss": 0.0003,
+      "step": 10345
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 2.492450073063809e-07,
+      "loss": 0.0003,
+      "step": 10346
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.4922065270336095e-07,
+      "loss": 0.0003,
+      "step": 10347
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.4919629810034095e-07,
+      "loss": 0.0005,
+      "step": 10348
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.49171943497321e-07,
+      "loss": 0.0003,
+      "step": 10349
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.49147588894301e-07,
+      "loss": 0.1772,
+      "step": 10350
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.4912323429128105e-07,
+      "loss": 0.0005,
+      "step": 10351
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.4909887968826105e-07,
+      "loss": 0.0004,
+      "step": 10352
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.490745250852411e-07,
+      "loss": 0.0004,
+      "step": 10353
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.490501704822211e-07,
+      "loss": 0.0003,
+      "step": 10354
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.4902581587920115e-07,
+      "loss": 0.0003,
+      "step": 10355
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 2.490014612761812e-07,
+      "loss": 0.0004,
+      "step": 10356
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.489771066731612e-07,
+      "loss": 0.0002,
+      "step": 10357
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.4895275207014126e-07,
+      "loss": 0.0005,
+      "step": 10358
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.4892839746712125e-07,
+      "loss": 0.0003,
+      "step": 10359
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.489040428641013e-07,
+      "loss": 0.0003,
+      "step": 10360
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.488796882610813e-07,
+      "loss": 0.0005,
+      "step": 10361
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.4885533365806136e-07,
+      "loss": 0.0005,
+      "step": 10362
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.488309790550414e-07,
+      "loss": 0.0005,
+      "step": 10363
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.488066244520214e-07,
+      "loss": 0.0005,
+      "step": 10364
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.4878226984900146e-07,
+      "loss": 0.0004,
+      "step": 10365
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.4875791524598146e-07,
+      "loss": 0.0002,
+      "step": 10366
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 2.487335606429615e-07,
+      "loss": 0.0126,
+      "step": 10367
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.487092060399415e-07,
+      "loss": 0.0005,
+      "step": 10368
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.4868485143692156e-07,
+      "loss": 0.0003,
+      "step": 10369
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.4866049683390156e-07,
+      "loss": 0.0004,
+      "step": 10370
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.486361422308816e-07,
+      "loss": 0.0004,
+      "step": 10371
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.4861178762786166e-07,
+      "loss": 0.0003,
+      "step": 10372
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.485874330248417e-07,
+      "loss": 0.0002,
+      "step": 10373
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.485630784218217e-07,
+      "loss": 0.0003,
+      "step": 10374
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.4853872381880176e-07,
+      "loss": 0.0005,
+      "step": 10375
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.4851436921578176e-07,
+      "loss": 0.0005,
+      "step": 10376
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 2.484900146127618e-07,
+      "loss": 0.0005,
+      "step": 10377
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.4846566000974187e-07,
+      "loss": 0.0004,
+      "step": 10378
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.4844130540672186e-07,
+      "loss": 0.0005,
+      "step": 10379
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.484169508037019e-07,
+      "loss": 0.0004,
+      "step": 10380
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.483925962006819e-07,
+      "loss": 0.0004,
+      "step": 10381
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.4836824159766197e-07,
+      "loss": 0.0003,
+      "step": 10382
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.4834388699464196e-07,
+      "loss": 0.0005,
+      "step": 10383
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.48319532391622e-07,
+      "loss": 0.0004,
+      "step": 10384
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.48295177788602e-07,
+      "loss": 0.0003,
+      "step": 10385
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.4827082318558207e-07,
+      "loss": 0.0005,
+      "step": 10386
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 2.482464685825621e-07,
+      "loss": 0.0005,
+      "step": 10387
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.482221139795421e-07,
+      "loss": 0.0003,
+      "step": 10388
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.4819775937652217e-07,
+      "loss": 0.0005,
+      "step": 10389
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.4817340477350217e-07,
+      "loss": 0.0006,
+      "step": 10390
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.481490501704822e-07,
+      "loss": 0.0004,
+      "step": 10391
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.481246955674622e-07,
+      "loss": 0.0008,
+      "step": 10392
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.4810034096444227e-07,
+      "loss": 0.0004,
+      "step": 10393
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.480759863614223e-07,
+      "loss": 0.0003,
+      "step": 10394
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.480516317584023e-07,
+      "loss": 0.0003,
+      "step": 10395
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.4802727715538237e-07,
+      "loss": 0.0004,
+      "step": 10396
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.4800292255236237e-07,
+      "loss": 0.0004,
+      "step": 10397
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 2.479785679493424e-07,
+      "loss": 0.0008,
+      "step": 10398
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.479542133463224e-07,
+      "loss": 0.0005,
+      "step": 10399
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.4792985874330247e-07,
+      "loss": 0.0006,
+      "step": 10400
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.4790550414028247e-07,
+      "loss": 0.0003,
+      "step": 10401
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.478811495372625e-07,
+      "loss": 0.0004,
+      "step": 10402
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.478567949342426e-07,
+      "loss": 0.0004,
+      "step": 10403
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.4783244033122257e-07,
+      "loss": 0.0004,
+      "step": 10404
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.478080857282026e-07,
+      "loss": 0.0005,
+      "step": 10405
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.477837311251826e-07,
+      "loss": 0.0003,
+      "step": 10406
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.477593765221627e-07,
+      "loss": 0.0016,
+      "step": 10407
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 2.4773502191914267e-07,
+      "loss": 0.0005,
+      "step": 10408
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.477106673161227e-07,
+      "loss": 0.0005,
+      "step": 10409
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.476863127131028e-07,
+      "loss": 0.0003,
+      "step": 10410
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.4766195811008283e-07,
+      "loss": 0.0003,
+      "step": 10411
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.4763760350706283e-07,
+      "loss": 0.0004,
+      "step": 10412
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.476132489040429e-07,
+      "loss": 0.0003,
+      "step": 10413
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.475888943010229e-07,
+      "loss": 0.0004,
+      "step": 10414
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.4756453969800293e-07,
+      "loss": 0.0003,
+      "step": 10415
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.4754018509498293e-07,
+      "loss": 0.0005,
+      "step": 10416
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.47515830491963e-07,
+      "loss": 0.0004,
+      "step": 10417
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 2.4749147588894303e-07,
+      "loss": 0.0003,
+      "step": 10418
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.4746712128592303e-07,
+      "loss": 0.0003,
+      "step": 10419
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.474427666829031e-07,
+      "loss": 0.0006,
+      "step": 10420
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.474184120798831e-07,
+      "loss": 0.0003,
+      "step": 10421
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.4739405747686313e-07,
+      "loss": 0.0004,
+      "step": 10422
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.4736970287384313e-07,
+      "loss": 0.0004,
+      "step": 10423
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.473453482708232e-07,
+      "loss": 0.0002,
+      "step": 10424
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.4732099366780323e-07,
+      "loss": 0.0003,
+      "step": 10425
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.4729663906478323e-07,
+      "loss": 0.0003,
+      "step": 10426
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.472722844617633e-07,
+      "loss": 0.0004,
+      "step": 10427
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 2.472479298587433e-07,
+      "loss": 0.0003,
+      "step": 10428
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.4722357525572333e-07,
+      "loss": 0.0002,
+      "step": 10429
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.4719922065270333e-07,
+      "loss": 0.0003,
+      "step": 10430
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.471748660496834e-07,
+      "loss": 0.0005,
+      "step": 10431
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.471505114466634e-07,
+      "loss": 0.0005,
+      "step": 10432
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.4712615684364343e-07,
+      "loss": 0.0003,
+      "step": 10433
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.471018022406235e-07,
+      "loss": 0.0007,
+      "step": 10434
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.470774476376035e-07,
+      "loss": 0.0003,
+      "step": 10435
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.4705309303458354e-07,
+      "loss": 0.0006,
+      "step": 10436
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.4702873843156354e-07,
+      "loss": 0.0003,
+      "step": 10437
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.470043838285436e-07,
+      "loss": 0.0008,
+      "step": 10438
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 2.469800292255236e-07,
+      "loss": 0.0004,
+      "step": 10439
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.4695567462250364e-07,
+      "loss": 0.0005,
+      "step": 10440
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.469313200194837e-07,
+      "loss": 0.0002,
+      "step": 10441
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.469069654164637e-07,
+      "loss": 0.0003,
+      "step": 10442
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.4688261081344374e-07,
+      "loss": 0.0004,
+      "step": 10443
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.4685825621042374e-07,
+      "loss": 0.0005,
+      "step": 10444
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.468339016074038e-07,
+      "loss": 0.0004,
+      "step": 10445
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.468095470043838e-07,
+      "loss": 0.0005,
+      "step": 10446
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.4678519240136384e-07,
+      "loss": 0.0004,
+      "step": 10447
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.4676083779834384e-07,
+      "loss": 0.0005,
+      "step": 10448
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 2.467364831953239e-07,
+      "loss": 0.0005,
+      "step": 10449
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.4671212859230394e-07,
+      "loss": 0.0004,
+      "step": 10450
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.46687773989284e-07,
+      "loss": 0.0006,
+      "step": 10451
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.46663419386264e-07,
+      "loss": 0.0004,
+      "step": 10452
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.4663906478324404e-07,
+      "loss": 0.0004,
+      "step": 10453
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.4661471018022404e-07,
+      "loss": 0.0005,
+      "step": 10454
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.465903555772041e-07,
+      "loss": 0.0004,
+      "step": 10455
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.465660009741841e-07,
+      "loss": 0.0004,
+      "step": 10456
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.4654164637116414e-07,
+      "loss": 0.0003,
+      "step": 10457
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.465172917681442e-07,
+      "loss": 0.0004,
+      "step": 10458
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 2.464929371651242e-07,
+      "loss": 0.0005,
+      "step": 10459
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.4646858256210425e-07,
+      "loss": 0.0003,
+      "step": 10460
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.4644422795908425e-07,
+      "loss": 0.0002,
+      "step": 10461
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.464198733560643e-07,
+      "loss": 0.0005,
+      "step": 10462
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.463955187530443e-07,
+      "loss": 0.0004,
+      "step": 10463
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.4637116415002435e-07,
+      "loss": 0.0003,
+      "step": 10464
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.463468095470044e-07,
+      "loss": 0.0006,
+      "step": 10465
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.463224549439844e-07,
+      "loss": 0.0003,
+      "step": 10466
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.4629810034096445e-07,
+      "loss": 0.0004,
+      "step": 10467
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.4627374573794445e-07,
+      "loss": 0.0003,
+      "step": 10468
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.462493911349245e-07,
+      "loss": 0.0004,
+      "step": 10469
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 2.462250365319045e-07,
+      "loss": 0.0004,
+      "step": 10470
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.4620068192888455e-07,
+      "loss": 0.0003,
+      "step": 10471
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.4617632732586455e-07,
+      "loss": 0.0004,
+      "step": 10472
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.461519727228446e-07,
+      "loss": 0.0003,
+      "step": 10473
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.4612761811982465e-07,
+      "loss": 0.0004,
+      "step": 10474
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.4610326351680465e-07,
+      "loss": 0.0004,
+      "step": 10475
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.460789089137847e-07,
+      "loss": 0.0004,
+      "step": 10476
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.460545543107647e-07,
+      "loss": 0.0003,
+      "step": 10477
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.4603019970774475e-07,
+      "loss": 0.0004,
+      "step": 10478
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.4600584510472475e-07,
+      "loss": 0.0004,
+      "step": 10479
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 2.459814905017048e-07,
+      "loss": 0.0007,
+      "step": 10480
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.4595713589868485e-07,
+      "loss": 0.0005,
+      "step": 10481
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.4593278129566485e-07,
+      "loss": 0.0003,
+      "step": 10482
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.459084266926449e-07,
+      "loss": 0.0003,
+      "step": 10483
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.458840720896249e-07,
+      "loss": 0.0005,
+      "step": 10484
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.4585971748660495e-07,
+      "loss": 0.0004,
+      "step": 10485
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.4583536288358495e-07,
+      "loss": 0.0005,
+      "step": 10486
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.45811008280565e-07,
+      "loss": 1.7569,
+      "step": 10487
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.4578665367754506e-07,
+      "loss": 0.0003,
+      "step": 10488
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.457622990745251e-07,
+      "loss": 0.0004,
+      "step": 10489
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 2.457379444715051e-07,
+      "loss": 0.0003,
+      "step": 10490
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.4571358986848516e-07,
+      "loss": 0.0003,
+      "step": 10491
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.4568923526546516e-07,
+      "loss": 0.0004,
+      "step": 10492
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.456648806624452e-07,
+      "loss": 0.0003,
+      "step": 10493
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.456405260594252e-07,
+      "loss": 0.0003,
+      "step": 10494
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.4561617145640526e-07,
+      "loss": 0.0004,
+      "step": 10495
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.455918168533853e-07,
+      "loss": 0.0004,
+      "step": 10496
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.455674622503653e-07,
+      "loss": 0.0004,
+      "step": 10497
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.4554310764734536e-07,
+      "loss": 0.0003,
+      "step": 10498
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.4551875304432536e-07,
+      "loss": 0.0003,
+      "step": 10499
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 2.454943984413054e-07,
+      "loss": 0.0003,
+      "step": 10500
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.454700438382854e-07,
+      "loss": 0.0003,
+      "step": 10501
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.4544568923526546e-07,
+      "loss": 0.0004,
+      "step": 10502
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.4542133463224546e-07,
+      "loss": 0.0004,
+      "step": 10503
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.453969800292255e-07,
+      "loss": 0.0006,
+      "step": 10504
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.4537262542620556e-07,
+      "loss": 0.0004,
+      "step": 10505
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.4534827082318556e-07,
+      "loss": 0.0003,
+      "step": 10506
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.453239162201656e-07,
+      "loss": 0.0004,
+      "step": 10507
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.452995616171456e-07,
+      "loss": 0.0003,
+      "step": 10508
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.4527520701412566e-07,
+      "loss": 0.0006,
+      "step": 10509
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.4525085241110566e-07,
+      "loss": 0.0003,
+      "step": 10510
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 2.452264978080857e-07,
+      "loss": 0.0003,
+      "step": 10511
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.4520214320506577e-07,
+      "loss": 0.0003,
+      "step": 10512
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.4517778860204577e-07,
+      "loss": 0.0003,
+      "step": 10513
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.451534339990258e-07,
+      "loss": 0.0003,
+      "step": 10514
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.451290793960058e-07,
+      "loss": 0.0004,
+      "step": 10515
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.4510472479298587e-07,
+      "loss": 0.0003,
+      "step": 10516
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.4508037018996587e-07,
+      "loss": 0.0003,
+      "step": 10517
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.450560155869459e-07,
+      "loss": 0.0006,
+      "step": 10518
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.450316609839259e-07,
+      "loss": 0.0003,
+      "step": 10519
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.4500730638090597e-07,
+      "loss": 0.0004,
+      "step": 10520
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 2.44982951777886e-07,
+      "loss": 0.0004,
+      "step": 10521
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.44958597174866e-07,
+      "loss": 0.0003,
+      "step": 10522
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.4493424257184607e-07,
+      "loss": 0.0003,
+      "step": 10523
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.4490988796882607e-07,
+      "loss": 0.0004,
+      "step": 10524
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.448855333658061e-07,
+      "loss": 0.0004,
+      "step": 10525
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.4486117876278617e-07,
+      "loss": 0.0003,
+      "step": 10526
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.448368241597662e-07,
+      "loss": 0.0003,
+      "step": 10527
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.448124695567462e-07,
+      "loss": 0.0003,
+      "step": 10528
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.447881149537263e-07,
+      "loss": 0.0003,
+      "step": 10529
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.4476376035070627e-07,
+      "loss": 0.4335,
+      "step": 10530
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 2.447394057476863e-07,
+      "loss": 0.0005,
+      "step": 10531
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.447150511446663e-07,
+      "loss": 0.0006,
+      "step": 10532
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.446906965416464e-07,
+      "loss": 0.0003,
+      "step": 10533
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.4466634193862637e-07,
+      "loss": 0.0003,
+      "step": 10534
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.446419873356064e-07,
+      "loss": 0.0008,
+      "step": 10535
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.446176327325865e-07,
+      "loss": 0.0004,
+      "step": 10536
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.445932781295665e-07,
+      "loss": 0.0004,
+      "step": 10537
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.4456892352654653e-07,
+      "loss": 0.0004,
+      "step": 10538
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.445445689235265e-07,
+      "loss": 0.0005,
+      "step": 10539
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.445202143205066e-07,
+      "loss": 0.0004,
+      "step": 10540
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.444958597174866e-07,
+      "loss": 0.0003,
+      "step": 10541
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 2.4447150511446663e-07,
+      "loss": 0.0005,
+      "step": 10542
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.444471505114467e-07,
+      "loss": 0.0003,
+      "step": 10543
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.444227959084267e-07,
+      "loss": 0.0003,
+      "step": 10544
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.4439844130540673e-07,
+      "loss": 0.0003,
+      "step": 10545
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.4437408670238673e-07,
+      "loss": 0.0002,
+      "step": 10546
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.443497320993668e-07,
+      "loss": 0.0005,
+      "step": 10547
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.443253774963468e-07,
+      "loss": 0.0004,
+      "step": 10548
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.4430102289332683e-07,
+      "loss": 0.0003,
+      "step": 10549
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.4427666829030683e-07,
+      "loss": 0.0011,
+      "step": 10550
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.442523136872869e-07,
+      "loss": 0.0003,
+      "step": 10551
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 2.4422795908426693e-07,
+      "loss": 0.0004,
+      "step": 10552
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.4420360448124693e-07,
+      "loss": 0.0006,
+      "step": 10553
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.44179249878227e-07,
+      "loss": 0.0007,
+      "step": 10554
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.44154895275207e-07,
+      "loss": 0.0004,
+      "step": 10555
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.4413054067218703e-07,
+      "loss": 0.0003,
+      "step": 10556
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.4410618606916703e-07,
+      "loss": 0.0006,
+      "step": 10557
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.440818314661471e-07,
+      "loss": 0.0003,
+      "step": 10558
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.4405747686312713e-07,
+      "loss": 0.0007,
+      "step": 10559
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.4403312226010713e-07,
+      "loss": 0.0003,
+      "step": 10560
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.440087676570872e-07,
+      "loss": 0.0003,
+      "step": 10561
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 2.439844130540672e-07,
+      "loss": 0.0003,
+      "step": 10562
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.4396005845104724e-07,
+      "loss": 0.0005,
+      "step": 10563
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.439357038480273e-07,
+      "loss": 0.0004,
+      "step": 10564
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.439113492450073e-07,
+      "loss": 0.0004,
+      "step": 10565
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.4388699464198734e-07,
+      "loss": 0.0003,
+      "step": 10566
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.438626400389674e-07,
+      "loss": 0.0006,
+      "step": 10567
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.438382854359474e-07,
+      "loss": 0.0005,
+      "step": 10568
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.4381393083292744e-07,
+      "loss": 0.0003,
+      "step": 10569
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.4378957622990744e-07,
+      "loss": 0.0003,
+      "step": 10570
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.437652216268875e-07,
+      "loss": 0.0003,
+      "step": 10571
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 2.437408670238675e-07,
+      "loss": 0.0002,
+      "step": 10572
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.4371651242084754e-07,
+      "loss": 0.0003,
+      "step": 10573
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.4369215781782754e-07,
+      "loss": 0.0004,
+      "step": 10574
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.436678032148076e-07,
+      "loss": 0.0005,
+      "step": 10575
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.4364344861178764e-07,
+      "loss": 0.0004,
+      "step": 10576
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.4361909400876764e-07,
+      "loss": 0.0003,
+      "step": 10577
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.435947394057477e-07,
+      "loss": 0.0004,
+      "step": 10578
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.435703848027277e-07,
+      "loss": 0.0004,
+      "step": 10579
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.4354603019970774e-07,
+      "loss": 0.0004,
+      "step": 10580
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.4352167559668774e-07,
+      "loss": 0.0005,
+      "step": 10581
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.434973209936678e-07,
+      "loss": 0.0004,
+      "step": 10582
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 2.4347296639064784e-07,
+      "loss": 0.0003,
+      "step": 10583
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.4344861178762784e-07,
+      "loss": 0.0006,
+      "step": 10584
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.434242571846079e-07,
+      "loss": 0.0005,
+      "step": 10585
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.433999025815879e-07,
+      "loss": 0.0004,
+      "step": 10586
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.4337554797856795e-07,
+      "loss": 0.0005,
+      "step": 10587
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.4335119337554794e-07,
+      "loss": 0.0005,
+      "step": 10588
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.43326838772528e-07,
+      "loss": 0.0004,
+      "step": 10589
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.4330248416950805e-07,
+      "loss": 1.2901,
+      "step": 10590
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.4327812956648805e-07,
+      "loss": 0.0005,
+      "step": 10591
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.432537749634681e-07,
+      "loss": 0.0003,
+      "step": 10592
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 2.432294203604481e-07,
+      "loss": 0.0005,
+      "step": 10593
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.4320506575742815e-07,
+      "loss": 0.0006,
+      "step": 10594
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.4318071115440815e-07,
+      "loss": 0.0003,
+      "step": 10595
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.431563565513882e-07,
+      "loss": 0.0003,
+      "step": 10596
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.431320019483682e-07,
+      "loss": 0.0003,
+      "step": 10597
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.4310764734534825e-07,
+      "loss": 0.0004,
+      "step": 10598
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.430832927423283e-07,
+      "loss": 0.0005,
+      "step": 10599
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.430589381393083e-07,
+      "loss": 0.0003,
+      "step": 10600
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.4303458353628835e-07,
+      "loss": 0.0005,
+      "step": 10601
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.4301022893326835e-07,
+      "loss": 0.0004,
+      "step": 10602
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 2.429858743302484e-07,
+      "loss": 0.0006,
+      "step": 10603
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.4296151972722845e-07,
+      "loss": 0.0004,
+      "step": 10604
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.4293716512420845e-07,
+      "loss": 0.0004,
+      "step": 10605
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.429128105211885e-07,
+      "loss": 0.0006,
+      "step": 10606
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.4288845591816855e-07,
+      "loss": 0.0512,
+      "step": 10607
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.4286410131514855e-07,
+      "loss": 0.0002,
+      "step": 10608
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.428397467121286e-07,
+      "loss": 0.0002,
+      "step": 10609
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.428153921091086e-07,
+      "loss": 0.0003,
+      "step": 10610
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.4279103750608866e-07,
+      "loss": 0.0003,
+      "step": 10611
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.4276668290306865e-07,
+      "loss": 0.0005,
+      "step": 10612
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.427423283000487e-07,
+      "loss": 0.0004,
+      "step": 10613
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 2.4271797369702876e-07,
+      "loss": 0.0002,
+      "step": 10614
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.4269361909400876e-07,
+      "loss": 0.0004,
+      "step": 10615
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.426692644909888e-07,
+      "loss": 0.0004,
+      "step": 10616
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.426449098879688e-07,
+      "loss": 0.0003,
+      "step": 10617
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.4262055528494886e-07,
+      "loss": 0.0005,
+      "step": 10618
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.4259620068192886e-07,
+      "loss": 0.0005,
+      "step": 10619
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.425718460789089e-07,
+      "loss": 0.0003,
+      "step": 10620
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.425474914758889e-07,
+      "loss": 0.0004,
+      "step": 10621
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.4252313687286896e-07,
+      "loss": 0.0004,
+      "step": 10622
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.42498782269849e-07,
+      "loss": 0.0005,
+      "step": 10623
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 2.42474427666829e-07,
+      "loss": 0.0002,
+      "step": 10624
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.4245007306380906e-07,
+      "loss": 0.0004,
+      "step": 10625
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.4242571846078906e-07,
+      "loss": 0.0004,
+      "step": 10626
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.424013638577691e-07,
+      "loss": 0.0004,
+      "step": 10627
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.423770092547491e-07,
+      "loss": 0.0003,
+      "step": 10628
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.4235265465172916e-07,
+      "loss": 0.0006,
+      "step": 10629
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.423283000487092e-07,
+      "loss": 0.0003,
+      "step": 10630
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.423039454456892e-07,
+      "loss": 0.0004,
+      "step": 10631
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.4227959084266926e-07,
+      "loss": 0.0002,
+      "step": 10632
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.4225523623964926e-07,
+      "loss": 0.0004,
+      "step": 10633
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 2.422308816366293e-07,
+      "loss": 0.0005,
+      "step": 10634
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.422065270336093e-07,
+      "loss": 0.0004,
+      "step": 10635
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.4218217243058936e-07,
+      "loss": 0.0003,
+      "step": 10636
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.4215781782756936e-07,
+      "loss": 0.0003,
+      "step": 10637
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.421334632245494e-07,
+      "loss": 0.0003,
+      "step": 10638
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.4210910862152947e-07,
+      "loss": 0.0006,
+      "step": 10639
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.4208475401850946e-07,
+      "loss": 0.0003,
+      "step": 10640
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.420603994154895e-07,
+      "loss": 0.0006,
+      "step": 10641
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.4203604481246957e-07,
+      "loss": 0.0004,
+      "step": 10642
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.4201169020944957e-07,
+      "loss": 0.0003,
+      "step": 10643
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.419873356064296e-07,
+      "loss": 0.0004,
+      "step": 10644
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 2.4196298100340967e-07,
+      "loss": 0.0004,
+      "step": 10645
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.4193862640038967e-07,
+      "loss": 0.0003,
+      "step": 10646
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.419142717973697e-07,
+      "loss": 0.0004,
+      "step": 10647
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.418899171943497e-07,
+      "loss": 0.0004,
+      "step": 10648
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.4186556259132977e-07,
+      "loss": 0.0004,
+      "step": 10649
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.4184120798830977e-07,
+      "loss": 0.0004,
+      "step": 10650
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.418168533852898e-07,
+      "loss": 0.0027,
+      "step": 10651
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.417924987822698e-07,
+      "loss": 0.0004,
+      "step": 10652
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.4176814417924987e-07,
+      "loss": 0.0004,
+      "step": 10653
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.417437895762299e-07,
+      "loss": 0.0004,
+      "step": 10654
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 2.417194349732099e-07,
+      "loss": 0.0002,
+      "step": 10655
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.4169508037018997e-07,
+      "loss": 0.0003,
+      "step": 10656
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.4167072576716997e-07,
+      "loss": 0.0004,
+      "step": 10657
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.4164637116415e-07,
+      "loss": 0.0005,
+      "step": 10658
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.4162201656113e-07,
+      "loss": 0.0004,
+      "step": 10659
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.415976619581101e-07,
+      "loss": 0.0004,
+      "step": 10660
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.415733073550901e-07,
+      "loss": 0.0002,
+      "step": 10661
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.415489527520701e-07,
+      "loss": 0.0003,
+      "step": 10662
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.415245981490502e-07,
+      "loss": 0.0003,
+      "step": 10663
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.415002435460302e-07,
+      "loss": 0.0004,
+      "step": 10664
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 2.414758889430102e-07,
+      "loss": 0.0004,
+      "step": 10665
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.414515343399902e-07,
+      "loss": 0.0004,
+      "step": 10666
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.414271797369703e-07,
+      "loss": 0.0003,
+      "step": 10667
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.414028251339503e-07,
+      "loss": 0.0004,
+      "step": 10668
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.4137847053093033e-07,
+      "loss": 0.0004,
+      "step": 10669
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.413541159279104e-07,
+      "loss": 0.0004,
+      "step": 10670
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.413297613248904e-07,
+      "loss": 0.0003,
+      "step": 10671
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.4130540672187043e-07,
+      "loss": 0.0004,
+      "step": 10672
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.4128105211885043e-07,
+      "loss": 0.0005,
+      "step": 10673
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.412566975158305e-07,
+      "loss": 0.0002,
+      "step": 10674
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 2.412323429128105e-07,
+      "loss": 0.0004,
+      "step": 10675
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.4120798830979053e-07,
+      "loss": 0.0005,
+      "step": 10676
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.411836337067706e-07,
+      "loss": 0.0003,
+      "step": 10677
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.411592791037506e-07,
+      "loss": 0.0004,
+      "step": 10678
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.4113492450073063e-07,
+      "loss": 0.0005,
+      "step": 10679
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.411105698977107e-07,
+      "loss": 0.0006,
+      "step": 10680
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.410862152946907e-07,
+      "loss": 0.0002,
+      "step": 10681
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.4106186069167073e-07,
+      "loss": 0.0003,
+      "step": 10682
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.4103750608865073e-07,
+      "loss": 0.0006,
+      "step": 10683
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.410131514856308e-07,
+      "loss": 0.0005,
+      "step": 10684
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.4098879688261084e-07,
+      "loss": 0.0005,
+      "step": 10685
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 2.4096444227959083e-07,
+      "loss": 0.0005,
+      "step": 10686
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.409400876765709e-07,
+      "loss": 0.0004,
+      "step": 10687
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.409157330735509e-07,
+      "loss": 0.0002,
+      "step": 10688
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.4089137847053094e-07,
+      "loss": 0.0004,
+      "step": 10689
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.4086702386751093e-07,
+      "loss": 0.0003,
+      "step": 10690
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.40842669264491e-07,
+      "loss": 0.0004,
+      "step": 10691
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.4081831466147104e-07,
+      "loss": 0.0004,
+      "step": 10692
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.4079396005845104e-07,
+      "loss": 0.0004,
+      "step": 10693
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.407696054554311e-07,
+      "loss": 0.0004,
+      "step": 10694
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.407452508524111e-07,
+      "loss": 0.0004,
+      "step": 10695
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 2.4072089624939114e-07,
+      "loss": 0.0003,
+      "step": 10696
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.4069654164637114e-07,
+      "loss": 0.0004,
+      "step": 10697
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.406721870433512e-07,
+      "loss": 0.0003,
+      "step": 10698
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.406478324403312e-07,
+      "loss": 0.0003,
+      "step": 10699
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.4062347783731124e-07,
+      "loss": 0.0003,
+      "step": 10700
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.405991232342913e-07,
+      "loss": 0.0003,
+      "step": 10701
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.405747686312713e-07,
+      "loss": 0.0003,
+      "step": 10702
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.4055041402825134e-07,
+      "loss": 0.0004,
+      "step": 10703
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.4052605942523134e-07,
+      "loss": 0.0006,
+      "step": 10704
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.405017048222114e-07,
+      "loss": 0.0003,
+      "step": 10705
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 2.404773502191914e-07,
+      "loss": 0.0004,
+      "step": 10706
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.4045299561617144e-07,
+      "loss": 0.0004,
+      "step": 10707
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.404286410131515e-07,
+      "loss": 0.0003,
+      "step": 10708
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.404042864101315e-07,
+      "loss": 1.8644,
+      "step": 10709
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.4037993180711154e-07,
+      "loss": 0.0004,
+      "step": 10710
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.4035557720409154e-07,
+      "loss": 0.0004,
+      "step": 10711
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.403312226010716e-07,
+      "loss": 0.0004,
+      "step": 10712
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.403068679980516e-07,
+      "loss": 0.0004,
+      "step": 10713
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.4028251339503164e-07,
+      "loss": 0.0004,
+      "step": 10714
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.4025815879201164e-07,
+      "loss": 0.0003,
+      "step": 10715
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.402338041889917e-07,
+      "loss": 0.0003,
+      "step": 10716
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 2.4020944958597175e-07,
+      "loss": 0.0004,
+      "step": 10717
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.401850949829518e-07,
+      "loss": 0.0004,
+      "step": 10718
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.401607403799318e-07,
+      "loss": 0.0004,
+      "step": 10719
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.4013638577691185e-07,
+      "loss": 0.0003,
+      "step": 10720
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.4011203117389185e-07,
+      "loss": 0.0005,
+      "step": 10721
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.400876765708719e-07,
+      "loss": 2.108,
+      "step": 10722
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.400633219678519e-07,
+      "loss": 0.0006,
+      "step": 10723
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.4003896736483195e-07,
+      "loss": 0.0005,
+      "step": 10724
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.40014612761812e-07,
+      "loss": 0.0004,
+      "step": 10725
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.39990258158792e-07,
+      "loss": 0.0002,
+      "step": 10726
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 2.3996590355577205e-07,
+      "loss": 0.0005,
+      "step": 10727
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.3994154895275205e-07,
+      "loss": 0.0004,
+      "step": 10728
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.399171943497321e-07,
+      "loss": 0.0003,
+      "step": 10729
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.398928397467121e-07,
+      "loss": 0.0005,
+      "step": 10730
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.3986848514369215e-07,
+      "loss": 0.0003,
+      "step": 10731
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.398441305406722e-07,
+      "loss": 0.0003,
+      "step": 10732
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.398197759376522e-07,
+      "loss": 0.0005,
+      "step": 10733
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.3979542133463225e-07,
+      "loss": 0.0002,
+      "step": 10734
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.3977106673161225e-07,
+      "loss": 0.0003,
+      "step": 10735
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.397467121285923e-07,
+      "loss": 0.0003,
+      "step": 10736
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 2.397223575255723e-07,
+      "loss": 0.0004,
+      "step": 10737
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3969800292255235e-07,
+      "loss": 0.0005,
+      "step": 10738
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3967364831953235e-07,
+      "loss": 0.0003,
+      "step": 10739
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.396492937165124e-07,
+      "loss": 0.0005,
+      "step": 10740
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3962493911349246e-07,
+      "loss": 0.0005,
+      "step": 10741
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3960058451047245e-07,
+      "loss": 0.0007,
+      "step": 10742
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.395762299074525e-07,
+      "loss": 0.0004,
+      "step": 10743
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.395518753044325e-07,
+      "loss": 0.0003,
+      "step": 10744
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3952752070141256e-07,
+      "loss": 0.0002,
+      "step": 10745
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.3950316609839256e-07,
+      "loss": 0.0014,
+      "step": 10746
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 2.394788114953726e-07,
+      "loss": 0.0003,
+      "step": 10747
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.3945445689235266e-07,
+      "loss": 0.0004,
+      "step": 10748
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.3943010228933266e-07,
+      "loss": 0.0005,
+      "step": 10749
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.394057476863127e-07,
+      "loss": 0.0003,
+      "step": 10750
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.393813930832927e-07,
+      "loss": 0.0002,
+      "step": 10751
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.3935703848027276e-07,
+      "loss": 0.0004,
+      "step": 10752
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.3933268387725276e-07,
+      "loss": 0.0005,
+      "step": 10753
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.393083292742328e-07,
+      "loss": 0.0004,
+      "step": 10754
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.392839746712128e-07,
+      "loss": 0.0005,
+      "step": 10755
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.392596200681929e-07,
+      "loss": 0.0005,
+      "step": 10756
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.392352654651729e-07,
+      "loss": 0.0003,
+      "step": 10757
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 2.3921091086215296e-07,
+      "loss": 0.0003,
+      "step": 10758
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.3918655625913296e-07,
+      "loss": 0.0004,
+      "step": 10759
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.39162201656113e-07,
+      "loss": 0.0006,
+      "step": 10760
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.39137847053093e-07,
+      "loss": 0.0003,
+      "step": 10761
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.3911349245007306e-07,
+      "loss": 0.0004,
+      "step": 10762
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.390891378470531e-07,
+      "loss": 0.0003,
+      "step": 10763
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.390647832440331e-07,
+      "loss": 0.0003,
+      "step": 10764
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.3904042864101317e-07,
+      "loss": 0.0089,
+      "step": 10765
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.3901607403799316e-07,
+      "loss": 0.0004,
+      "step": 10766
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.389917194349732e-07,
+      "loss": 0.0004,
+      "step": 10767
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 2.389673648319532e-07,
+      "loss": 0.0003,
+      "step": 10768
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3894301022893327e-07,
+      "loss": 0.0005,
+      "step": 10769
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3891865562591327e-07,
+      "loss": 0.0005,
+      "step": 10770
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.388943010228933e-07,
+      "loss": 0.0003,
+      "step": 10771
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3886994641987337e-07,
+      "loss": 0.0004,
+      "step": 10772
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3884559181685337e-07,
+      "loss": 0.0004,
+      "step": 10773
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.388212372138334e-07,
+      "loss": 0.0003,
+      "step": 10774
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.387968826108134e-07,
+      "loss": 0.0003,
+      "step": 10775
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3877252800779347e-07,
+      "loss": 0.0005,
+      "step": 10776
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.3874817340477347e-07,
+      "loss": 0.0003,
+      "step": 10777
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 2.387238188017535e-07,
+      "loss": 0.0003,
+      "step": 10778
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3869946419873357e-07,
+      "loss": 0.0003,
+      "step": 10779
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3867510959571357e-07,
+      "loss": 0.0005,
+      "step": 10780
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.386507549926936e-07,
+      "loss": 0.0003,
+      "step": 10781
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.386264003896736e-07,
+      "loss": 0.0003,
+      "step": 10782
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3860204578665367e-07,
+      "loss": 0.001,
+      "step": 10783
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3857769118363367e-07,
+      "loss": 0.0004,
+      "step": 10784
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.385533365806137e-07,
+      "loss": 0.0004,
+      "step": 10785
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.385289819775937e-07,
+      "loss": 0.0004,
+      "step": 10786
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.3850462737457377e-07,
+      "loss": 0.0003,
+      "step": 10787
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.384802727715538e-07,
+      "loss": 0.0003,
+      "step": 10788
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 2.384559181685338e-07,
+      "loss": 0.0005,
+      "step": 10789
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.384315635655139e-07,
+      "loss": 0.0004,
+      "step": 10790
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.384072089624939e-07,
+      "loss": 0.0004,
+      "step": 10791
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3838285435947392e-07,
+      "loss": 0.0002,
+      "step": 10792
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3835849975645395e-07,
+      "loss": 0.0002,
+      "step": 10793
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.38334145153434e-07,
+      "loss": 0.0006,
+      "step": 10794
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3830979055041403e-07,
+      "loss": 0.0004,
+      "step": 10795
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3828543594739405e-07,
+      "loss": 0.0004,
+      "step": 10796
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3826108134437408e-07,
+      "loss": 0.0003,
+      "step": 10797
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3823672674135413e-07,
+      "loss": 0.0003,
+      "step": 10798
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 2.3821237213833415e-07,
+      "loss": 0.0005,
+      "step": 10799
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3818801753531418e-07,
+      "loss": 0.0003,
+      "step": 10800
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.381636629322942e-07,
+      "loss": 0.0003,
+      "step": 10801
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3813930832927423e-07,
+      "loss": 0.0003,
+      "step": 10802
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3811495372625425e-07,
+      "loss": 0.0002,
+      "step": 10803
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3809059912323428e-07,
+      "loss": 0.0003,
+      "step": 10804
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.380662445202143e-07,
+      "loss": 0.0003,
+      "step": 10805
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3804188991719436e-07,
+      "loss": 0.0002,
+      "step": 10806
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3801753531417438e-07,
+      "loss": 0.0003,
+      "step": 10807
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.379931807111544e-07,
+      "loss": 0.0003,
+      "step": 10808
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 2.3796882610813443e-07,
+      "loss": 0.0005,
+      "step": 10809
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3794447150511446e-07,
+      "loss": 0.0004,
+      "step": 10810
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3792011690209448e-07,
+      "loss": 0.0005,
+      "step": 10811
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.378957622990745e-07,
+      "loss": 0.0003,
+      "step": 10812
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3787140769605453e-07,
+      "loss": 0.0003,
+      "step": 10813
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3784705309303458e-07,
+      "loss": 0.0003,
+      "step": 10814
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.378226984900146e-07,
+      "loss": 0.0005,
+      "step": 10815
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3779834388699463e-07,
+      "loss": 0.0012,
+      "step": 10816
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3777398928397466e-07,
+      "loss": 0.0003,
+      "step": 10817
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.3774963468095468e-07,
+      "loss": 0.0003,
+      "step": 10818
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 2.377252800779347e-07,
+      "loss": 0.0003,
+      "step": 10819
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3770092547491474e-07,
+      "loss": 0.0004,
+      "step": 10820
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3767657087189476e-07,
+      "loss": 0.0003,
+      "step": 10821
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.376522162688748e-07,
+      "loss": 0.0005,
+      "step": 10822
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3762786166585484e-07,
+      "loss": 0.0005,
+      "step": 10823
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3760350706283486e-07,
+      "loss": 0.0005,
+      "step": 10824
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.375791524598149e-07,
+      "loss": 0.0003,
+      "step": 10825
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.375547978567949e-07,
+      "loss": 0.0003,
+      "step": 10826
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3753044325377494e-07,
+      "loss": 0.0004,
+      "step": 10827
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3750608865075496e-07,
+      "loss": 0.0002,
+      "step": 10828
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.37481734047735e-07,
+      "loss": 0.0003,
+      "step": 10829
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 2.3745737944471504e-07,
+      "loss": 0.0005,
+      "step": 10830
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3743302484169507e-07,
+      "loss": 0.0003,
+      "step": 10831
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.374086702386751e-07,
+      "loss": 0.0003,
+      "step": 10832
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3738431563565514e-07,
+      "loss": 0.0004,
+      "step": 10833
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3735996103263517e-07,
+      "loss": 0.0003,
+      "step": 10834
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.373356064296152e-07,
+      "loss": 0.0005,
+      "step": 10835
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3731125182659522e-07,
+      "loss": 0.0007,
+      "step": 10836
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3728689722357524e-07,
+      "loss": 0.0004,
+      "step": 10837
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.372625426205553e-07,
+      "loss": 0.0002,
+      "step": 10838
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3723818801753532e-07,
+      "loss": 0.0004,
+      "step": 10839
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 2.3721383341451534e-07,
+      "loss": 0.0004,
+      "step": 10840
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3718947881149537e-07,
+      "loss": 0.0002,
+      "step": 10841
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.371651242084754e-07,
+      "loss": 0.0002,
+      "step": 10842
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3714076960545542e-07,
+      "loss": 0.0002,
+      "step": 10843
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3711641500243545e-07,
+      "loss": 0.0003,
+      "step": 10844
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3709206039941547e-07,
+      "loss": 0.0003,
+      "step": 10845
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3706770579639552e-07,
+      "loss": 0.0006,
+      "step": 10846
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3704335119337555e-07,
+      "loss": 0.0003,
+      "step": 10847
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3701899659035557e-07,
+      "loss": 0.0003,
+      "step": 10848
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.369946419873356e-07,
+      "loss": 0.0006,
+      "step": 10849
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 2.3697028738431562e-07,
+      "loss": 0.0004,
+      "step": 10850
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3694593278129565e-07,
+      "loss": 0.0004,
+      "step": 10851
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3692157817827567e-07,
+      "loss": 0.0004,
+      "step": 10852
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.368972235752557e-07,
+      "loss": 0.0004,
+      "step": 10853
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3687286897223575e-07,
+      "loss": 0.0002,
+      "step": 10854
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3684851436921577e-07,
+      "loss": 0.0004,
+      "step": 10855
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.368241597661958e-07,
+      "loss": 0.0005,
+      "step": 10856
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3679980516317583e-07,
+      "loss": 0.0004,
+      "step": 10857
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3677545056015585e-07,
+      "loss": 0.0003,
+      "step": 10858
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3675109595713588e-07,
+      "loss": 0.0004,
+      "step": 10859
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.367267413541159e-07,
+      "loss": 0.001,
+      "step": 10860
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 2.3670238675109595e-07,
+      "loss": 0.0004,
+      "step": 10861
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3667803214807598e-07,
+      "loss": 1.8213,
+      "step": 10862
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.36653677545056e-07,
+      "loss": 0.0003,
+      "step": 10863
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3662932294203603e-07,
+      "loss": 0.0004,
+      "step": 10864
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3660496833901605e-07,
+      "loss": 0.0003,
+      "step": 10865
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3658061373599608e-07,
+      "loss": 0.0003,
+      "step": 10866
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.365562591329761e-07,
+      "loss": 0.0004,
+      "step": 10867
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3653190452995613e-07,
+      "loss": 0.0003,
+      "step": 10868
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3650754992693618e-07,
+      "loss": 0.0003,
+      "step": 10869
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.364831953239162e-07,
+      "loss": 0.0003,
+      "step": 10870
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 2.3645884072089626e-07,
+      "loss": 0.0003,
+      "step": 10871
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3643448611787628e-07,
+      "loss": 0.0003,
+      "step": 10872
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.364101315148563e-07,
+      "loss": 0.0003,
+      "step": 10873
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3638577691183633e-07,
+      "loss": 0.0004,
+      "step": 10874
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3636142230881636e-07,
+      "loss": 0.0003,
+      "step": 10875
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3633706770579638e-07,
+      "loss": 0.0006,
+      "step": 10876
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3631271310277643e-07,
+      "loss": 0.0004,
+      "step": 10877
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3628835849975646e-07,
+      "loss": 0.0005,
+      "step": 10878
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3626400389673648e-07,
+      "loss": 0.0004,
+      "step": 10879
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.362396492937165e-07,
+      "loss": 0.0006,
+      "step": 10880
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 2.3621529469069654e-07,
+      "loss": 0.0004,
+      "step": 10881
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.3619094008767656e-07,
+      "loss": 0.0003,
+      "step": 10882
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.3616658548465659e-07,
+      "loss": 0.0003,
+      "step": 10883
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.361422308816366e-07,
+      "loss": 0.0004,
+      "step": 10884
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.3611787627861666e-07,
+      "loss": 0.0003,
+      "step": 10885
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.360935216755967e-07,
+      "loss": 0.0004,
+      "step": 10886
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.360691670725767e-07,
+      "loss": 0.0003,
+      "step": 10887
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.3604481246955674e-07,
+      "loss": 0.0002,
+      "step": 10888
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.3602045786653676e-07,
+      "loss": 0.0003,
+      "step": 10889
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.359961032635168e-07,
+      "loss": 0.0004,
+      "step": 10890
+    },
+    {
+      "epoch": 10.58,
+      "learning_rate": 2.359717486604968e-07,
+      "loss": 0.0003,
+      "step": 10891
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3594739405747684e-07,
+      "loss": 0.0002,
+      "step": 10892
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.359230394544569e-07,
+      "loss": 0.0003,
+      "step": 10893
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3589868485143692e-07,
+      "loss": 0.0002,
+      "step": 10894
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3587433024841694e-07,
+      "loss": 0.0003,
+      "step": 10895
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3584997564539697e-07,
+      "loss": 0.0003,
+      "step": 10896
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.35825621042377e-07,
+      "loss": 0.0008,
+      "step": 10897
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3580126643935702e-07,
+      "loss": 0.0003,
+      "step": 10898
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3577691183633704e-07,
+      "loss": 0.0003,
+      "step": 10899
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3575255723331707e-07,
+      "loss": 0.0003,
+      "step": 10900
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3572820263029712e-07,
+      "loss": 0.0004,
+      "step": 10901
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 2.3570384802727714e-07,
+      "loss": 0.0004,
+      "step": 10902
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3567949342425717e-07,
+      "loss": 0.0004,
+      "step": 10903
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.356551388212372e-07,
+      "loss": 0.0003,
+      "step": 10904
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3563078421821722e-07,
+      "loss": 0.0007,
+      "step": 10905
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3560642961519724e-07,
+      "loss": 0.0008,
+      "step": 10906
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3558207501217727e-07,
+      "loss": 0.0003,
+      "step": 10907
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.355577204091573e-07,
+      "loss": 0.0004,
+      "step": 10908
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3553336580613737e-07,
+      "loss": 0.0036,
+      "step": 10909
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.355090112031174e-07,
+      "loss": 0.0007,
+      "step": 10910
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3548465660009742e-07,
+      "loss": 0.0005,
+      "step": 10911
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 2.3546030199707745e-07,
+      "loss": 0.0018,
+      "step": 10912
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3543594739405747e-07,
+      "loss": 0.0003,
+      "step": 10913
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.354115927910375e-07,
+      "loss": 0.0571,
+      "step": 10914
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3538723818801752e-07,
+      "loss": 0.0004,
+      "step": 10915
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3536288358499757e-07,
+      "loss": 0.0005,
+      "step": 10916
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.353385289819776e-07,
+      "loss": 0.0005,
+      "step": 10917
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3531417437895763e-07,
+      "loss": 0.0004,
+      "step": 10918
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3528981977593765e-07,
+      "loss": 0.0004,
+      "step": 10919
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3526546517291768e-07,
+      "loss": 0.0003,
+      "step": 10920
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.352411105698977e-07,
+      "loss": 0.0003,
+      "step": 10921
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 2.3521675596687773e-07,
+      "loss": 0.0004,
+      "step": 10922
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3519240136385775e-07,
+      "loss": 0.0004,
+      "step": 10923
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.351680467608378e-07,
+      "loss": 0.0003,
+      "step": 10924
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3514369215781783e-07,
+      "loss": 0.0004,
+      "step": 10925
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3511933755479785e-07,
+      "loss": 0.0004,
+      "step": 10926
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3509498295177788e-07,
+      "loss": 0.0005,
+      "step": 10927
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.350706283487579e-07,
+      "loss": 0.0005,
+      "step": 10928
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3504627374573793e-07,
+      "loss": 0.0005,
+      "step": 10929
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3502191914271795e-07,
+      "loss": 0.0003,
+      "step": 10930
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3499756453969798e-07,
+      "loss": 0.0004,
+      "step": 10931
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3497320993667803e-07,
+      "loss": 0.0005,
+      "step": 10932
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 2.3494885533365806e-07,
+      "loss": 0.0002,
+      "step": 10933
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3492450073063808e-07,
+      "loss": 0.0004,
+      "step": 10934
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.349001461276181e-07,
+      "loss": 0.0003,
+      "step": 10935
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3487579152459813e-07,
+      "loss": 0.0003,
+      "step": 10936
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3485143692157816e-07,
+      "loss": 0.0002,
+      "step": 10937
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3482708231855818e-07,
+      "loss": 0.0003,
+      "step": 10938
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.348027277155382e-07,
+      "loss": 0.0003,
+      "step": 10939
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3477837311251826e-07,
+      "loss": 0.0004,
+      "step": 10940
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3475401850949828e-07,
+      "loss": 2.4463,
+      "step": 10941
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.347296639064783e-07,
+      "loss": 0.0005,
+      "step": 10942
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 2.3470530930345833e-07,
+      "loss": 0.0003,
+      "step": 10943
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3468095470043836e-07,
+      "loss": 0.0004,
+      "step": 10944
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3465660009741838e-07,
+      "loss": 0.0003,
+      "step": 10945
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.346322454943984e-07,
+      "loss": 0.0003,
+      "step": 10946
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3460789089137846e-07,
+      "loss": 0.0004,
+      "step": 10947
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.345835362883585e-07,
+      "loss": 0.001,
+      "step": 10948
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3455918168533854e-07,
+      "loss": 0.0002,
+      "step": 10949
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3453482708231856e-07,
+      "loss": 0.0003,
+      "step": 10950
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.345104724792986e-07,
+      "loss": 0.0005,
+      "step": 10951
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.344861178762786e-07,
+      "loss": 0.0003,
+      "step": 10952
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 2.3446176327325864e-07,
+      "loss": 0.0003,
+      "step": 10953
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3443740867023866e-07,
+      "loss": 0.0006,
+      "step": 10954
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3441305406721871e-07,
+      "loss": 0.0004,
+      "step": 10955
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3438869946419874e-07,
+      "loss": 0.0002,
+      "step": 10956
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3436434486117877e-07,
+      "loss": 0.0005,
+      "step": 10957
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.343399902581588e-07,
+      "loss": 0.0006,
+      "step": 10958
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3431563565513882e-07,
+      "loss": 0.0003,
+      "step": 10959
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3429128105211884e-07,
+      "loss": 0.0003,
+      "step": 10960
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3426692644909887e-07,
+      "loss": 0.0003,
+      "step": 10961
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.342425718460789e-07,
+      "loss": 0.0004,
+      "step": 10962
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 2.3421821724305894e-07,
+      "loss": 0.0005,
+      "step": 10963
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3419386264003897e-07,
+      "loss": 0.0003,
+      "step": 10964
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.34169508037019e-07,
+      "loss": 0.0003,
+      "step": 10965
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3414515343399902e-07,
+      "loss": 0.0003,
+      "step": 10966
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3412079883097904e-07,
+      "loss": 0.0003,
+      "step": 10967
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3409644422795907e-07,
+      "loss": 0.0003,
+      "step": 10968
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.340720896249391e-07,
+      "loss": 0.0004,
+      "step": 10969
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3404773502191912e-07,
+      "loss": 0.0003,
+      "step": 10970
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3402338041889917e-07,
+      "loss": 0.0004,
+      "step": 10971
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.339990258158792e-07,
+      "loss": 0.0003,
+      "step": 10972
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3397467121285922e-07,
+      "loss": 0.0004,
+      "step": 10973
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 2.3395031660983925e-07,
+      "loss": 0.0005,
+      "step": 10974
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3392596200681927e-07,
+      "loss": 0.0004,
+      "step": 10975
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.339016074037993e-07,
+      "loss": 0.0004,
+      "step": 10976
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3387725280077932e-07,
+      "loss": 0.0007,
+      "step": 10977
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3385289819775935e-07,
+      "loss": 0.015,
+      "step": 10978
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.338285435947394e-07,
+      "loss": 0.0005,
+      "step": 10979
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3380418899171942e-07,
+      "loss": 0.0004,
+      "step": 10980
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3377983438869945e-07,
+      "loss": 0.0005,
+      "step": 10981
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3375547978567947e-07,
+      "loss": 0.0004,
+      "step": 10982
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.337311251826595e-07,
+      "loss": 0.0005,
+      "step": 10983
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 2.3370677057963952e-07,
+      "loss": 0.0008,
+      "step": 10984
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3368241597661955e-07,
+      "loss": 0.0005,
+      "step": 10985
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.336580613735996e-07,
+      "loss": 0.0006,
+      "step": 10986
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3363370677057965e-07,
+      "loss": 0.0005,
+      "step": 10987
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3360935216755968e-07,
+      "loss": 0.0005,
+      "step": 10988
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.335849975645397e-07,
+      "loss": 0.0003,
+      "step": 10989
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3356064296151973e-07,
+      "loss": 0.0003,
+      "step": 10990
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3353628835849975e-07,
+      "loss": 0.0004,
+      "step": 10991
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3351193375547978e-07,
+      "loss": 0.0003,
+      "step": 10992
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.334875791524598e-07,
+      "loss": 0.0002,
+      "step": 10993
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 2.3346322454943983e-07,
+      "loss": 0.0003,
+      "step": 10994
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3343886994641988e-07,
+      "loss": 0.0005,
+      "step": 10995
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.334145153433999e-07,
+      "loss": 0.0002,
+      "step": 10996
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3339016074037993e-07,
+      "loss": 0.0004,
+      "step": 10997
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3336580613735996e-07,
+      "loss": 0.0003,
+      "step": 10998
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3334145153433998e-07,
+      "loss": 0.0003,
+      "step": 10999
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3331709693132e-07,
+      "loss": 0.0005,
+      "step": 11000
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3329274232830003e-07,
+      "loss": 0.0004,
+      "step": 11001
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3326838772528006e-07,
+      "loss": 0.0013,
+      "step": 11002
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.332440331222601e-07,
+      "loss": 0.0003,
+      "step": 11003
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3321967851924013e-07,
+      "loss": 0.0004,
+      "step": 11004
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 2.3319532391622016e-07,
+      "loss": 0.0008,
+      "step": 11005
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3317096931320018e-07,
+      "loss": 0.0006,
+      "step": 11006
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.331466147101802e-07,
+      "loss": 0.0003,
+      "step": 11007
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3312226010716023e-07,
+      "loss": 0.0006,
+      "step": 11008
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3309790550414026e-07,
+      "loss": 0.0004,
+      "step": 11009
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3307355090112028e-07,
+      "loss": 0.0004,
+      "step": 11010
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3304919629810034e-07,
+      "loss": 0.0003,
+      "step": 11011
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3302484169508036e-07,
+      "loss": 0.0003,
+      "step": 11012
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3300048709206039e-07,
+      "loss": 0.0005,
+      "step": 11013
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.329761324890404e-07,
+      "loss": 0.0003,
+      "step": 11014
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 2.3295177788602044e-07,
+      "loss": 0.0008,
+      "step": 11015
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3292742328300046e-07,
+      "loss": 0.0005,
+      "step": 11016
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.329030686799805e-07,
+      "loss": 0.0004,
+      "step": 11017
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.328787140769605e-07,
+      "loss": 0.0004,
+      "step": 11018
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3285435947394056e-07,
+      "loss": 0.0004,
+      "step": 11019
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.328300048709206e-07,
+      "loss": 0.0005,
+      "step": 11020
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3280565026790061e-07,
+      "loss": 0.0003,
+      "step": 11021
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3278129566488064e-07,
+      "loss": 0.0005,
+      "step": 11022
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3275694106186066e-07,
+      "loss": 0.0002,
+      "step": 11023
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3273258645884072e-07,
+      "loss": 0.0004,
+      "step": 11024
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 2.3270823185582074e-07,
+      "loss": 0.0003,
+      "step": 11025
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.326838772528008e-07,
+      "loss": 0.0003,
+      "step": 11026
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3265952264978082e-07,
+      "loss": 0.0003,
+      "step": 11027
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3263516804676084e-07,
+      "loss": 0.0004,
+      "step": 11028
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3261081344374087e-07,
+      "loss": 0.0003,
+      "step": 11029
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.325864588407209e-07,
+      "loss": 0.0004,
+      "step": 11030
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3256210423770092e-07,
+      "loss": 0.0003,
+      "step": 11031
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3253774963468094e-07,
+      "loss": 0.0003,
+      "step": 11032
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3251339503166097e-07,
+      "loss": 0.0003,
+      "step": 11033
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3248904042864102e-07,
+      "loss": 0.0006,
+      "step": 11034
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3246468582562105e-07,
+      "loss": 0.0003,
+      "step": 11035
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 2.3244033122260107e-07,
+      "loss": 0.0004,
+      "step": 11036
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.324159766195811e-07,
+      "loss": 0.0004,
+      "step": 11037
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3239162201656112e-07,
+      "loss": 0.0004,
+      "step": 11038
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3236726741354115e-07,
+      "loss": 0.0004,
+      "step": 11039
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3234291281052117e-07,
+      "loss": 0.0005,
+      "step": 11040
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.323185582075012e-07,
+      "loss": 0.0003,
+      "step": 11041
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3229420360448125e-07,
+      "loss": 0.0002,
+      "step": 11042
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3226984900146127e-07,
+      "loss": 0.0006,
+      "step": 11043
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.322454943984413e-07,
+      "loss": 0.0003,
+      "step": 11044
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3222113979542132e-07,
+      "loss": 0.0002,
+      "step": 11045
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 2.3219678519240135e-07,
+      "loss": 0.0005,
+      "step": 11046
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3217243058938137e-07,
+      "loss": 0.0003,
+      "step": 11047
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.321480759863614e-07,
+      "loss": 0.0003,
+      "step": 11048
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3212372138334142e-07,
+      "loss": 0.0004,
+      "step": 11049
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3209936678032148e-07,
+      "loss": 0.0002,
+      "step": 11050
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.320750121773015e-07,
+      "loss": 0.0004,
+      "step": 11051
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3205065757428153e-07,
+      "loss": 0.0021,
+      "step": 11052
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3202630297126155e-07,
+      "loss": 0.0006,
+      "step": 11053
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3200194836824158e-07,
+      "loss": 0.0005,
+      "step": 11054
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.319775937652216e-07,
+      "loss": 0.0003,
+      "step": 11055
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 2.3195323916220163e-07,
+      "loss": 0.0003,
+      "step": 11056
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3192888455918165e-07,
+      "loss": 0.0002,
+      "step": 11057
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.319045299561617e-07,
+      "loss": 0.0013,
+      "step": 11058
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3188017535314173e-07,
+      "loss": 0.0002,
+      "step": 11059
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3185582075012175e-07,
+      "loss": 0.0003,
+      "step": 11060
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3183146614710178e-07,
+      "loss": 0.0005,
+      "step": 11061
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3180711154408183e-07,
+      "loss": 0.0004,
+      "step": 11062
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3178275694106186e-07,
+      "loss": 0.0004,
+      "step": 11063
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3175840233804188e-07,
+      "loss": 0.0004,
+      "step": 11064
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3173404773502193e-07,
+      "loss": 0.0003,
+      "step": 11065
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 2.3170969313200196e-07,
+      "loss": 0.0002,
+      "step": 11066
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3168533852898198e-07,
+      "loss": 0.0004,
+      "step": 11067
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.31660983925962e-07,
+      "loss": 0.0003,
+      "step": 11068
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3163662932294203e-07,
+      "loss": 0.0003,
+      "step": 11069
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3161227471992206e-07,
+      "loss": 0.0003,
+      "step": 11070
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3158792011690208e-07,
+      "loss": 0.0002,
+      "step": 11071
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.315635655138821e-07,
+      "loss": 0.0003,
+      "step": 11072
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3153921091086216e-07,
+      "loss": 0.0003,
+      "step": 11073
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3151485630784219e-07,
+      "loss": 0.0005,
+      "step": 11074
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.314905017048222e-07,
+      "loss": 0.0004,
+      "step": 11075
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3146614710180224e-07,
+      "loss": 0.0003,
+      "step": 11076
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 2.3144179249878226e-07,
+      "loss": 0.0004,
+      "step": 11077
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3141743789576229e-07,
+      "loss": 0.0005,
+      "step": 11078
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.313930832927423e-07,
+      "loss": 0.0003,
+      "step": 11079
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3136872868972234e-07,
+      "loss": 0.0004,
+      "step": 11080
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.313443740867024e-07,
+      "loss": 0.0006,
+      "step": 11081
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3132001948368241e-07,
+      "loss": 0.0002,
+      "step": 11082
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3129566488066244e-07,
+      "loss": 0.0002,
+      "step": 11083
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3127131027764246e-07,
+      "loss": 0.0004,
+      "step": 11084
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.312469556746225e-07,
+      "loss": 0.0002,
+      "step": 11085
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3122260107160251e-07,
+      "loss": 0.0003,
+      "step": 11086
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 2.3119824646858254e-07,
+      "loss": 0.0003,
+      "step": 11087
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3117389186556256e-07,
+      "loss": 0.0002,
+      "step": 11088
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3114953726254262e-07,
+      "loss": 0.0003,
+      "step": 11089
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3112518265952264e-07,
+      "loss": 0.0002,
+      "step": 11090
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3110082805650267e-07,
+      "loss": 0.0003,
+      "step": 11091
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.310764734534827e-07,
+      "loss": 0.0004,
+      "step": 11092
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3105211885046272e-07,
+      "loss": 0.0004,
+      "step": 11093
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3102776424744274e-07,
+      "loss": 0.0004,
+      "step": 11094
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3100340964442277e-07,
+      "loss": 0.0002,
+      "step": 11095
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.309790550414028e-07,
+      "loss": 0.0003,
+      "step": 11096
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 2.3095470043838284e-07,
+      "loss": 0.0005,
+      "step": 11097
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3093034583536287e-07,
+      "loss": 0.0002,
+      "step": 11098
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.309059912323429e-07,
+      "loss": 0.0002,
+      "step": 11099
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3088163662932292e-07,
+      "loss": 0.0003,
+      "step": 11100
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3085728202630297e-07,
+      "loss": 0.0004,
+      "step": 11101
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.30832927423283e-07,
+      "loss": 0.0005,
+      "step": 11102
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3080857282026302e-07,
+      "loss": 0.0004,
+      "step": 11103
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3078421821724305e-07,
+      "loss": 0.0003,
+      "step": 11104
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.307598636142231e-07,
+      "loss": 0.0004,
+      "step": 11105
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3073550901120312e-07,
+      "loss": 0.0005,
+      "step": 11106
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3071115440818315e-07,
+      "loss": 0.0007,
+      "step": 11107
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 2.3068679980516317e-07,
+      "loss": 0.0003,
+      "step": 11108
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.306624452021432e-07,
+      "loss": 0.0003,
+      "step": 11109
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3063809059912322e-07,
+      "loss": 0.0005,
+      "step": 11110
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3061373599610325e-07,
+      "loss": 0.0004,
+      "step": 11111
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3058938139308327e-07,
+      "loss": 0.0003,
+      "step": 11112
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3056502679006333e-07,
+      "loss": 0.0006,
+      "step": 11113
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3054067218704335e-07,
+      "loss": 0.0004,
+      "step": 11114
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3051631758402338e-07,
+      "loss": 0.0002,
+      "step": 11115
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.304919629810034e-07,
+      "loss": 0.0004,
+      "step": 11116
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3046760837798343e-07,
+      "loss": 0.0004,
+      "step": 11117
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 2.3044325377496345e-07,
+      "loss": 0.0005,
+      "step": 11118
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3041889917194348e-07,
+      "loss": 0.0002,
+      "step": 11119
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.303945445689235e-07,
+      "loss": 0.0007,
+      "step": 11120
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3037018996590355e-07,
+      "loss": 0.0003,
+      "step": 11121
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3034583536288358e-07,
+      "loss": 0.0005,
+      "step": 11122
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.303214807598636e-07,
+      "loss": 0.0003,
+      "step": 11123
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3029712615684363e-07,
+      "loss": 0.0003,
+      "step": 11124
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3027277155382365e-07,
+      "loss": 0.0004,
+      "step": 11125
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3024841695080368e-07,
+      "loss": 0.0004,
+      "step": 11126
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.302240623477837e-07,
+      "loss": 0.0003,
+      "step": 11127
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 2.3019970774476376e-07,
+      "loss": 0.0003,
+      "step": 11128
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3017535314174378e-07,
+      "loss": 0.0003,
+      "step": 11129
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.301509985387238e-07,
+      "loss": 0.0003,
+      "step": 11130
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3012664393570383e-07,
+      "loss": 0.0005,
+      "step": 11131
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3010228933268386e-07,
+      "loss": 0.0005,
+      "step": 11132
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3007793472966388e-07,
+      "loss": 0.0003,
+      "step": 11133
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.300535801266439e-07,
+      "loss": 0.0004,
+      "step": 11134
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3002922552362393e-07,
+      "loss": 0.0004,
+      "step": 11135
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.3000487092060398e-07,
+      "loss": 0.0003,
+      "step": 11136
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.29980516317584e-07,
+      "loss": 0.0003,
+      "step": 11137
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 2.2995616171456403e-07,
+      "loss": 0.0003,
+      "step": 11138
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.2993180711154409e-07,
+      "loss": 0.0006,
+      "step": 11139
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.299074525085241e-07,
+      "loss": 0.0003,
+      "step": 11140
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.2988309790550414e-07,
+      "loss": 0.2341,
+      "step": 11141
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.2985874330248416e-07,
+      "loss": 0.0005,
+      "step": 11142
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.298343886994642e-07,
+      "loss": 0.0004,
+      "step": 11143
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.2981003409644424e-07,
+      "loss": 0.0003,
+      "step": 11144
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.2978567949342426e-07,
+      "loss": 0.0003,
+      "step": 11145
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.297613248904043e-07,
+      "loss": 0.0003,
+      "step": 11146
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.2973697028738431e-07,
+      "loss": 0.0003,
+      "step": 11147
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.2971261568436434e-07,
+      "loss": 0.0003,
+      "step": 11148
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 2.2968826108134436e-07,
+      "loss": 0.0004,
+      "step": 11149
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.296639064783244e-07,
+      "loss": 0.0004,
+      "step": 11150
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.2963955187530441e-07,
+      "loss": 0.0004,
+      "step": 11151
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.2961519727228447e-07,
+      "loss": 0.0004,
+      "step": 11152
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.295908426692645e-07,
+      "loss": 0.0005,
+      "step": 11153
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.2956648806624452e-07,
+      "loss": 0.0006,
+      "step": 11154
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.2954213346322454e-07,
+      "loss": 0.0003,
+      "step": 11155
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.2951777886020457e-07,
+      "loss": 0.0003,
+      "step": 11156
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.294934242571846e-07,
+      "loss": 0.0005,
+      "step": 11157
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.2946906965416462e-07,
+      "loss": 0.0003,
+      "step": 11158
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 2.2944471505114464e-07,
+      "loss": 0.0006,
+      "step": 11159
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.294203604481247e-07,
+      "loss": 0.0003,
+      "step": 11160
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.2939600584510472e-07,
+      "loss": 0.0003,
+      "step": 11161
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.2937165124208474e-07,
+      "loss": 0.0003,
+      "step": 11162
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.2934729663906477e-07,
+      "loss": 0.0004,
+      "step": 11163
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.293229420360448e-07,
+      "loss": 0.0004,
+      "step": 11164
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.2929858743302482e-07,
+      "loss": 0.0004,
+      "step": 11165
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.2927423283000485e-07,
+      "loss": 0.0003,
+      "step": 11166
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.2924987822698487e-07,
+      "loss": 0.0004,
+      "step": 11167
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.2922552362396492e-07,
+      "loss": 0.0005,
+      "step": 11168
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 2.2920116902094495e-07,
+      "loss": 0.0005,
+      "step": 11169
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.2917681441792497e-07,
+      "loss": 0.0004,
+      "step": 11170
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.29152459814905e-07,
+      "loss": 0.0003,
+      "step": 11171
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.2912810521188502e-07,
+      "loss": 0.0004,
+      "step": 11172
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.2910375060886505e-07,
+      "loss": 0.0003,
+      "step": 11173
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.2907939600584507e-07,
+      "loss": 0.0005,
+      "step": 11174
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.290550414028251e-07,
+      "loss": 0.0005,
+      "step": 11175
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.2903068679980515e-07,
+      "loss": 0.0004,
+      "step": 11176
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.290063321967852e-07,
+      "loss": 0.0003,
+      "step": 11177
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.2898197759376523e-07,
+      "loss": 0.0003,
+      "step": 11178
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.2895762299074525e-07,
+      "loss": 0.0003,
+      "step": 11179
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 2.2893326838772528e-07,
+      "loss": 0.0005,
+      "step": 11180
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.289089137847053e-07,
+      "loss": 0.0006,
+      "step": 11181
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.2888455918168533e-07,
+      "loss": 0.0004,
+      "step": 11182
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.2886020457866538e-07,
+      "loss": 0.0004,
+      "step": 11183
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.288358499756454e-07,
+      "loss": 0.0007,
+      "step": 11184
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.2881149537262543e-07,
+      "loss": 0.0004,
+      "step": 11185
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.2878714076960545e-07,
+      "loss": 0.0003,
+      "step": 11186
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.2876278616658548e-07,
+      "loss": 0.0004,
+      "step": 11187
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.287384315635655e-07,
+      "loss": 1.2222,
+      "step": 11188
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.2871407696054553e-07,
+      "loss": 0.0003,
+      "step": 11189
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 2.2868972235752556e-07,
+      "loss": 0.0003,
+      "step": 11190
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.286653677545056e-07,
+      "loss": 0.0003,
+      "step": 11191
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.2864101315148563e-07,
+      "loss": 0.0003,
+      "step": 11192
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.2861665854846566e-07,
+      "loss": 0.0004,
+      "step": 11193
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.2859230394544568e-07,
+      "loss": 0.0003,
+      "step": 11194
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.285679493424257e-07,
+      "loss": 0.0005,
+      "step": 11195
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.2854359473940573e-07,
+      "loss": 0.0004,
+      "step": 11196
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.2851924013638576e-07,
+      "loss": 0.0003,
+      "step": 11197
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.2849488553336578e-07,
+      "loss": 0.0005,
+      "step": 11198
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.2847053093034583e-07,
+      "loss": 0.0003,
+      "step": 11199
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 2.2844617632732586e-07,
+      "loss": 0.0004,
+      "step": 11200
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.2842182172430588e-07,
+      "loss": 0.0003,
+      "step": 11201
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.283974671212859e-07,
+      "loss": 0.0003,
+      "step": 11202
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.2837311251826594e-07,
+      "loss": 0.0003,
+      "step": 11203
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.2834875791524596e-07,
+      "loss": 0.0004,
+      "step": 11204
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.2832440331222599e-07,
+      "loss": 0.0002,
+      "step": 11205
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.28300048709206e-07,
+      "loss": 0.0006,
+      "step": 11206
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.2827569410618606e-07,
+      "loss": 0.0006,
+      "step": 11207
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.282513395031661e-07,
+      "loss": 0.0003,
+      "step": 11208
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.282269849001461e-07,
+      "loss": 0.0003,
+      "step": 11209
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 2.2820263029712614e-07,
+      "loss": 0.0003,
+      "step": 11210
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.2817827569410616e-07,
+      "loss": 0.0003,
+      "step": 11211
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.281539210910862e-07,
+      "loss": 0.0003,
+      "step": 11212
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.2812956648806621e-07,
+      "loss": 0.0003,
+      "step": 11213
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.2810521188504624e-07,
+      "loss": 0.0002,
+      "step": 11214
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.280808572820263e-07,
+      "loss": 0.0004,
+      "step": 11215
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.2805650267900634e-07,
+      "loss": 0.0004,
+      "step": 11216
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.2803214807598637e-07,
+      "loss": 0.0004,
+      "step": 11217
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.280077934729664e-07,
+      "loss": 0.0003,
+      "step": 11218
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.2798343886994642e-07,
+      "loss": 0.0003,
+      "step": 11219
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.2795908426692644e-07,
+      "loss": 0.0006,
+      "step": 11220
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 2.2793472966390647e-07,
+      "loss": 0.0004,
+      "step": 11221
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.2791037506088652e-07,
+      "loss": 0.0004,
+      "step": 11222
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.2788602045786654e-07,
+      "loss": 0.0004,
+      "step": 11223
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.2786166585484657e-07,
+      "loss": 0.0003,
+      "step": 11224
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.278373112518266e-07,
+      "loss": 0.0004,
+      "step": 11225
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.2781295664880662e-07,
+      "loss": 0.1602,
+      "step": 11226
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.2778860204578665e-07,
+      "loss": 0.0003,
+      "step": 11227
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.2776424744276667e-07,
+      "loss": 0.0003,
+      "step": 11228
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.277398928397467e-07,
+      "loss": 0.0004,
+      "step": 11229
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.2771553823672675e-07,
+      "loss": 0.0006,
+      "step": 11230
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 2.2769118363370677e-07,
+      "loss": 0.0003,
+      "step": 11231
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.276668290306868e-07,
+      "loss": 0.0004,
+      "step": 11232
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.2764247442766682e-07,
+      "loss": 0.0004,
+      "step": 11233
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.2761811982464685e-07,
+      "loss": 0.0003,
+      "step": 11234
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.2759376522162687e-07,
+      "loss": 0.0002,
+      "step": 11235
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.275694106186069e-07,
+      "loss": 0.0005,
+      "step": 11236
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.2754505601558692e-07,
+      "loss": 0.0018,
+      "step": 11237
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.2752070141256697e-07,
+      "loss": 0.0006,
+      "step": 11238
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.27496346809547e-07,
+      "loss": 0.0004,
+      "step": 11239
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.2747199220652703e-07,
+      "loss": 0.0003,
+      "step": 11240
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 2.2744763760350705e-07,
+      "loss": 0.0003,
+      "step": 11241
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.2742328300048708e-07,
+      "loss": 0.0003,
+      "step": 11242
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.273989283974671e-07,
+      "loss": 0.0534,
+      "step": 11243
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.2737457379444713e-07,
+      "loss": 0.0005,
+      "step": 11244
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.2735021919142715e-07,
+      "loss": 0.0004,
+      "step": 11245
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.273258645884072e-07,
+      "loss": 0.0002,
+      "step": 11246
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.2730150998538723e-07,
+      "loss": 0.0003,
+      "step": 11247
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.2727715538236725e-07,
+      "loss": 0.0004,
+      "step": 11248
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.2725280077934728e-07,
+      "loss": 0.0003,
+      "step": 11249
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.272284461763273e-07,
+      "loss": 0.0002,
+      "step": 11250
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.2720409157330733e-07,
+      "loss": 0.0003,
+      "step": 11251
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 2.2717973697028735e-07,
+      "loss": 0.0002,
+      "step": 11252
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.2715538236726738e-07,
+      "loss": 0.0003,
+      "step": 11253
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.2713102776424746e-07,
+      "loss": 0.0003,
+      "step": 11254
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.2710667316122748e-07,
+      "loss": 0.0004,
+      "step": 11255
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.270823185582075e-07,
+      "loss": 0.0006,
+      "step": 11256
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.2705796395518753e-07,
+      "loss": 0.0005,
+      "step": 11257
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.2703360935216756e-07,
+      "loss": 0.0005,
+      "step": 11258
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.2700925474914758e-07,
+      "loss": 0.0021,
+      "step": 11259
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.269849001461276e-07,
+      "loss": 0.0003,
+      "step": 11260
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.2696054554310763e-07,
+      "loss": 0.0005,
+      "step": 11261
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 2.2693619094008768e-07,
+      "loss": 0.0004,
+      "step": 11262
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.269118363370677e-07,
+      "loss": 0.0003,
+      "step": 11263
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.2688748173404774e-07,
+      "loss": 0.0004,
+      "step": 11264
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.2686312713102776e-07,
+      "loss": 0.0005,
+      "step": 11265
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.2683877252800779e-07,
+      "loss": 0.0003,
+      "step": 11266
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.268144179249878e-07,
+      "loss": 0.0005,
+      "step": 11267
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.2679006332196784e-07,
+      "loss": 0.0004,
+      "step": 11268
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.2676570871894786e-07,
+      "loss": 0.0004,
+      "step": 11269
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.267413541159279e-07,
+      "loss": 0.0003,
+      "step": 11270
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.2671699951290794e-07,
+      "loss": 0.0003,
+      "step": 11271
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 2.2669264490988796e-07,
+      "loss": 0.0002,
+      "step": 11272
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.26668290306868e-07,
+      "loss": 0.0004,
+      "step": 11273
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.26643935703848e-07,
+      "loss": 0.0004,
+      "step": 11274
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.2661958110082804e-07,
+      "loss": 0.0003,
+      "step": 11275
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.2659522649780806e-07,
+      "loss": 0.0003,
+      "step": 11276
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.265708718947881e-07,
+      "loss": 0.0002,
+      "step": 11277
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.2654651729176814e-07,
+      "loss": 0.0003,
+      "step": 11278
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.2652216268874817e-07,
+      "loss": 0.0006,
+      "step": 11279
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.264978080857282e-07,
+      "loss": 0.0004,
+      "step": 11280
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.2647345348270822e-07,
+      "loss": 0.0003,
+      "step": 11281
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 2.2644909887968824e-07,
+      "loss": 0.0004,
+      "step": 11282
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.2642474427666827e-07,
+      "loss": 0.0002,
+      "step": 11283
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.264003896736483e-07,
+      "loss": 0.0003,
+      "step": 11284
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.2637603507062832e-07,
+      "loss": 0.0003,
+      "step": 11285
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.2635168046760837e-07,
+      "loss": 0.0003,
+      "step": 11286
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.263273258645884e-07,
+      "loss": 0.0002,
+      "step": 11287
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.2630297126156842e-07,
+      "loss": 0.0004,
+      "step": 11288
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.2627861665854844e-07,
+      "loss": 0.0004,
+      "step": 11289
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.2625426205552847e-07,
+      "loss": 0.0003,
+      "step": 11290
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.262299074525085e-07,
+      "loss": 0.0007,
+      "step": 11291
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.2620555284948855e-07,
+      "loss": 0.0002,
+      "step": 11292
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 2.261811982464686e-07,
+      "loss": 0.0003,
+      "step": 11293
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2615684364344862e-07,
+      "loss": 0.0003,
+      "step": 11294
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2613248904042865e-07,
+      "loss": 0.0003,
+      "step": 11295
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2610813443740867e-07,
+      "loss": 0.0004,
+      "step": 11296
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.260837798343887e-07,
+      "loss": 0.5617,
+      "step": 11297
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2605942523136872e-07,
+      "loss": 0.0002,
+      "step": 11298
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2603507062834875e-07,
+      "loss": 0.0004,
+      "step": 11299
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2601071602532877e-07,
+      "loss": 0.0002,
+      "step": 11300
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2598636142230883e-07,
+      "loss": 0.0002,
+      "step": 11301
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2596200681928885e-07,
+      "loss": 2.5304,
+      "step": 11302
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 2.2593765221626888e-07,
+      "loss": 0.0003,
+      "step": 11303
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.259132976132489e-07,
+      "loss": 0.0003,
+      "step": 11304
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.2588894301022893e-07,
+      "loss": 0.0003,
+      "step": 11305
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.2586458840720895e-07,
+      "loss": 0.0003,
+      "step": 11306
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.2584023380418898e-07,
+      "loss": 0.0004,
+      "step": 11307
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.25815879201169e-07,
+      "loss": 0.0003,
+      "step": 11308
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.2579152459814905e-07,
+      "loss": 0.0005,
+      "step": 11309
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.2576716999512908e-07,
+      "loss": 0.0003,
+      "step": 11310
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.257428153921091e-07,
+      "loss": 0.0002,
+      "step": 11311
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.2571846078908913e-07,
+      "loss": 0.0004,
+      "step": 11312
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 2.2569410618606915e-07,
+      "loss": 0.0004,
+      "step": 11313
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.2566975158304918e-07,
+      "loss": 0.0003,
+      "step": 11314
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.256453969800292e-07,
+      "loss": 0.0003,
+      "step": 11315
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.2562104237700923e-07,
+      "loss": 0.0002,
+      "step": 11316
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.2559668777398928e-07,
+      "loss": 0.0006,
+      "step": 11317
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.255723331709693e-07,
+      "loss": 0.0003,
+      "step": 11318
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.2554797856794933e-07,
+      "loss": 0.0003,
+      "step": 11319
+    },
+    {
+      "epoch": 11.0,
+      "eval_accuracy": 0.9985415653864852,
+      "eval_loss": 0.01568285934627056,
+      "eval_runtime": 2.8161,
+      "eval_samples_per_second": 730.437,
+      "eval_steps_per_second": 11.718,
+      "step": 11319
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.2552362396492936e-07,
+      "loss": 0.0004,
+      "step": 11320
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.2549926936190938e-07,
+      "loss": 0.0003,
+      "step": 11321
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.254749147588894e-07,
+      "loss": 0.0004,
+      "step": 11322
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.2545056015586943e-07,
+      "loss": 0.0003,
+      "step": 11323
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 2.2542620555284946e-07,
+      "loss": 0.0003,
+      "step": 11324
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.254018509498295e-07,
+      "loss": 0.0005,
+      "step": 11325
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.2537749634680953e-07,
+      "loss": 0.0003,
+      "step": 11326
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.2535314174378956e-07,
+      "loss": 0.0003,
+      "step": 11327
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.2532878714076958e-07,
+      "loss": 0.0003,
+      "step": 11328
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.253044325377496e-07,
+      "loss": 0.0003,
+      "step": 11329
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.2528007793472963e-07,
+      "loss": 0.0003,
+      "step": 11330
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.2525572333170969e-07,
+      "loss": 0.0004,
+      "step": 11331
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.2523136872868974e-07,
+      "loss": 0.0006,
+      "step": 11332
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.2520701412566976e-07,
+      "loss": 0.0004,
+      "step": 11333
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 2.251826595226498e-07,
+      "loss": 0.0004,
+      "step": 11334
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.251583049196298e-07,
+      "loss": 0.0003,
+      "step": 11335
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.2513395031660984e-07,
+      "loss": 0.0004,
+      "step": 11336
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.2510959571358986e-07,
+      "loss": 0.0002,
+      "step": 11337
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.250852411105699e-07,
+      "loss": 0.0003,
+      "step": 11338
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.2506088650754991e-07,
+      "loss": 0.0002,
+      "step": 11339
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.2503653190452997e-07,
+      "loss": 0.0003,
+      "step": 11340
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.2501217730151e-07,
+      "loss": 0.0003,
+      "step": 11341
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.2498782269849002e-07,
+      "loss": 0.0004,
+      "step": 11342
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.2496346809547004e-07,
+      "loss": 0.0014,
+      "step": 11343
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 2.2493911349245007e-07,
+      "loss": 0.0003,
+      "step": 11344
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.249147588894301e-07,
+      "loss": 0.0005,
+      "step": 11345
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.2489040428641012e-07,
+      "loss": 0.0003,
+      "step": 11346
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.2486604968339014e-07,
+      "loss": 0.0002,
+      "step": 11347
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.248416950803702e-07,
+      "loss": 0.0003,
+      "step": 11348
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.2481734047735022e-07,
+      "loss": 0.0004,
+      "step": 11349
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.2479298587433024e-07,
+      "loss": 0.0005,
+      "step": 11350
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.2476863127131027e-07,
+      "loss": 0.0002,
+      "step": 11351
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.247442766682903e-07,
+      "loss": 0.0005,
+      "step": 11352
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.2471992206527032e-07,
+      "loss": 0.0003,
+      "step": 11353
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.2469556746225034e-07,
+      "loss": 0.0004,
+      "step": 11354
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 2.2467121285923037e-07,
+      "loss": 0.0004,
+      "step": 11355
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.2464685825621042e-07,
+      "loss": 0.0005,
+      "step": 11356
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.2462250365319045e-07,
+      "loss": 0.0003,
+      "step": 11357
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.2459814905017047e-07,
+      "loss": 0.0009,
+      "step": 11358
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.245737944471505e-07,
+      "loss": 0.0003,
+      "step": 11359
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.2454943984413052e-07,
+      "loss": 0.0003,
+      "step": 11360
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.2452508524111055e-07,
+      "loss": 0.0005,
+      "step": 11361
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.2450073063809057e-07,
+      "loss": 0.0004,
+      "step": 11362
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.244763760350706e-07,
+      "loss": 0.0003,
+      "step": 11363
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.2445202143205065e-07,
+      "loss": 0.0004,
+      "step": 11364
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 2.2442766682903067e-07,
+      "loss": 0.0002,
+      "step": 11365
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.244033122260107e-07,
+      "loss": 0.0003,
+      "step": 11366
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.2437895762299072e-07,
+      "loss": 0.0004,
+      "step": 11367
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.2435460301997075e-07,
+      "loss": 0.0004,
+      "step": 11368
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.243302484169508e-07,
+      "loss": 0.0003,
+      "step": 11369
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.2430589381393083e-07,
+      "loss": 0.0003,
+      "step": 11370
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.2428153921091085e-07,
+      "loss": 0.0004,
+      "step": 11371
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.242571846078909e-07,
+      "loss": 0.0005,
+      "step": 11372
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.2423283000487093e-07,
+      "loss": 0.0002,
+      "step": 11373
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.2420847540185095e-07,
+      "loss": 0.0003,
+      "step": 11374
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 2.2418412079883098e-07,
+      "loss": 0.0003,
+      "step": 11375
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.24159766195811e-07,
+      "loss": 0.0003,
+      "step": 11376
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.2413541159279103e-07,
+      "loss": 0.0003,
+      "step": 11377
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.2411105698977105e-07,
+      "loss": 0.0003,
+      "step": 11378
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.2408670238675108e-07,
+      "loss": 0.0003,
+      "step": 11379
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.2406234778373113e-07,
+      "loss": 0.0003,
+      "step": 11380
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.2403799318071116e-07,
+      "loss": 0.0005,
+      "step": 11381
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.2401363857769118e-07,
+      "loss": 0.0003,
+      "step": 11382
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.239892839746712e-07,
+      "loss": 0.0004,
+      "step": 11383
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.2396492937165123e-07,
+      "loss": 0.0003,
+      "step": 11384
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 2.2394057476863126e-07,
+      "loss": 0.0003,
+      "step": 11385
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.2391622016561128e-07,
+      "loss": 0.0004,
+      "step": 11386
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.238918655625913e-07,
+      "loss": 0.0003,
+      "step": 11387
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.2386751095957136e-07,
+      "loss": 0.0002,
+      "step": 11388
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.2384315635655138e-07,
+      "loss": 0.0004,
+      "step": 11389
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.238188017535314e-07,
+      "loss": 0.0003,
+      "step": 11390
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.2379444715051143e-07,
+      "loss": 0.0005,
+      "step": 11391
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.2377009254749146e-07,
+      "loss": 0.0002,
+      "step": 11392
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.2374573794447148e-07,
+      "loss": 0.0002,
+      "step": 11393
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.237213833414515e-07,
+      "loss": 0.0004,
+      "step": 11394
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.2369702873843156e-07,
+      "loss": 0.0005,
+      "step": 11395
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 2.2367267413541159e-07,
+      "loss": 0.0003,
+      "step": 11396
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.236483195323916e-07,
+      "loss": 0.0004,
+      "step": 11397
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.2362396492937164e-07,
+      "loss": 0.0004,
+      "step": 11398
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.2359961032635166e-07,
+      "loss": 0.0002,
+      "step": 11399
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.235752557233317e-07,
+      "loss": 0.0003,
+      "step": 11400
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.235509011203117e-07,
+      "loss": 0.0003,
+      "step": 11401
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.2352654651729174e-07,
+      "loss": 0.0004,
+      "step": 11402
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.235021919142718e-07,
+      "loss": 0.0002,
+      "step": 11403
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.2347783731125181e-07,
+      "loss": 0.0004,
+      "step": 11404
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.2345348270823184e-07,
+      "loss": 0.0002,
+      "step": 11405
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 2.2342912810521186e-07,
+      "loss": 0.0003,
+      "step": 11406
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.2340477350219192e-07,
+      "loss": 0.0005,
+      "step": 11407
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.2338041889917194e-07,
+      "loss": 0.0003,
+      "step": 11408
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.2335606429615197e-07,
+      "loss": 0.0004,
+      "step": 11409
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.23331709693132e-07,
+      "loss": 0.0002,
+      "step": 11410
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.2330735509011204e-07,
+      "loss": 0.0003,
+      "step": 11411
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.2328300048709207e-07,
+      "loss": 0.0003,
+      "step": 11412
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.232586458840721e-07,
+      "loss": 0.0003,
+      "step": 11413
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.2323429128105212e-07,
+      "loss": 0.0003,
+      "step": 11414
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.2320993667803214e-07,
+      "loss": 0.0004,
+      "step": 11415
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 2.2318558207501217e-07,
+      "loss": 0.0003,
+      "step": 11416
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.231612274719922e-07,
+      "loss": 2.2223,
+      "step": 11417
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.2313687286897222e-07,
+      "loss": 0.0003,
+      "step": 11418
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.2311251826595227e-07,
+      "loss": 0.0003,
+      "step": 11419
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.230881636629323e-07,
+      "loss": 0.0003,
+      "step": 11420
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.2306380905991232e-07,
+      "loss": 0.0003,
+      "step": 11421
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.2303945445689235e-07,
+      "loss": 0.0004,
+      "step": 11422
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.2301509985387237e-07,
+      "loss": 0.0004,
+      "step": 11423
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.229907452508524e-07,
+      "loss": 0.0005,
+      "step": 11424
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.2296639064783242e-07,
+      "loss": 0.0002,
+      "step": 11425
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.2294203604481245e-07,
+      "loss": 0.0003,
+      "step": 11426
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 2.229176814417925e-07,
+      "loss": 0.0002,
+      "step": 11427
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2289332683877252e-07,
+      "loss": 0.0019,
+      "step": 11428
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2286897223575255e-07,
+      "loss": 0.0004,
+      "step": 11429
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2284461763273257e-07,
+      "loss": 0.0003,
+      "step": 11430
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.228202630297126e-07,
+      "loss": 0.0003,
+      "step": 11431
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2279590842669262e-07,
+      "loss": 0.0004,
+      "step": 11432
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2277155382367265e-07,
+      "loss": 0.0004,
+      "step": 11433
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2274719922065267e-07,
+      "loss": 0.0003,
+      "step": 11434
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2272284461763273e-07,
+      "loss": 0.0003,
+      "step": 11435
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2269849001461275e-07,
+      "loss": 0.0003,
+      "step": 11436
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 2.2267413541159278e-07,
+      "loss": 0.0003,
+      "step": 11437
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.226497808085728e-07,
+      "loss": 0.0003,
+      "step": 11438
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.2262542620555283e-07,
+      "loss": 0.0004,
+      "step": 11439
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.2260107160253285e-07,
+      "loss": 0.0003,
+      "step": 11440
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.2257671699951288e-07,
+      "loss": 0.0003,
+      "step": 11441
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.225523623964929e-07,
+      "loss": 0.0005,
+      "step": 11442
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.2252800779347295e-07,
+      "loss": 0.0005,
+      "step": 11443
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.2250365319045298e-07,
+      "loss": 0.0003,
+      "step": 11444
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.2247929858743303e-07,
+      "loss": 0.0003,
+      "step": 11445
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.2245494398441306e-07,
+      "loss": 0.0005,
+      "step": 11446
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 2.2243058938139308e-07,
+      "loss": 0.0005,
+      "step": 11447
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.224062347783731e-07,
+      "loss": 0.0004,
+      "step": 11448
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.2238188017535313e-07,
+      "loss": 0.0003,
+      "step": 11449
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.2235752557233318e-07,
+      "loss": 0.0002,
+      "step": 11450
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.223331709693132e-07,
+      "loss": 0.0003,
+      "step": 11451
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.2230881636629323e-07,
+      "loss": 0.0005,
+      "step": 11452
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.2228446176327326e-07,
+      "loss": 0.0004,
+      "step": 11453
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.2226010716025328e-07,
+      "loss": 0.0003,
+      "step": 11454
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.222357525572333e-07,
+      "loss": 0.0003,
+      "step": 11455
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.2221139795421333e-07,
+      "loss": 0.0005,
+      "step": 11456
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 2.2218704335119336e-07,
+      "loss": 0.0009,
+      "step": 11457
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.221626887481734e-07,
+      "loss": 0.0003,
+      "step": 11458
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.2213833414515344e-07,
+      "loss": 0.0003,
+      "step": 11459
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.2211397954213346e-07,
+      "loss": 0.0005,
+      "step": 11460
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.2208962493911349e-07,
+      "loss": 0.0006,
+      "step": 11461
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.220652703360935e-07,
+      "loss": 0.0006,
+      "step": 11462
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.2204091573307354e-07,
+      "loss": 0.0004,
+      "step": 11463
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.2201656113005356e-07,
+      "loss": 0.0004,
+      "step": 11464
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.219922065270336e-07,
+      "loss": 0.0004,
+      "step": 11465
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.2196785192401364e-07,
+      "loss": 0.0003,
+      "step": 11466
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.2194349732099366e-07,
+      "loss": 0.0005,
+      "step": 11467
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 2.219191427179737e-07,
+      "loss": 0.2348,
+      "step": 11468
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.2189478811495371e-07,
+      "loss": 0.0004,
+      "step": 11469
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.2187043351193374e-07,
+      "loss": 0.0005,
+      "step": 11470
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.2184607890891376e-07,
+      "loss": 0.0003,
+      "step": 11471
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.218217243058938e-07,
+      "loss": 0.0006,
+      "step": 11472
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.2179736970287382e-07,
+      "loss": 0.0003,
+      "step": 11473
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.2177301509985387e-07,
+      "loss": 0.0003,
+      "step": 11474
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.217486604968339e-07,
+      "loss": 0.0003,
+      "step": 11475
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.2172430589381392e-07,
+      "loss": 0.0003,
+      "step": 11476
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.2169995129079394e-07,
+      "loss": 0.0005,
+      "step": 11477
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 2.2167559668777397e-07,
+      "loss": 0.0003,
+      "step": 11478
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.21651242084754e-07,
+      "loss": 0.0004,
+      "step": 11479
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.2162688748173402e-07,
+      "loss": 0.0002,
+      "step": 11480
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.2160253287871404e-07,
+      "loss": 0.0003,
+      "step": 11481
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.215781782756941e-07,
+      "loss": 0.0004,
+      "step": 11482
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.2155382367267412e-07,
+      "loss": 0.0004,
+      "step": 11483
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.2152946906965417e-07,
+      "loss": 0.0004,
+      "step": 11484
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.215051144666342e-07,
+      "loss": 0.0003,
+      "step": 11485
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.2148075986361422e-07,
+      "loss": 0.0004,
+      "step": 11486
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.2145640526059425e-07,
+      "loss": 0.0005,
+      "step": 11487
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 2.2143205065757427e-07,
+      "loss": 0.0002,
+      "step": 11488
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.214076960545543e-07,
+      "loss": 0.0002,
+      "step": 11489
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.2138334145153435e-07,
+      "loss": 0.0006,
+      "step": 11490
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.2135898684851437e-07,
+      "loss": 0.0002,
+      "step": 11491
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.213346322454944e-07,
+      "loss": 0.0003,
+      "step": 11492
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.2131027764247442e-07,
+      "loss": 0.0002,
+      "step": 11493
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.2128592303945445e-07,
+      "loss": 0.0003,
+      "step": 11494
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.2126156843643447e-07,
+      "loss": 0.0008,
+      "step": 11495
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.212372138334145e-07,
+      "loss": 0.0005,
+      "step": 11496
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.2121285923039455e-07,
+      "loss": 0.0003,
+      "step": 11497
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.2118850462737458e-07,
+      "loss": 0.0003,
+      "step": 11498
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 2.211641500243546e-07,
+      "loss": 0.0003,
+      "step": 11499
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.2113979542133463e-07,
+      "loss": 0.0002,
+      "step": 11500
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.2111544081831465e-07,
+      "loss": 0.0003,
+      "step": 11501
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.2109108621529468e-07,
+      "loss": 0.0003,
+      "step": 11502
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.210667316122747e-07,
+      "loss": 0.0004,
+      "step": 11503
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.2104237700925473e-07,
+      "loss": 0.0003,
+      "step": 11504
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.2101802240623478e-07,
+      "loss": 0.0004,
+      "step": 11505
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.209936678032148e-07,
+      "loss": 0.0004,
+      "step": 11506
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.2096931320019483e-07,
+      "loss": 0.0004,
+      "step": 11507
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.2094495859717485e-07,
+      "loss": 0.2879,
+      "step": 11508
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 2.2092060399415488e-07,
+      "loss": 0.0004,
+      "step": 11509
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.208962493911349e-07,
+      "loss": 0.0003,
+      "step": 11510
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.2087189478811493e-07,
+      "loss": 0.0012,
+      "step": 11511
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.2084754018509496e-07,
+      "loss": 0.0003,
+      "step": 11512
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.20823185582075e-07,
+      "loss": 0.0002,
+      "step": 11513
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.2079883097905503e-07,
+      "loss": 0.0003,
+      "step": 11514
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.2077447637603506e-07,
+      "loss": 0.0003,
+      "step": 11515
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.2075012177301508e-07,
+      "loss": 0.0003,
+      "step": 11516
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.207257671699951e-07,
+      "loss": 0.0003,
+      "step": 11517
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.2070141256697513e-07,
+      "loss": 0.0006,
+      "step": 11518
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 2.2067705796395516e-07,
+      "loss": 0.0005,
+      "step": 11519
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.2065270336093518e-07,
+      "loss": 0.0004,
+      "step": 11520
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.2062834875791523e-07,
+      "loss": 0.0003,
+      "step": 11521
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.2060399415489529e-07,
+      "loss": 0.0003,
+      "step": 11522
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.205796395518753e-07,
+      "loss": 0.0004,
+      "step": 11523
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.2055528494885534e-07,
+      "loss": 0.0004,
+      "step": 11524
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.2053093034583536e-07,
+      "loss": 0.0007,
+      "step": 11525
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.205065757428154e-07,
+      "loss": 0.0003,
+      "step": 11526
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.204822211397954e-07,
+      "loss": 0.0004,
+      "step": 11527
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.2045786653677544e-07,
+      "loss": 0.0016,
+      "step": 11528
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 2.204335119337555e-07,
+      "loss": 0.0004,
+      "step": 11529
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2040915733073551e-07,
+      "loss": 0.0003,
+      "step": 11530
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2038480272771554e-07,
+      "loss": 0.0003,
+      "step": 11531
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2036044812469556e-07,
+      "loss": 0.0002,
+      "step": 11532
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.203360935216756e-07,
+      "loss": 0.0003,
+      "step": 11533
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2031173891865561e-07,
+      "loss": 0.0004,
+      "step": 11534
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2028738431563564e-07,
+      "loss": 0.0002,
+      "step": 11535
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2026302971261567e-07,
+      "loss": 0.0002,
+      "step": 11536
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2023867510959572e-07,
+      "loss": 0.0003,
+      "step": 11537
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2021432050657574e-07,
+      "loss": 0.0003,
+      "step": 11538
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.2018996590355577e-07,
+      "loss": 0.0005,
+      "step": 11539
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 2.201656113005358e-07,
+      "loss": 0.0004,
+      "step": 11540
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.2014125669751582e-07,
+      "loss": 0.0003,
+      "step": 11541
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.2011690209449584e-07,
+      "loss": 0.0002,
+      "step": 11542
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.2009254749147587e-07,
+      "loss": 0.0003,
+      "step": 11543
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.200681928884559e-07,
+      "loss": 0.0006,
+      "step": 11544
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.2004383828543594e-07,
+      "loss": 0.0004,
+      "step": 11545
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.2001948368241597e-07,
+      "loss": 0.0004,
+      "step": 11546
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.19995129079396e-07,
+      "loss": 0.0003,
+      "step": 11547
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.1997077447637602e-07,
+      "loss": 0.0002,
+      "step": 11548
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.1994641987335605e-07,
+      "loss": 0.0003,
+      "step": 11549
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 2.1992206527033607e-07,
+      "loss": 0.0002,
+      "step": 11550
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.198977106673161e-07,
+      "loss": 0.0003,
+      "step": 11551
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.1987335606429612e-07,
+      "loss": 0.0003,
+      "step": 11552
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.1984900146127617e-07,
+      "loss": 0.0005,
+      "step": 11553
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.198246468582562e-07,
+      "loss": 0.0003,
+      "step": 11554
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.1980029225523622e-07,
+      "loss": 0.0004,
+      "step": 11555
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.1977593765221625e-07,
+      "loss": 0.0003,
+      "step": 11556
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.1975158304919627e-07,
+      "loss": 0.0003,
+      "step": 11557
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.197272284461763e-07,
+      "loss": 0.0003,
+      "step": 11558
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.1970287384315632e-07,
+      "loss": 0.0004,
+      "step": 11559
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 2.196785192401364e-07,
+      "loss": 0.0006,
+      "step": 11560
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1965416463711643e-07,
+      "loss": 0.0004,
+      "step": 11561
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1962981003409645e-07,
+      "loss": 0.0005,
+      "step": 11562
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1960545543107648e-07,
+      "loss": 0.0004,
+      "step": 11563
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.195811008280565e-07,
+      "loss": 0.0002,
+      "step": 11564
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1955674622503653e-07,
+      "loss": 0.0004,
+      "step": 11565
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1953239162201655e-07,
+      "loss": 0.0005,
+      "step": 11566
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1950803701899658e-07,
+      "loss": 0.0002,
+      "step": 11567
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1948368241597663e-07,
+      "loss": 0.0003,
+      "step": 11568
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1945932781295665e-07,
+      "loss": 0.0003,
+      "step": 11569
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.1943497320993668e-07,
+      "loss": 0.0004,
+      "step": 11570
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 2.194106186069167e-07,
+      "loss": 0.0004,
+      "step": 11571
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.1938626400389673e-07,
+      "loss": 0.001,
+      "step": 11572
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.1936190940087676e-07,
+      "loss": 0.0002,
+      "step": 11573
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.1933755479785678e-07,
+      "loss": 0.0002,
+      "step": 11574
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.193132001948368e-07,
+      "loss": 0.0002,
+      "step": 11575
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.1928884559181686e-07,
+      "loss": 0.0003,
+      "step": 11576
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.1926449098879688e-07,
+      "loss": 0.0003,
+      "step": 11577
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.192401363857769e-07,
+      "loss": 0.0003,
+      "step": 11578
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.1921578178275693e-07,
+      "loss": 0.0005,
+      "step": 11579
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.1919142717973696e-07,
+      "loss": 0.0004,
+      "step": 11580
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 2.1916707257671698e-07,
+      "loss": 0.0003,
+      "step": 11581
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.19142717973697e-07,
+      "loss": 0.0002,
+      "step": 11582
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.1911836337067703e-07,
+      "loss": 0.0002,
+      "step": 11583
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.1909400876765709e-07,
+      "loss": 0.0004,
+      "step": 11584
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.190696541646371e-07,
+      "loss": 0.0004,
+      "step": 11585
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.1904529956161714e-07,
+      "loss": 0.0002,
+      "step": 11586
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.1902094495859716e-07,
+      "loss": 0.0003,
+      "step": 11587
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.1899659035557719e-07,
+      "loss": 0.0003,
+      "step": 11588
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.189722357525572e-07,
+      "loss": 0.0002,
+      "step": 11589
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.1894788114953724e-07,
+      "loss": 0.0004,
+      "step": 11590
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 2.1892352654651726e-07,
+      "loss": 0.0002,
+      "step": 11591
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.188991719434973e-07,
+      "loss": 0.0003,
+      "step": 11592
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.1887481734047734e-07,
+      "loss": 0.0005,
+      "step": 11593
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.1885046273745736e-07,
+      "loss": 0.0003,
+      "step": 11594
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.188261081344374e-07,
+      "loss": 0.0012,
+      "step": 11595
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.1880175353141741e-07,
+      "loss": 0.0005,
+      "step": 11596
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.1877739892839744e-07,
+      "loss": 0.0004,
+      "step": 11597
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.1875304432537746e-07,
+      "loss": 0.0004,
+      "step": 11598
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.1872868972235754e-07,
+      "loss": 0.0002,
+      "step": 11599
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.1870433511933757e-07,
+      "loss": 0.0003,
+      "step": 11600
+    },
+    {
+      "epoch": 11.27,
+      "learning_rate": 2.186799805163176e-07,
+      "loss": 0.0002,
+      "step": 11601
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.1865562591329762e-07,
+      "loss": 0.7813,
+      "step": 11602
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.1863127131027764e-07,
+      "loss": 0.0004,
+      "step": 11603
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.1860691670725767e-07,
+      "loss": 0.0187,
+      "step": 11604
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.185825621042377e-07,
+      "loss": 0.0003,
+      "step": 11605
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.1855820750121772e-07,
+      "loss": 0.0002,
+      "step": 11606
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.1853385289819777e-07,
+      "loss": 0.0003,
+      "step": 11607
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.185094982951778e-07,
+      "loss": 0.0003,
+      "step": 11608
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.1848514369215782e-07,
+      "loss": 0.0005,
+      "step": 11609
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.1846078908913785e-07,
+      "loss": 0.0003,
+      "step": 11610
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.1843643448611787e-07,
+      "loss": 0.0002,
+      "step": 11611
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 2.184120798830979e-07,
+      "loss": 0.0003,
+      "step": 11612
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.1838772528007792e-07,
+      "loss": 0.0003,
+      "step": 11613
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.1836337067705795e-07,
+      "loss": 0.0005,
+      "step": 11614
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.18339016074038e-07,
+      "loss": 0.0002,
+      "step": 11615
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.1831466147101802e-07,
+      "loss": 0.0003,
+      "step": 11616
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.1829030686799805e-07,
+      "loss": 0.0003,
+      "step": 11617
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.1826595226497807e-07,
+      "loss": 0.0002,
+      "step": 11618
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.182415976619581e-07,
+      "loss": 0.0004,
+      "step": 11619
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.1821724305893812e-07,
+      "loss": 0.0003,
+      "step": 11620
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.1819288845591815e-07,
+      "loss": 0.0005,
+      "step": 11621
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 2.1816853385289817e-07,
+      "loss": 0.0003,
+      "step": 11622
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.1814417924987823e-07,
+      "loss": 0.0003,
+      "step": 11623
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.1811982464685825e-07,
+      "loss": 0.0005,
+      "step": 11624
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.1809547004383828e-07,
+      "loss": 0.0003,
+      "step": 11625
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.180711154408183e-07,
+      "loss": 0.0003,
+      "step": 11626
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.1804676083779833e-07,
+      "loss": 0.0003,
+      "step": 11627
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.1802240623477835e-07,
+      "loss": 0.0003,
+      "step": 11628
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.1799805163175838e-07,
+      "loss": 0.0003,
+      "step": 11629
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.179736970287384e-07,
+      "loss": 0.0005,
+      "step": 11630
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.1794934242571845e-07,
+      "loss": 0.0003,
+      "step": 11631
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 2.1792498782269848e-07,
+      "loss": 0.0006,
+      "step": 11632
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.179006332196785e-07,
+      "loss": 0.0006,
+      "step": 11633
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.1787627861665853e-07,
+      "loss": 0.0003,
+      "step": 11634
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.1785192401363855e-07,
+      "loss": 0.0004,
+      "step": 11635
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.1782756941061858e-07,
+      "loss": 0.0003,
+      "step": 11636
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.1780321480759863e-07,
+      "loss": 0.0003,
+      "step": 11637
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.1777886020457866e-07,
+      "loss": 0.0004,
+      "step": 11638
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.177545056015587e-07,
+      "loss": 0.0004,
+      "step": 11639
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.1773015099853873e-07,
+      "loss": 0.0003,
+      "step": 11640
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.1770579639551876e-07,
+      "loss": 0.0003,
+      "step": 11641
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.1768144179249878e-07,
+      "loss": 0.0003,
+      "step": 11642
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 2.176570871894788e-07,
+      "loss": 0.0004,
+      "step": 11643
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1763273258645883e-07,
+      "loss": 0.0004,
+      "step": 11644
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1760837798343886e-07,
+      "loss": 0.0005,
+      "step": 11645
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1758402338041888e-07,
+      "loss": 0.2828,
+      "step": 11646
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1755966877739894e-07,
+      "loss": 0.0003,
+      "step": 11647
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1753531417437896e-07,
+      "loss": 0.0004,
+      "step": 11648
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1751095957135899e-07,
+      "loss": 0.0005,
+      "step": 11649
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.17486604968339e-07,
+      "loss": 0.0003,
+      "step": 11650
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1746225036531904e-07,
+      "loss": 0.0006,
+      "step": 11651
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1743789576229906e-07,
+      "loss": 0.0003,
+      "step": 11652
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 2.1741354115927909e-07,
+      "loss": 0.0003,
+      "step": 11653
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.173891865562591e-07,
+      "loss": 0.0002,
+      "step": 11654
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.1736483195323916e-07,
+      "loss": 0.0003,
+      "step": 11655
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.173404773502192e-07,
+      "loss": 0.0003,
+      "step": 11656
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.1731612274719921e-07,
+      "loss": 0.0004,
+      "step": 11657
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.1729176814417924e-07,
+      "loss": 0.0002,
+      "step": 11658
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.1726741354115926e-07,
+      "loss": 0.0003,
+      "step": 11659
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.172430589381393e-07,
+      "loss": 0.0002,
+      "step": 11660
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.1721870433511931e-07,
+      "loss": 0.0013,
+      "step": 11661
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.1719434973209934e-07,
+      "loss": 0.0003,
+      "step": 11662
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 2.171699951290794e-07,
+      "loss": 0.0003,
+      "step": 11663
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.1714564052605942e-07,
+      "loss": 0.0003,
+      "step": 11664
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.1712128592303944e-07,
+      "loss": 0.0004,
+      "step": 11665
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.1709693132001947e-07,
+      "loss": 0.0003,
+      "step": 11666
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.170725767169995e-07,
+      "loss": 0.0002,
+      "step": 11667
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.1704822211397952e-07,
+      "loss": 0.0004,
+      "step": 11668
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.1702386751095954e-07,
+      "loss": 0.0008,
+      "step": 11669
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.169995129079396e-07,
+      "loss": 0.0003,
+      "step": 11670
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.1697515830491962e-07,
+      "loss": 0.0004,
+      "step": 11671
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.1695080370189964e-07,
+      "loss": 0.0002,
+      "step": 11672
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.1692644909887967e-07,
+      "loss": 0.0003,
+      "step": 11673
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 2.169020944958597e-07,
+      "loss": 0.0002,
+      "step": 11674
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.1687773989283975e-07,
+      "loss": 0.0431,
+      "step": 11675
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.1685338528981977e-07,
+      "loss": 0.0003,
+      "step": 11676
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.168290306867998e-07,
+      "loss": 0.0005,
+      "step": 11677
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.1680467608377985e-07,
+      "loss": 0.0004,
+      "step": 11678
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.1678032148075987e-07,
+      "loss": 0.0004,
+      "step": 11679
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.167559668777399e-07,
+      "loss": 0.0004,
+      "step": 11680
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.1673161227471992e-07,
+      "loss": 0.0004,
+      "step": 11681
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.1670725767169995e-07,
+      "loss": 0.0002,
+      "step": 11682
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.1668290306867997e-07,
+      "loss": 0.0004,
+      "step": 11683
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 2.1665854846566e-07,
+      "loss": 0.0004,
+      "step": 11684
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.1663419386264002e-07,
+      "loss": 0.0003,
+      "step": 11685
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.1660983925962008e-07,
+      "loss": 0.0004,
+      "step": 11686
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.165854846566001e-07,
+      "loss": 0.0002,
+      "step": 11687
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.1656113005358013e-07,
+      "loss": 0.0002,
+      "step": 11688
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.1653677545056015e-07,
+      "loss": 0.0005,
+      "step": 11689
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.1651242084754018e-07,
+      "loss": 0.0003,
+      "step": 11690
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.164880662445202e-07,
+      "loss": 0.0003,
+      "step": 11691
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.1646371164150023e-07,
+      "loss": 0.0012,
+      "step": 11692
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.1643935703848025e-07,
+      "loss": 0.0003,
+      "step": 11693
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 2.164150024354603e-07,
+      "loss": 0.0004,
+      "step": 11694
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1639064783244033e-07,
+      "loss": 0.0003,
+      "step": 11695
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1636629322942035e-07,
+      "loss": 0.0002,
+      "step": 11696
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1634193862640038e-07,
+      "loss": 0.0003,
+      "step": 11697
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.163175840233804e-07,
+      "loss": 0.0003,
+      "step": 11698
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1629322942036043e-07,
+      "loss": 0.0004,
+      "step": 11699
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1626887481734045e-07,
+      "loss": 0.0003,
+      "step": 11700
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1624452021432048e-07,
+      "loss": 0.0003,
+      "step": 11701
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1622016561130053e-07,
+      "loss": 0.0003,
+      "step": 11702
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1619581100828056e-07,
+      "loss": 0.0014,
+      "step": 11703
+    },
+    {
+      "epoch": 11.37,
+      "learning_rate": 2.1617145640526058e-07,
+      "loss": 0.0003,
+      "step": 11704
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.161471018022406e-07,
+      "loss": 0.0005,
+      "step": 11705
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.1612274719922063e-07,
+      "loss": 0.0004,
+      "step": 11706
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.1609839259620066e-07,
+      "loss": 0.0002,
+      "step": 11707
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.1607403799318068e-07,
+      "loss": 0.0006,
+      "step": 11708
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.160496833901607e-07,
+      "loss": 0.0003,
+      "step": 11709
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.1602532878714076e-07,
+      "loss": 0.0004,
+      "step": 11710
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.1600097418412078e-07,
+      "loss": 0.0002,
+      "step": 11711
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.159766195811008e-07,
+      "loss": 0.0003,
+      "step": 11712
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.1595226497808083e-07,
+      "loss": 0.0003,
+      "step": 11713
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.1592791037506089e-07,
+      "loss": 0.0005,
+      "step": 11714
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 2.159035557720409e-07,
+      "loss": 0.0003,
+      "step": 11715
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.1587920116902094e-07,
+      "loss": 0.0004,
+      "step": 11716
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.15854846566001e-07,
+      "loss": 0.0003,
+      "step": 11717
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.15830491962981e-07,
+      "loss": 0.0003,
+      "step": 11718
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.1580613735996104e-07,
+      "loss": 0.0002,
+      "step": 11719
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.1578178275694106e-07,
+      "loss": 0.0004,
+      "step": 11720
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.157574281539211e-07,
+      "loss": 0.0003,
+      "step": 11721
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.1573307355090111e-07,
+      "loss": 0.0002,
+      "step": 11722
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.1570871894788114e-07,
+      "loss": 0.0003,
+      "step": 11723
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.1568436434486116e-07,
+      "loss": 0.0004,
+      "step": 11724
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 2.1566000974184122e-07,
+      "loss": 0.0003,
+      "step": 11725
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.1563565513882124e-07,
+      "loss": 0.0002,
+      "step": 11726
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.1561130053580127e-07,
+      "loss": 0.0004,
+      "step": 11727
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.155869459327813e-07,
+      "loss": 0.0003,
+      "step": 11728
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.1556259132976132e-07,
+      "loss": 0.0003,
+      "step": 11729
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.1553823672674134e-07,
+      "loss": 0.0004,
+      "step": 11730
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.1551388212372137e-07,
+      "loss": 0.0002,
+      "step": 11731
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.154895275207014e-07,
+      "loss": 0.0002,
+      "step": 11732
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.1546517291768144e-07,
+      "loss": 0.0003,
+      "step": 11733
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.1544081831466147e-07,
+      "loss": 0.0003,
+      "step": 11734
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 2.154164637116415e-07,
+      "loss": 0.0003,
+      "step": 11735
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.1539210910862152e-07,
+      "loss": 0.0003,
+      "step": 11736
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.1536775450560154e-07,
+      "loss": 0.0004,
+      "step": 11737
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.1534339990258157e-07,
+      "loss": 0.0004,
+      "step": 11738
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.153190452995616e-07,
+      "loss": 0.0004,
+      "step": 11739
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.1529469069654162e-07,
+      "loss": 0.0004,
+      "step": 11740
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.1527033609352167e-07,
+      "loss": 0.0002,
+      "step": 11741
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.152459814905017e-07,
+      "loss": 0.0005,
+      "step": 11742
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.1522162688748172e-07,
+      "loss": 0.0003,
+      "step": 11743
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.1519727228446175e-07,
+      "loss": 0.0003,
+      "step": 11744
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.1517291768144177e-07,
+      "loss": 0.0002,
+      "step": 11745
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 2.151485630784218e-07,
+      "loss": 0.0003,
+      "step": 11746
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.1512420847540182e-07,
+      "loss": 0.0004,
+      "step": 11747
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.1509985387238185e-07,
+      "loss": 0.0004,
+      "step": 11748
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.150754992693619e-07,
+      "loss": 0.0003,
+      "step": 11749
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.1505114466634192e-07,
+      "loss": 0.0005,
+      "step": 11750
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.1502679006332195e-07,
+      "loss": 0.0004,
+      "step": 11751
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.15002435460302e-07,
+      "loss": 0.0003,
+      "step": 11752
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.1497808085728203e-07,
+      "loss": 0.0003,
+      "step": 11753
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.1495372625426205e-07,
+      "loss": 0.0003,
+      "step": 11754
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.1492937165124208e-07,
+      "loss": 0.0003,
+      "step": 11755
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 2.149050170482221e-07,
+      "loss": 0.0002,
+      "step": 11756
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.1488066244520215e-07,
+      "loss": 0.0005,
+      "step": 11757
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.1485630784218218e-07,
+      "loss": 0.0003,
+      "step": 11758
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.148319532391622e-07,
+      "loss": 0.0003,
+      "step": 11759
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.1480759863614223e-07,
+      "loss": 0.0004,
+      "step": 11760
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.1478324403312225e-07,
+      "loss": 0.0003,
+      "step": 11761
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.1475888943010228e-07,
+      "loss": 0.0003,
+      "step": 11762
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.147345348270823e-07,
+      "loss": 0.0004,
+      "step": 11763
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.1471018022406236e-07,
+      "loss": 0.0004,
+      "step": 11764
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.1468582562104238e-07,
+      "loss": 0.0004,
+      "step": 11765
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 2.146614710180224e-07,
+      "loss": 0.0003,
+      "step": 11766
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.1463711641500243e-07,
+      "loss": 1.6726,
+      "step": 11767
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.1461276181198246e-07,
+      "loss": 0.0043,
+      "step": 11768
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.1458840720896248e-07,
+      "loss": 0.0004,
+      "step": 11769
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.145640526059425e-07,
+      "loss": 0.0003,
+      "step": 11770
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.1453969800292253e-07,
+      "loss": 0.0004,
+      "step": 11771
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.1451534339990258e-07,
+      "loss": 0.0002,
+      "step": 11772
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.144909887968826e-07,
+      "loss": 0.0002,
+      "step": 11773
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.1446663419386263e-07,
+      "loss": 0.0003,
+      "step": 11774
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.1444227959084266e-07,
+      "loss": 0.0003,
+      "step": 11775
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 2.1441792498782268e-07,
+      "loss": 0.0003,
+      "step": 11776
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.143935703848027e-07,
+      "loss": 0.0003,
+      "step": 11777
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.1436921578178273e-07,
+      "loss": 0.0004,
+      "step": 11778
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.1434486117876276e-07,
+      "loss": 0.0004,
+      "step": 11779
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.143205065757428e-07,
+      "loss": 0.0003,
+      "step": 11780
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.1429615197272284e-07,
+      "loss": 0.0002,
+      "step": 11781
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.1427179736970286e-07,
+      "loss": 0.0003,
+      "step": 11782
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.142474427666829e-07,
+      "loss": 0.0003,
+      "step": 11783
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.142230881636629e-07,
+      "loss": 0.0004,
+      "step": 11784
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.1419873356064294e-07,
+      "loss": 0.0003,
+      "step": 11785
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.1417437895762296e-07,
+      "loss": 0.0003,
+      "step": 11786
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 2.14150024354603e-07,
+      "loss": 0.0003,
+      "step": 11787
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.1412566975158304e-07,
+      "loss": 0.0004,
+      "step": 11788
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.1410131514856306e-07,
+      "loss": 0.0003,
+      "step": 11789
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.1407696054554312e-07,
+      "loss": 0.0002,
+      "step": 11790
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.1405260594252314e-07,
+      "loss": 0.0004,
+      "step": 11791
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.1402825133950317e-07,
+      "loss": 0.0003,
+      "step": 11792
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.140038967364832e-07,
+      "loss": 0.0003,
+      "step": 11793
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.1397954213346322e-07,
+      "loss": 0.0005,
+      "step": 11794
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.1395518753044324e-07,
+      "loss": 0.0004,
+      "step": 11795
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.139308329274233e-07,
+      "loss": 0.0004,
+      "step": 11796
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 2.1390647832440332e-07,
+      "loss": 0.0005,
+      "step": 11797
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.1388212372138334e-07,
+      "loss": 0.0004,
+      "step": 11798
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.1385776911836337e-07,
+      "loss": 0.0003,
+      "step": 11799
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.138334145153434e-07,
+      "loss": 0.0004,
+      "step": 11800
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.1380905991232342e-07,
+      "loss": 0.0005,
+      "step": 11801
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.1378470530930344e-07,
+      "loss": 0.0003,
+      "step": 11802
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.1376035070628347e-07,
+      "loss": 0.0003,
+      "step": 11803
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.1373599610326352e-07,
+      "loss": 0.0004,
+      "step": 11804
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.1371164150024355e-07,
+      "loss": 0.0005,
+      "step": 11805
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.1368728689722357e-07,
+      "loss": 0.0002,
+      "step": 11806
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 2.136629322942036e-07,
+      "loss": 0.0005,
+      "step": 11807
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.1363857769118362e-07,
+      "loss": 0.0003,
+      "step": 11808
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.1361422308816365e-07,
+      "loss": 0.0002,
+      "step": 11809
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.1358986848514367e-07,
+      "loss": 0.0002,
+      "step": 11810
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.135655138821237e-07,
+      "loss": 0.0003,
+      "step": 11811
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.1354115927910375e-07,
+      "loss": 0.0002,
+      "step": 11812
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.1351680467608377e-07,
+      "loss": 0.0005,
+      "step": 11813
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.134924500730638e-07,
+      "loss": 0.0002,
+      "step": 11814
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.1346809547004382e-07,
+      "loss": 0.0002,
+      "step": 11815
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.1344374086702385e-07,
+      "loss": 0.0003,
+      "step": 11816
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.1341938626400387e-07,
+      "loss": 0.0003,
+      "step": 11817
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 2.133950316609839e-07,
+      "loss": 0.0004,
+      "step": 11818
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.1337067705796393e-07,
+      "loss": 0.0003,
+      "step": 11819
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.1334632245494398e-07,
+      "loss": 0.0003,
+      "step": 11820
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.13321967851924e-07,
+      "loss": 0.0003,
+      "step": 11821
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.1329761324890403e-07,
+      "loss": 0.0002,
+      "step": 11822
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.1327325864588405e-07,
+      "loss": 0.0003,
+      "step": 11823
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.1324890404286408e-07,
+      "loss": 0.0003,
+      "step": 11824
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.132245494398441e-07,
+      "loss": 0.0003,
+      "step": 11825
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.1320019483682413e-07,
+      "loss": 0.0004,
+      "step": 11826
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.1317584023380415e-07,
+      "loss": 0.0003,
+      "step": 11827
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 2.1315148563078423e-07,
+      "loss": 0.0003,
+      "step": 11828
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.1312713102776426e-07,
+      "loss": 0.0005,
+      "step": 11829
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.1310277642474428e-07,
+      "loss": 0.0005,
+      "step": 11830
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.130784218217243e-07,
+      "loss": 0.0002,
+      "step": 11831
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.1305406721870433e-07,
+      "loss": 0.0002,
+      "step": 11832
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.1302971261568436e-07,
+      "loss": 0.0004,
+      "step": 11833
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.1300535801266438e-07,
+      "loss": 0.0004,
+      "step": 11834
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.1298100340964443e-07,
+      "loss": 0.0004,
+      "step": 11835
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.1295664880662446e-07,
+      "loss": 0.0003,
+      "step": 11836
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.1293229420360448e-07,
+      "loss": 0.2357,
+      "step": 11837
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 2.129079396005845e-07,
+      "loss": 0.0003,
+      "step": 11838
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.1288358499756453e-07,
+      "loss": 0.0013,
+      "step": 11839
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.1285923039454456e-07,
+      "loss": 0.0003,
+      "step": 11840
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.1283487579152458e-07,
+      "loss": 0.0005,
+      "step": 11841
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.128105211885046e-07,
+      "loss": 0.0004,
+      "step": 11842
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.1278616658548466e-07,
+      "loss": 0.0002,
+      "step": 11843
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.127618119824647e-07,
+      "loss": 0.0006,
+      "step": 11844
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.127374573794447e-07,
+      "loss": 0.0003,
+      "step": 11845
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.1271310277642474e-07,
+      "loss": 0.0003,
+      "step": 11846
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.1268874817340476e-07,
+      "loss": 0.0002,
+      "step": 11847
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 2.126643935703848e-07,
+      "loss": 0.0003,
+      "step": 11848
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.126400389673648e-07,
+      "loss": 0.0004,
+      "step": 11849
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.1261568436434484e-07,
+      "loss": 0.0004,
+      "step": 11850
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.125913297613249e-07,
+      "loss": 0.0003,
+      "step": 11851
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.1256697515830491e-07,
+      "loss": 0.0005,
+      "step": 11852
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.1254262055528494e-07,
+      "loss": 0.0002,
+      "step": 11853
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.1251826595226496e-07,
+      "loss": 0.0002,
+      "step": 11854
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.12493911349245e-07,
+      "loss": 0.0003,
+      "step": 11855
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.1246955674622502e-07,
+      "loss": 0.0006,
+      "step": 11856
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.1244520214320504e-07,
+      "loss": 0.0005,
+      "step": 11857
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.1242084754018507e-07,
+      "loss": 0.0004,
+      "step": 11858
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 2.1239649293716512e-07,
+      "loss": 0.0004,
+      "step": 11859
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.1237213833414514e-07,
+      "loss": 0.0003,
+      "step": 11860
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.1234778373112517e-07,
+      "loss": 0.0003,
+      "step": 11861
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.123234291281052e-07,
+      "loss": 0.0003,
+      "step": 11862
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.1229907452508522e-07,
+      "loss": 0.0005,
+      "step": 11863
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.1227471992206524e-07,
+      "loss": 0.0002,
+      "step": 11864
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.1225036531904527e-07,
+      "loss": 0.0004,
+      "step": 11865
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.122260107160253e-07,
+      "loss": 0.0005,
+      "step": 11866
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.1220165611300537e-07,
+      "loss": 0.0002,
+      "step": 11867
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.121773015099854e-07,
+      "loss": 2.5919,
+      "step": 11868
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 2.1215294690696542e-07,
+      "loss": 0.0004,
+      "step": 11869
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.1212859230394545e-07,
+      "loss": 0.0005,
+      "step": 11870
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.1210423770092547e-07,
+      "loss": 0.0003,
+      "step": 11871
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.120798830979055e-07,
+      "loss": 0.0003,
+      "step": 11872
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.1205552849488552e-07,
+      "loss": 0.0004,
+      "step": 11873
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.1203117389186557e-07,
+      "loss": 0.0003,
+      "step": 11874
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.120068192888456e-07,
+      "loss": 0.0003,
+      "step": 11875
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.1198246468582562e-07,
+      "loss": 0.0005,
+      "step": 11876
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.1195811008280565e-07,
+      "loss": 0.0003,
+      "step": 11877
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.1193375547978567e-07,
+      "loss": 0.0005,
+      "step": 11878
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 2.119094008767657e-07,
+      "loss": 0.0004,
+      "step": 11879
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1188504627374573e-07,
+      "loss": 0.0004,
+      "step": 11880
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1186069167072575e-07,
+      "loss": 0.0003,
+      "step": 11881
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.118363370677058e-07,
+      "loss": 0.0002,
+      "step": 11882
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1181198246468583e-07,
+      "loss": 0.0002,
+      "step": 11883
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1178762786166585e-07,
+      "loss": 0.0004,
+      "step": 11884
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1176327325864588e-07,
+      "loss": 0.0003,
+      "step": 11885
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.117389186556259e-07,
+      "loss": 0.0004,
+      "step": 11886
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1171456405260593e-07,
+      "loss": 0.0002,
+      "step": 11887
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1169020944958595e-07,
+      "loss": 0.0003,
+      "step": 11888
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1166585484656598e-07,
+      "loss": 0.0004,
+      "step": 11889
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 2.1164150024354603e-07,
+      "loss": 0.0003,
+      "step": 11890
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.1161714564052605e-07,
+      "loss": 0.0003,
+      "step": 11891
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.1159279103750608e-07,
+      "loss": 0.0003,
+      "step": 11892
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.115684364344861e-07,
+      "loss": 0.0002,
+      "step": 11893
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.1154408183146613e-07,
+      "loss": 0.0004,
+      "step": 11894
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.1151972722844616e-07,
+      "loss": 0.0003,
+      "step": 11895
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.1149537262542618e-07,
+      "loss": 0.0003,
+      "step": 11896
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.114710180224062e-07,
+      "loss": 0.0004,
+      "step": 11897
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.1144666341938626e-07,
+      "loss": 0.0003,
+      "step": 11898
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.1142230881636628e-07,
+      "loss": 0.0002,
+      "step": 11899
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 2.113979542133463e-07,
+      "loss": 0.0003,
+      "step": 11900
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.1137359961032633e-07,
+      "loss": 0.0008,
+      "step": 11901
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.1134924500730636e-07,
+      "loss": 0.0002,
+      "step": 11902
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.1132489040428638e-07,
+      "loss": 0.0005,
+      "step": 11903
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.113005358012664e-07,
+      "loss": 0.0004,
+      "step": 11904
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.1127618119824646e-07,
+      "loss": 0.0002,
+      "step": 11905
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.112518265952265e-07,
+      "loss": 0.0002,
+      "step": 11906
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.1122747199220654e-07,
+      "loss": 0.0003,
+      "step": 11907
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.1120311738918656e-07,
+      "loss": 0.0004,
+      "step": 11908
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.111787627861666e-07,
+      "loss": 0.0003,
+      "step": 11909
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 2.111544081831466e-07,
+      "loss": 0.0003,
+      "step": 11910
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.1113005358012664e-07,
+      "loss": 0.0005,
+      "step": 11911
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.1110569897710666e-07,
+      "loss": 0.0003,
+      "step": 11912
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.110813443740867e-07,
+      "loss": 0.0004,
+      "step": 11913
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.1105698977106674e-07,
+      "loss": 0.0004,
+      "step": 11914
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.1103263516804676e-07,
+      "loss": 0.0004,
+      "step": 11915
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.110082805650268e-07,
+      "loss": 0.0005,
+      "step": 11916
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.1098392596200682e-07,
+      "loss": 0.0004,
+      "step": 11917
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.1095957135898684e-07,
+      "loss": 0.0002,
+      "step": 11918
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.1093521675596687e-07,
+      "loss": 0.0002,
+      "step": 11919
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 2.109108621529469e-07,
+      "loss": 0.0003,
+      "step": 11920
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.1088650754992692e-07,
+      "loss": 0.0004,
+      "step": 11921
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.1086215294690697e-07,
+      "loss": 0.0003,
+      "step": 11922
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.10837798343887e-07,
+      "loss": 0.0005,
+      "step": 11923
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.1081344374086702e-07,
+      "loss": 0.0003,
+      "step": 11924
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.1078908913784704e-07,
+      "loss": 0.0006,
+      "step": 11925
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.1076473453482707e-07,
+      "loss": 0.0003,
+      "step": 11926
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.107403799318071e-07,
+      "loss": 0.0004,
+      "step": 11927
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.1071602532878712e-07,
+      "loss": 0.0003,
+      "step": 11928
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.1069167072576714e-07,
+      "loss": 0.0002,
+      "step": 11929
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.106673161227472e-07,
+      "loss": 0.0004,
+      "step": 11930
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 2.1064296151972722e-07,
+      "loss": 0.0692,
+      "step": 11931
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.1061860691670725e-07,
+      "loss": 0.0002,
+      "step": 11932
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.1059425231368727e-07,
+      "loss": 0.0003,
+      "step": 11933
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.105698977106673e-07,
+      "loss": 0.0003,
+      "step": 11934
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.1054554310764732e-07,
+      "loss": 0.0005,
+      "step": 11935
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.1052118850462735e-07,
+      "loss": 0.0004,
+      "step": 11936
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.104968339016074e-07,
+      "loss": 0.0003,
+      "step": 11937
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.1047247929858742e-07,
+      "loss": 0.0004,
+      "step": 11938
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.1044812469556745e-07,
+      "loss": 0.0002,
+      "step": 11939
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.1042377009254747e-07,
+      "loss": 0.0002,
+      "step": 11940
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 2.103994154895275e-07,
+      "loss": 0.0003,
+      "step": 11941
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.1037506088650752e-07,
+      "loss": 0.0005,
+      "step": 11942
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.1035070628348758e-07,
+      "loss": 0.0004,
+      "step": 11943
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.103263516804676e-07,
+      "loss": 0.0003,
+      "step": 11944
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.1030199707744765e-07,
+      "loss": 0.0003,
+      "step": 11945
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.1027764247442768e-07,
+      "loss": 0.0002,
+      "step": 11946
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.102532878714077e-07,
+      "loss": 0.0003,
+      "step": 11947
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.1022893326838773e-07,
+      "loss": 0.0002,
+      "step": 11948
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.1020457866536775e-07,
+      "loss": 0.0029,
+      "step": 11949
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.1018022406234778e-07,
+      "loss": 0.0003,
+      "step": 11950
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 2.101558694593278e-07,
+      "loss": 0.0003,
+      "step": 11951
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.1013151485630783e-07,
+      "loss": 0.0003,
+      "step": 11952
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.1010716025328788e-07,
+      "loss": 0.0004,
+      "step": 11953
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.100828056502679e-07,
+      "loss": 0.0005,
+      "step": 11954
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.1005845104724793e-07,
+      "loss": 0.0003,
+      "step": 11955
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.1003409644422796e-07,
+      "loss": 0.0003,
+      "step": 11956
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.1000974184120798e-07,
+      "loss": 0.0004,
+      "step": 11957
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.09985387238188e-07,
+      "loss": 0.0001,
+      "step": 11958
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.0996103263516803e-07,
+      "loss": 0.0003,
+      "step": 11959
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.0993667803214806e-07,
+      "loss": 0.0004,
+      "step": 11960
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.099123234291281e-07,
+      "loss": 0.0002,
+      "step": 11961
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 2.0988796882610813e-07,
+      "loss": 0.0003,
+      "step": 11962
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.0986361422308816e-07,
+      "loss": 0.0004,
+      "step": 11963
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.0983925962006818e-07,
+      "loss": 0.0003,
+      "step": 11964
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.098149050170482e-07,
+      "loss": 0.0004,
+      "step": 11965
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.0979055041402823e-07,
+      "loss": 0.0003,
+      "step": 11966
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.0976619581100826e-07,
+      "loss": 0.0003,
+      "step": 11967
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.0974184120798828e-07,
+      "loss": 0.0004,
+      "step": 11968
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.0971748660496834e-07,
+      "loss": 0.0003,
+      "step": 11969
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.0969313200194836e-07,
+      "loss": 0.0001,
+      "step": 11970
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.0966877739892839e-07,
+      "loss": 0.0002,
+      "step": 11971
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 2.096444227959084e-07,
+      "loss": 0.0004,
+      "step": 11972
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.0962006819288844e-07,
+      "loss": 0.0003,
+      "step": 11973
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.0959571358986846e-07,
+      "loss": 0.0003,
+      "step": 11974
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.0957135898684849e-07,
+      "loss": 0.0002,
+      "step": 11975
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.095470043838285e-07,
+      "loss": 0.0002,
+      "step": 11976
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.0952264978080856e-07,
+      "loss": 0.0005,
+      "step": 11977
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.094982951777886e-07,
+      "loss": 0.0006,
+      "step": 11978
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.0947394057476861e-07,
+      "loss": 0.0003,
+      "step": 11979
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.0944958597174864e-07,
+      "loss": 0.0004,
+      "step": 11980
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.0942523136872866e-07,
+      "loss": 0.0002,
+      "step": 11981
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 2.0940087676570872e-07,
+      "loss": 0.0003,
+      "step": 11982
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.0937652216268874e-07,
+      "loss": 0.0004,
+      "step": 11983
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.093521675596688e-07,
+      "loss": 0.0004,
+      "step": 11984
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.0932781295664882e-07,
+      "loss": 0.0003,
+      "step": 11985
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.0930345835362884e-07,
+      "loss": 0.0003,
+      "step": 11986
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.0927910375060887e-07,
+      "loss": 0.0004,
+      "step": 11987
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.092547491475889e-07,
+      "loss": 0.0002,
+      "step": 11988
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.0923039454456892e-07,
+      "loss": 0.0003,
+      "step": 11989
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.0920603994154894e-07,
+      "loss": 0.0002,
+      "step": 11990
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.0918168533852897e-07,
+      "loss": 0.0002,
+      "step": 11991
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 2.0915733073550902e-07,
+      "loss": 0.0004,
+      "step": 11992
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.0913297613248905e-07,
+      "loss": 0.0003,
+      "step": 11993
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.0910862152946907e-07,
+      "loss": 0.0003,
+      "step": 11994
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.090842669264491e-07,
+      "loss": 0.0005,
+      "step": 11995
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.0905991232342912e-07,
+      "loss": 0.0004,
+      "step": 11996
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.0903555772040915e-07,
+      "loss": 0.0004,
+      "step": 11997
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.0901120311738917e-07,
+      "loss": 0.0002,
+      "step": 11998
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.089868485143692e-07,
+      "loss": 0.0003,
+      "step": 11999
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.0896249391134925e-07,
+      "loss": 0.0005,
+      "step": 12000
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.0893813930832927e-07,
+      "loss": 0.0004,
+      "step": 12001
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.089137847053093e-07,
+      "loss": 0.0004,
+      "step": 12002
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 2.0888943010228932e-07,
+      "loss": 0.0002,
+      "step": 12003
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.0886507549926935e-07,
+      "loss": 0.0003,
+      "step": 12004
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.0884072089624937e-07,
+      "loss": 0.0003,
+      "step": 12005
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.088163662932294e-07,
+      "loss": 0.0002,
+      "step": 12006
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.0879201169020942e-07,
+      "loss": 0.0002,
+      "step": 12007
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.0876765708718948e-07,
+      "loss": 0.0003,
+      "step": 12008
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.087433024841695e-07,
+      "loss": 0.0002,
+      "step": 12009
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.0871894788114953e-07,
+      "loss": 0.0003,
+      "step": 12010
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.0869459327812955e-07,
+      "loss": 0.0002,
+      "step": 12011
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.0867023867510958e-07,
+      "loss": 0.0003,
+      "step": 12012
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 2.086458840720896e-07,
+      "loss": 0.0003,
+      "step": 12013
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.0862152946906963e-07,
+      "loss": 0.0004,
+      "step": 12014
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.0859717486604965e-07,
+      "loss": 0.0002,
+      "step": 12015
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.085728202630297e-07,
+      "loss": 0.0002,
+      "step": 12016
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.0854846566000973e-07,
+      "loss": 0.0005,
+      "step": 12017
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.0852411105698975e-07,
+      "loss": 0.0002,
+      "step": 12018
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.0849975645396978e-07,
+      "loss": 0.0003,
+      "step": 12019
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.0847540185094983e-07,
+      "loss": 0.0003,
+      "step": 12020
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.0845104724792986e-07,
+      "loss": 0.0004,
+      "step": 12021
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.0842669264490988e-07,
+      "loss": 0.0002,
+      "step": 12022
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 2.084023380418899e-07,
+      "loss": 0.0005,
+      "step": 12023
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0837798343886996e-07,
+      "loss": 0.0003,
+      "step": 12024
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0835362883584998e-07,
+      "loss": 0.0004,
+      "step": 12025
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0832927423283e-07,
+      "loss": 0.0003,
+      "step": 12026
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0830491962981003e-07,
+      "loss": 0.0002,
+      "step": 12027
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0828056502679006e-07,
+      "loss": 0.0004,
+      "step": 12028
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0825621042377008e-07,
+      "loss": 0.0003,
+      "step": 12029
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.082318558207501e-07,
+      "loss": 0.0005,
+      "step": 12030
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0820750121773016e-07,
+      "loss": 0.0002,
+      "step": 12031
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0818314661471019e-07,
+      "loss": 0.0003,
+      "step": 12032
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.081587920116902e-07,
+      "loss": 0.0004,
+      "step": 12033
+    },
+    {
+      "epoch": 11.69,
+      "learning_rate": 2.0813443740867024e-07,
+      "loss": 0.0002,
+      "step": 12034
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.0811008280565026e-07,
+      "loss": 0.0004,
+      "step": 12035
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.0808572820263029e-07,
+      "loss": 0.0003,
+      "step": 12036
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.080613735996103e-07,
+      "loss": 0.0003,
+      "step": 12037
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.0803701899659034e-07,
+      "loss": 0.0002,
+      "step": 12038
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.080126643935704e-07,
+      "loss": 0.0005,
+      "step": 12039
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.0798830979055041e-07,
+      "loss": 0.0003,
+      "step": 12040
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.0796395518753044e-07,
+      "loss": 0.0003,
+      "step": 12041
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.0793960058451046e-07,
+      "loss": 0.0003,
+      "step": 12042
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.079152459814905e-07,
+      "loss": 0.0002,
+      "step": 12043
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 2.0789089137847051e-07,
+      "loss": 0.0004,
+      "step": 12044
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.0786653677545054e-07,
+      "loss": 0.0003,
+      "step": 12045
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.0784218217243056e-07,
+      "loss": 0.0003,
+      "step": 12046
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.0781782756941062e-07,
+      "loss": 0.0005,
+      "step": 12047
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.0779347296639064e-07,
+      "loss": 0.0002,
+      "step": 12048
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.0776911836337067e-07,
+      "loss": 0.0005,
+      "step": 12049
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.077447637603507e-07,
+      "loss": 0.0003,
+      "step": 12050
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.0772040915733072e-07,
+      "loss": 0.0003,
+      "step": 12051
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.0769605455431074e-07,
+      "loss": 0.0002,
+      "step": 12052
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.0767169995129077e-07,
+      "loss": 0.0005,
+      "step": 12053
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 2.076473453482708e-07,
+      "loss": 0.0002,
+      "step": 12054
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.0762299074525084e-07,
+      "loss": 0.0002,
+      "step": 12055
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.0759863614223087e-07,
+      "loss": 0.0004,
+      "step": 12056
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.075742815392109e-07,
+      "loss": 0.0002,
+      "step": 12057
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.0754992693619095e-07,
+      "loss": 0.0002,
+      "step": 12058
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.0752557233317097e-07,
+      "loss": 0.0006,
+      "step": 12059
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.07501217730151e-07,
+      "loss": 0.0004,
+      "step": 12060
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.0747686312713102e-07,
+      "loss": 0.0003,
+      "step": 12061
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.0745250852411105e-07,
+      "loss": 0.0003,
+      "step": 12062
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.074281539210911e-07,
+      "loss": 0.0002,
+      "step": 12063
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.0740379931807112e-07,
+      "loss": 0.0006,
+      "step": 12064
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 2.0737944471505115e-07,
+      "loss": 0.0004,
+      "step": 12065
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.0735509011203117e-07,
+      "loss": 0.0002,
+      "step": 12066
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.073307355090112e-07,
+      "loss": 0.0002,
+      "step": 12067
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.0730638090599122e-07,
+      "loss": 0.0003,
+      "step": 12068
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.0728202630297125e-07,
+      "loss": 0.0003,
+      "step": 12069
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.0725767169995127e-07,
+      "loss": 0.0003,
+      "step": 12070
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.0723331709693133e-07,
+      "loss": 0.0003,
+      "step": 12071
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.0720896249391135e-07,
+      "loss": 0.0004,
+      "step": 12072
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.0718460789089138e-07,
+      "loss": 2.2552,
+      "step": 12073
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.071602532878714e-07,
+      "loss": 0.0003,
+      "step": 12074
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 2.0713589868485143e-07,
+      "loss": 0.0002,
+      "step": 12075
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.0711154408183145e-07,
+      "loss": 0.0005,
+      "step": 12076
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.0708718947881148e-07,
+      "loss": 0.0004,
+      "step": 12077
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.070628348757915e-07,
+      "loss": 0.0004,
+      "step": 12078
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.0703848027277155e-07,
+      "loss": 0.0002,
+      "step": 12079
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.0701412566975158e-07,
+      "loss": 0.0003,
+      "step": 12080
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.069897710667316e-07,
+      "loss": 0.0002,
+      "step": 12081
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.0696541646371163e-07,
+      "loss": 0.0005,
+      "step": 12082
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.0694106186069165e-07,
+      "loss": 0.0004,
+      "step": 12083
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.0691670725767168e-07,
+      "loss": 0.0003,
+      "step": 12084
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 2.068923526546517e-07,
+      "loss": 0.0003,
+      "step": 12085
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.0686799805163173e-07,
+      "loss": 0.0003,
+      "step": 12086
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.0684364344861178e-07,
+      "loss": 0.0002,
+      "step": 12087
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.068192888455918e-07,
+      "loss": 0.0004,
+      "step": 12088
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.0679493424257183e-07,
+      "loss": 0.0003,
+      "step": 12089
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.0677057963955186e-07,
+      "loss": 0.0003,
+      "step": 12090
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.0674622503653188e-07,
+      "loss": 0.0003,
+      "step": 12091
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.067218704335119e-07,
+      "loss": 0.0003,
+      "step": 12092
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.0669751583049193e-07,
+      "loss": 0.0003,
+      "step": 12093
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.0667316122747196e-07,
+      "loss": 0.0003,
+      "step": 12094
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 2.06648806624452e-07,
+      "loss": 0.0005,
+      "step": 12095
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0662445202143203e-07,
+      "loss": 0.0004,
+      "step": 12096
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0660009741841209e-07,
+      "loss": 0.0013,
+      "step": 12097
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.065757428153921e-07,
+      "loss": 0.0003,
+      "step": 12098
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0655138821237214e-07,
+      "loss": 0.0003,
+      "step": 12099
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0652703360935216e-07,
+      "loss": 0.0002,
+      "step": 12100
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0650267900633219e-07,
+      "loss": 0.0006,
+      "step": 12101
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0647832440331224e-07,
+      "loss": 0.0002,
+      "step": 12102
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0645396980029226e-07,
+      "loss": 0.0005,
+      "step": 12103
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.064296151972723e-07,
+      "loss": 0.0002,
+      "step": 12104
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0640526059425231e-07,
+      "loss": 0.0003,
+      "step": 12105
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 2.0638090599123234e-07,
+      "loss": 0.0005,
+      "step": 12106
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.0635655138821236e-07,
+      "loss": 0.0004,
+      "step": 12107
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.063321967851924e-07,
+      "loss": 0.0005,
+      "step": 12108
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.0630784218217241e-07,
+      "loss": 0.0003,
+      "step": 12109
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.0628348757915247e-07,
+      "loss": 0.0003,
+      "step": 12110
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.062591329761325e-07,
+      "loss": 0.0003,
+      "step": 12111
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.0623477837311252e-07,
+      "loss": 0.0002,
+      "step": 12112
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.0621042377009254e-07,
+      "loss": 0.0004,
+      "step": 12113
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.0618606916707257e-07,
+      "loss": 0.0003,
+      "step": 12114
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.061617145640526e-07,
+      "loss": 0.0002,
+      "step": 12115
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 2.0613735996103262e-07,
+      "loss": 0.0002,
+      "step": 12116
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.0611300535801264e-07,
+      "loss": 0.0009,
+      "step": 12117
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.060886507549927e-07,
+      "loss": 0.0005,
+      "step": 12118
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.0606429615197272e-07,
+      "loss": 0.0003,
+      "step": 12119
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.0603994154895274e-07,
+      "loss": 0.0003,
+      "step": 12120
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.0601558694593277e-07,
+      "loss": 0.0003,
+      "step": 12121
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.059912323429128e-07,
+      "loss": 0.0004,
+      "step": 12122
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.0596687773989282e-07,
+      "loss": 0.0002,
+      "step": 12123
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.0594252313687284e-07,
+      "loss": 0.0003,
+      "step": 12124
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.0591816853385287e-07,
+      "loss": 0.0002,
+      "step": 12125
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 2.0589381393083292e-07,
+      "loss": 0.0002,
+      "step": 12126
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.0586945932781295e-07,
+      "loss": 0.0003,
+      "step": 12127
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.0584510472479297e-07,
+      "loss": 0.0003,
+      "step": 12128
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.05820750121773e-07,
+      "loss": 0.0003,
+      "step": 12129
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.0579639551875302e-07,
+      "loss": 0.0003,
+      "step": 12130
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.0577204091573305e-07,
+      "loss": 0.0003,
+      "step": 12131
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.0574768631271307e-07,
+      "loss": 0.0004,
+      "step": 12132
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.057233317096931e-07,
+      "loss": 0.0003,
+      "step": 12133
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.0569897710667315e-07,
+      "loss": 1.5667,
+      "step": 12134
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.056746225036532e-07,
+      "loss": 0.0005,
+      "step": 12135
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.0565026790063323e-07,
+      "loss": 0.0005,
+      "step": 12136
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 2.0562591329761325e-07,
+      "loss": 0.0003,
+      "step": 12137
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.0560155869459328e-07,
+      "loss": 0.0004,
+      "step": 12138
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.055772040915733e-07,
+      "loss": 0.0004,
+      "step": 12139
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.0555284948855333e-07,
+      "loss": 0.0003,
+      "step": 12140
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.0552849488553338e-07,
+      "loss": 0.0003,
+      "step": 12141
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.055041402825134e-07,
+      "loss": 0.0003,
+      "step": 12142
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.0547978567949343e-07,
+      "loss": 0.4549,
+      "step": 12143
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.0545543107647345e-07,
+      "loss": 0.0012,
+      "step": 12144
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.0543107647345348e-07,
+      "loss": 0.0004,
+      "step": 12145
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.054067218704335e-07,
+      "loss": 0.0003,
+      "step": 12146
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 2.0538236726741353e-07,
+      "loss": 0.0002,
+      "step": 12147
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.0535801266439355e-07,
+      "loss": 0.0004,
+      "step": 12148
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.053336580613736e-07,
+      "loss": 0.0003,
+      "step": 12149
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.0530930345835363e-07,
+      "loss": 0.0004,
+      "step": 12150
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.0528494885533366e-07,
+      "loss": 0.0004,
+      "step": 12151
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.0526059425231368e-07,
+      "loss": 0.0005,
+      "step": 12152
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.052362396492937e-07,
+      "loss": 0.0003,
+      "step": 12153
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.0521188504627373e-07,
+      "loss": 0.0002,
+      "step": 12154
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.0518753044325376e-07,
+      "loss": 0.0003,
+      "step": 12155
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.0516317584023378e-07,
+      "loss": 0.0006,
+      "step": 12156
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 2.0513882123721383e-07,
+      "loss": 0.024,
+      "step": 12157
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.0511446663419386e-07,
+      "loss": 0.0005,
+      "step": 12158
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.0509011203117388e-07,
+      "loss": 0.0005,
+      "step": 12159
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.050657574281539e-07,
+      "loss": 0.0005,
+      "step": 12160
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.0504140282513393e-07,
+      "loss": 0.0003,
+      "step": 12161
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.0501704822211396e-07,
+      "loss": 0.0004,
+      "step": 12162
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.0499269361909399e-07,
+      "loss": 0.0005,
+      "step": 12163
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.04968339016074e-07,
+      "loss": 0.0004,
+      "step": 12164
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.0494398441305406e-07,
+      "loss": 0.0002,
+      "step": 12165
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.049196298100341e-07,
+      "loss": 0.0004,
+      "step": 12166
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 2.048952752070141e-07,
+      "loss": 0.0003,
+      "step": 12167
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.0487092060399414e-07,
+      "loss": 0.0003,
+      "step": 12168
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.0484656600097416e-07,
+      "loss": 0.0003,
+      "step": 12169
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.048222113979542e-07,
+      "loss": 0.0005,
+      "step": 12170
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.047978567949342e-07,
+      "loss": 0.0003,
+      "step": 12171
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.0477350219191424e-07,
+      "loss": 0.0002,
+      "step": 12172
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.0474914758889432e-07,
+      "loss": 0.0003,
+      "step": 12173
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.0472479298587434e-07,
+      "loss": 0.0005,
+      "step": 12174
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.0470043838285437e-07,
+      "loss": 0.0003,
+      "step": 12175
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.046760837798344e-07,
+      "loss": 0.0003,
+      "step": 12176
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.0465172917681442e-07,
+      "loss": 0.0003,
+      "step": 12177
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 2.0462737457379444e-07,
+      "loss": 0.0002,
+      "step": 12178
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.0460301997077447e-07,
+      "loss": 0.0003,
+      "step": 12179
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.045786653677545e-07,
+      "loss": 0.0005,
+      "step": 12180
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.0455431076473454e-07,
+      "loss": 0.0002,
+      "step": 12181
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.0452995616171457e-07,
+      "loss": 0.0003,
+      "step": 12182
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.045056015586946e-07,
+      "loss": 0.0006,
+      "step": 12183
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.0448124695567462e-07,
+      "loss": 0.0004,
+      "step": 12184
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.0445689235265464e-07,
+      "loss": 0.0002,
+      "step": 12185
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.0443253774963467e-07,
+      "loss": 0.0005,
+      "step": 12186
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.044081831466147e-07,
+      "loss": 0.0002,
+      "step": 12187
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 2.0438382854359472e-07,
+      "loss": 0.0002,
+      "step": 12188
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.0435947394057477e-07,
+      "loss": 2.2576,
+      "step": 12189
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.043351193375548e-07,
+      "loss": 0.0001,
+      "step": 12190
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.0431076473453482e-07,
+      "loss": 0.0007,
+      "step": 12191
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.0428641013151485e-07,
+      "loss": 0.0004,
+      "step": 12192
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.0426205552849487e-07,
+      "loss": 0.0003,
+      "step": 12193
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.042377009254749e-07,
+      "loss": 0.0003,
+      "step": 12194
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.0421334632245492e-07,
+      "loss": 0.0007,
+      "step": 12195
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.0418899171943495e-07,
+      "loss": 0.0004,
+      "step": 12196
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.04164637116415e-07,
+      "loss": 0.0005,
+      "step": 12197
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 2.0414028251339502e-07,
+      "loss": 0.0002,
+      "step": 12198
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.0411592791037505e-07,
+      "loss": 0.0003,
+      "step": 12199
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.0409157330735507e-07,
+      "loss": 0.0011,
+      "step": 12200
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.040672187043351e-07,
+      "loss": 0.0003,
+      "step": 12201
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.0404286410131513e-07,
+      "loss": 0.0004,
+      "step": 12202
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.0401850949829515e-07,
+      "loss": 0.0003,
+      "step": 12203
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.039941548952752e-07,
+      "loss": 0.0002,
+      "step": 12204
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.0396980029225523e-07,
+      "loss": 0.0007,
+      "step": 12205
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.0394544568923525e-07,
+      "loss": 0.0004,
+      "step": 12206
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.0392109108621528e-07,
+      "loss": 0.0002,
+      "step": 12207
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.038967364831953e-07,
+      "loss": 0.0002,
+      "step": 12208
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 2.0387238188017533e-07,
+      "loss": 0.0003,
+      "step": 12209
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.0384802727715535e-07,
+      "loss": 0.0002,
+      "step": 12210
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.038236726741354e-07,
+      "loss": 0.0004,
+      "step": 12211
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.0379931807111546e-07,
+      "loss": 0.0003,
+      "step": 12212
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.0377496346809548e-07,
+      "loss": 0.0004,
+      "step": 12213
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.037506088650755e-07,
+      "loss": 0.0005,
+      "step": 12214
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.0372625426205553e-07,
+      "loss": 0.0003,
+      "step": 12215
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.0370189965903556e-07,
+      "loss": 0.0004,
+      "step": 12216
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.0367754505601558e-07,
+      "loss": 0.0004,
+      "step": 12217
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.036531904529956e-07,
+      "loss": 0.0004,
+      "step": 12218
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 2.0362883584997563e-07,
+      "loss": 0.0004,
+      "step": 12219
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.0360448124695568e-07,
+      "loss": 0.0002,
+      "step": 12220
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.035801266439357e-07,
+      "loss": 0.0005,
+      "step": 12221
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.0355577204091573e-07,
+      "loss": 0.0004,
+      "step": 12222
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.0353141743789576e-07,
+      "loss": 0.0003,
+      "step": 12223
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.0350706283487578e-07,
+      "loss": 0.0002,
+      "step": 12224
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.034827082318558e-07,
+      "loss": 0.0001,
+      "step": 12225
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.0345835362883584e-07,
+      "loss": 0.0004,
+      "step": 12226
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.0343399902581586e-07,
+      "loss": 0.0002,
+      "step": 12227
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.034096444227959e-07,
+      "loss": 0.0004,
+      "step": 12228
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 2.0338528981977594e-07,
+      "loss": 0.0003,
+      "step": 12229
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.0336093521675596e-07,
+      "loss": 0.0002,
+      "step": 12230
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.03336580613736e-07,
+      "loss": 0.0003,
+      "step": 12231
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.03312226010716e-07,
+      "loss": 0.0003,
+      "step": 12232
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.0328787140769604e-07,
+      "loss": 0.0003,
+      "step": 12233
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.0326351680467606e-07,
+      "loss": 0.0003,
+      "step": 12234
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.032391622016561e-07,
+      "loss": 0.0003,
+      "step": 12235
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.0321480759863614e-07,
+      "loss": 0.0003,
+      "step": 12236
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.0319045299561616e-07,
+      "loss": 0.0003,
+      "step": 12237
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.031660983925962e-07,
+      "loss": 0.0002,
+      "step": 12238
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 2.0314174378957622e-07,
+      "loss": 0.0005,
+      "step": 12239
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.0311738918655624e-07,
+      "loss": 0.0003,
+      "step": 12240
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.0309303458353627e-07,
+      "loss": 0.0003,
+      "step": 12241
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.030686799805163e-07,
+      "loss": 0.0005,
+      "step": 12242
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.0304432537749632e-07,
+      "loss": 0.0003,
+      "step": 12243
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.0301997077447637e-07,
+      "loss": 0.0003,
+      "step": 12244
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.029956161714564e-07,
+      "loss": 0.0005,
+      "step": 12245
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.0297126156843642e-07,
+      "loss": 0.0002,
+      "step": 12246
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.0294690696541644e-07,
+      "loss": 0.0003,
+      "step": 12247
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.0292255236239647e-07,
+      "loss": 0.0003,
+      "step": 12248
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.028981977593765e-07,
+      "loss": 0.0003,
+      "step": 12249
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 2.0287384315635654e-07,
+      "loss": 0.0002,
+      "step": 12250
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.028494885533366e-07,
+      "loss": 0.0002,
+      "step": 12251
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.0282513395031662e-07,
+      "loss": 0.0006,
+      "step": 12252
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.0280077934729665e-07,
+      "loss": 0.0003,
+      "step": 12253
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.0277642474427667e-07,
+      "loss": 0.0003,
+      "step": 12254
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.027520701412567e-07,
+      "loss": 0.0004,
+      "step": 12255
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.0272771553823672e-07,
+      "loss": 0.0002,
+      "step": 12256
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.0270336093521675e-07,
+      "loss": 0.0002,
+      "step": 12257
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.0267900633219677e-07,
+      "loss": 0.0003,
+      "step": 12258
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.0265465172917682e-07,
+      "loss": 0.0003,
+      "step": 12259
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 2.0263029712615685e-07,
+      "loss": 0.0002,
+      "step": 12260
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.0260594252313687e-07,
+      "loss": 0.0003,
+      "step": 12261
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.025815879201169e-07,
+      "loss": 0.0004,
+      "step": 12262
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.0255723331709693e-07,
+      "loss": 0.0004,
+      "step": 12263
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.0253287871407695e-07,
+      "loss": 0.0003,
+      "step": 12264
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.0250852411105698e-07,
+      "loss": 0.0004,
+      "step": 12265
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.02484169508037e-07,
+      "loss": 0.0002,
+      "step": 12266
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.0245981490501705e-07,
+      "loss": 0.0004,
+      "step": 12267
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.0243546030199708e-07,
+      "loss": 0.0003,
+      "step": 12268
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.024111056989771e-07,
+      "loss": 0.0002,
+      "step": 12269
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 2.0238675109595713e-07,
+      "loss": 0.0002,
+      "step": 12270
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.0236239649293715e-07,
+      "loss": 0.0004,
+      "step": 12271
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.0233804188991718e-07,
+      "loss": 0.0003,
+      "step": 12272
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.023136872868972e-07,
+      "loss": 1.7386,
+      "step": 12273
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.0228933268387723e-07,
+      "loss": 0.0004,
+      "step": 12274
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.0226497808085728e-07,
+      "loss": 0.0005,
+      "step": 12275
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.022406234778373e-07,
+      "loss": 0.0004,
+      "step": 12276
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.0221626887481733e-07,
+      "loss": 0.0003,
+      "step": 12277
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.0219191427179736e-07,
+      "loss": 0.0002,
+      "step": 12278
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.0216755966877738e-07,
+      "loss": 0.0005,
+      "step": 12279
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.021432050657574e-07,
+      "loss": 0.0004,
+      "step": 12280
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 2.0211885046273743e-07,
+      "loss": 0.0002,
+      "step": 12281
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.0209449585971746e-07,
+      "loss": 0.0003,
+      "step": 12282
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.020701412566975e-07,
+      "loss": 0.0003,
+      "step": 12283
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.0204578665367753e-07,
+      "loss": 0.0055,
+      "step": 12284
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.0202143205065756e-07,
+      "loss": 0.0004,
+      "step": 12285
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.0199707744763758e-07,
+      "loss": 0.0004,
+      "step": 12286
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.019727228446176e-07,
+      "loss": 0.0003,
+      "step": 12287
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.0194836824159766e-07,
+      "loss": 0.0003,
+      "step": 12288
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.0192401363857769e-07,
+      "loss": 0.0003,
+      "step": 12289
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.018996590355577e-07,
+      "loss": 0.0003,
+      "step": 12290
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 2.0187530443253776e-07,
+      "loss": 0.0005,
+      "step": 12291
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.018509498295178e-07,
+      "loss": 0.0006,
+      "step": 12292
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.018265952264978e-07,
+      "loss": 0.0003,
+      "step": 12293
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.0180224062347784e-07,
+      "loss": 0.0005,
+      "step": 12294
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.0177788602045786e-07,
+      "loss": 0.0005,
+      "step": 12295
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.017535314174379e-07,
+      "loss": 0.0003,
+      "step": 12296
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.017291768144179e-07,
+      "loss": 0.0002,
+      "step": 12297
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.0170482221139794e-07,
+      "loss": 0.0003,
+      "step": 12298
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.01680467608378e-07,
+      "loss": 0.0002,
+      "step": 12299
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.0165611300535802e-07,
+      "loss": 0.0002,
+      "step": 12300
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 2.0163175840233804e-07,
+      "loss": 0.0002,
+      "step": 12301
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.0160740379931807e-07,
+      "loss": 0.0004,
+      "step": 12302
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.015830491962981e-07,
+      "loss": 0.0003,
+      "step": 12303
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.0155869459327812e-07,
+      "loss": 0.0003,
+      "step": 12304
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.0153433999025814e-07,
+      "loss": 0.0003,
+      "step": 12305
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.015099853872382e-07,
+      "loss": 0.0013,
+      "step": 12306
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.0148563078421822e-07,
+      "loss": 0.0005,
+      "step": 12307
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.0146127618119824e-07,
+      "loss": 0.0003,
+      "step": 12308
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.0143692157817827e-07,
+      "loss": 0.0003,
+      "step": 12309
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.014125669751583e-07,
+      "loss": 0.0003,
+      "step": 12310
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 2.0138821237213832e-07,
+      "loss": 0.0003,
+      "step": 12311
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0136385776911834e-07,
+      "loss": 0.0004,
+      "step": 12312
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0133950316609837e-07,
+      "loss": 0.0003,
+      "step": 12313
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0131514856307842e-07,
+      "loss": 0.0003,
+      "step": 12314
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0129079396005845e-07,
+      "loss": 0.0003,
+      "step": 12315
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0126643935703847e-07,
+      "loss": 0.0005,
+      "step": 12316
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.012420847540185e-07,
+      "loss": 0.0002,
+      "step": 12317
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0121773015099852e-07,
+      "loss": 0.0003,
+      "step": 12318
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0119337554797855e-07,
+      "loss": 0.0006,
+      "step": 12319
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0116902094495857e-07,
+      "loss": 0.0003,
+      "step": 12320
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.011446663419386e-07,
+      "loss": 0.0003,
+      "step": 12321
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 2.0112031173891865e-07,
+      "loss": 0.0003,
+      "step": 12322
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.0109595713589867e-07,
+      "loss": 0.0002,
+      "step": 12323
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.010716025328787e-07,
+      "loss": 0.0005,
+      "step": 12324
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.0104724792985872e-07,
+      "loss": 0.0003,
+      "step": 12325
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.0102289332683878e-07,
+      "loss": 0.0003,
+      "step": 12326
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.009985387238188e-07,
+      "loss": 0.0003,
+      "step": 12327
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.0097418412079883e-07,
+      "loss": 0.0003,
+      "step": 12328
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.0094982951777885e-07,
+      "loss": 0.0003,
+      "step": 12329
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.009254749147589e-07,
+      "loss": 0.0004,
+      "step": 12330
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.0090112031173893e-07,
+      "loss": 0.0002,
+      "step": 12331
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 2.0087676570871895e-07,
+      "loss": 0.0003,
+      "step": 12332
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.0085241110569898e-07,
+      "loss": 0.0003,
+      "step": 12333
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.00828056502679e-07,
+      "loss": 0.0003,
+      "step": 12334
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.0080370189965903e-07,
+      "loss": 0.0003,
+      "step": 12335
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.0077934729663905e-07,
+      "loss": 0.0004,
+      "step": 12336
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.0075499269361908e-07,
+      "loss": 0.0003,
+      "step": 12337
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.0073063809059913e-07,
+      "loss": 0.0004,
+      "step": 12338
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.0070628348757916e-07,
+      "loss": 0.0003,
+      "step": 12339
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.0068192888455918e-07,
+      "loss": 0.0002,
+      "step": 12340
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.006575742815392e-07,
+      "loss": 0.0001,
+      "step": 12341
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 2.0063321967851923e-07,
+      "loss": 0.0004,
+      "step": 12342
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.0060886507549926e-07,
+      "loss": 0.0003,
+      "step": 12343
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.0058451047247928e-07,
+      "loss": 0.0004,
+      "step": 12344
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.005601558694593e-07,
+      "loss": 0.0004,
+      "step": 12345
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.0053580126643936e-07,
+      "loss": 0.0006,
+      "step": 12346
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.0051144666341938e-07,
+      "loss": 0.0003,
+      "step": 12347
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.004870920603994e-07,
+      "loss": 0.0002,
+      "step": 12348
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.9990277102576568,
+      "eval_loss": 0.014455861411988735,
+      "eval_runtime": 2.7849,
+      "eval_samples_per_second": 738.62,
+      "eval_steps_per_second": 11.85,
+      "step": 12348
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.0046273745737943e-07,
+      "loss": 0.0004,
+      "step": 12349
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.0043838285435946e-07,
+      "loss": 0.0004,
+      "step": 12350
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.0041402825133948e-07,
+      "loss": 0.0004,
+      "step": 12351
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.003896736483195e-07,
+      "loss": 0.0003,
+      "step": 12352
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 2.0036531904529953e-07,
+      "loss": 0.0003,
+      "step": 12353
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.0034096444227959e-07,
+      "loss": 0.0002,
+      "step": 12354
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.003166098392596e-07,
+      "loss": 0.0005,
+      "step": 12355
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.0029225523623964e-07,
+      "loss": 0.0004,
+      "step": 12356
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.0026790063321966e-07,
+      "loss": 0.0003,
+      "step": 12357
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.0024354603019969e-07,
+      "loss": 0.0005,
+      "step": 12358
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.002191914271797e-07,
+      "loss": 0.0002,
+      "step": 12359
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.0019483682415974e-07,
+      "loss": 0.0003,
+      "step": 12360
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.0017048222113976e-07,
+      "loss": 0.0005,
+      "step": 12361
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.0014612761811981e-07,
+      "loss": 0.0002,
+      "step": 12362
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 2.0012177301509984e-07,
+      "loss": 0.0005,
+      "step": 12363
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.0009741841207986e-07,
+      "loss": 0.0003,
+      "step": 12364
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.0007306380905992e-07,
+      "loss": 0.0005,
+      "step": 12365
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.0004870920603994e-07,
+      "loss": 0.0002,
+      "step": 12366
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2.0002435460301997e-07,
+      "loss": 0.0003,
+      "step": 12367
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 2e-07,
+      "loss": 0.0004,
+      "step": 12368
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 1.9997564539698004e-07,
+      "loss": 0.5181,
+      "step": 12369
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 1.9995129079396007e-07,
+      "loss": 0.0005,
+      "step": 12370
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 1.999269361909401e-07,
+      "loss": 0.0003,
+      "step": 12371
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 1.9990258158792012e-07,
+      "loss": 0.0003,
+      "step": 12372
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 1.9987822698490014e-07,
+      "loss": 0.0004,
+      "step": 12373
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.9985387238188017e-07,
+      "loss": 0.0004,
+      "step": 12374
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.998295177788602e-07,
+      "loss": 0.0002,
+      "step": 12375
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.9980516317584022e-07,
+      "loss": 0.0003,
+      "step": 12376
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.9978080857282027e-07,
+      "loss": 0.0003,
+      "step": 12377
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.997564539698003e-07,
+      "loss": 0.0003,
+      "step": 12378
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.9973209936678032e-07,
+      "loss": 0.0004,
+      "step": 12379
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.9970774476376035e-07,
+      "loss": 0.0003,
+      "step": 12380
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.9968339016074037e-07,
+      "loss": 0.0004,
+      "step": 12381
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.996590355577204e-07,
+      "loss": 0.0002,
+      "step": 12382
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.9963468095470042e-07,
+      "loss": 0.0004,
+      "step": 12383
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 1.9961032635168045e-07,
+      "loss": 0.0003,
+      "step": 12384
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.995859717486605e-07,
+      "loss": 0.0003,
+      "step": 12385
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.9956161714564052e-07,
+      "loss": 0.0004,
+      "step": 12386
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.9953726254262055e-07,
+      "loss": 0.0003,
+      "step": 12387
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.9951290793960057e-07,
+      "loss": 0.0002,
+      "step": 12388
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.994885533365806e-07,
+      "loss": 0.0004,
+      "step": 12389
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.9946419873356062e-07,
+      "loss": 0.0003,
+      "step": 12390
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.9943984413054065e-07,
+      "loss": 0.0003,
+      "step": 12391
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.9941548952752067e-07,
+      "loss": 0.0003,
+      "step": 12392
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.9939113492450073e-07,
+      "loss": 0.0003,
+      "step": 12393
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 1.9936678032148075e-07,
+      "loss": 0.0003,
+      "step": 12394
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.9934242571846078e-07,
+      "loss": 0.0004,
+      "step": 12395
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.993180711154408e-07,
+      "loss": 0.0003,
+      "step": 12396
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.9929371651242083e-07,
+      "loss": 0.0003,
+      "step": 12397
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.9926936190940085e-07,
+      "loss": 0.0003,
+      "step": 12398
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.9924500730638088e-07,
+      "loss": 0.0002,
+      "step": 12399
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.992206527033609e-07,
+      "loss": 0.0002,
+      "step": 12400
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.9919629810034095e-07,
+      "loss": 0.0002,
+      "step": 12401
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.9917194349732098e-07,
+      "loss": 0.0004,
+      "step": 12402
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.9914758889430103e-07,
+      "loss": 0.0004,
+      "step": 12403
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 1.9912323429128106e-07,
+      "loss": 0.0002,
+      "step": 12404
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.9909887968826108e-07,
+      "loss": 0.0002,
+      "step": 12405
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.990745250852411e-07,
+      "loss": 0.0002,
+      "step": 12406
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.9905017048222113e-07,
+      "loss": 0.0004,
+      "step": 12407
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.9902581587920118e-07,
+      "loss": 0.0003,
+      "step": 12408
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.990014612761812e-07,
+      "loss": 0.0005,
+      "step": 12409
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.9897710667316123e-07,
+      "loss": 0.0004,
+      "step": 12410
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.9895275207014126e-07,
+      "loss": 0.0003,
+      "step": 12411
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.9892839746712128e-07,
+      "loss": 0.0004,
+      "step": 12412
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.989040428641013e-07,
+      "loss": 0.0003,
+      "step": 12413
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 1.9887968826108133e-07,
+      "loss": 0.0004,
+      "step": 12414
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.9885533365806136e-07,
+      "loss": 0.0002,
+      "step": 12415
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.988309790550414e-07,
+      "loss": 0.0002,
+      "step": 12416
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.9880662445202144e-07,
+      "loss": 0.0003,
+      "step": 12417
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.9878226984900146e-07,
+      "loss": 0.0005,
+      "step": 12418
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.9875791524598149e-07,
+      "loss": 0.0003,
+      "step": 12419
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.987335606429615e-07,
+      "loss": 0.0002,
+      "step": 12420
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.9870920603994154e-07,
+      "loss": 0.0003,
+      "step": 12421
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.9868485143692156e-07,
+      "loss": 0.0003,
+      "step": 12422
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.986604968339016e-07,
+      "loss": 0.0005,
+      "step": 12423
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.9863614223088164e-07,
+      "loss": 0.0002,
+      "step": 12424
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 1.9861178762786166e-07,
+      "loss": 0.0002,
+      "step": 12425
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.985874330248417e-07,
+      "loss": 0.0004,
+      "step": 12426
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.9856307842182171e-07,
+      "loss": 0.0002,
+      "step": 12427
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.9853872381880174e-07,
+      "loss": 0.0003,
+      "step": 12428
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.9851436921578176e-07,
+      "loss": 0.0002,
+      "step": 12429
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.984900146127618e-07,
+      "loss": 0.0004,
+      "step": 12430
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.9846566000974181e-07,
+      "loss": 0.0003,
+      "step": 12431
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.9844130540672187e-07,
+      "loss": 0.0002,
+      "step": 12432
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.984169508037019e-07,
+      "loss": 0.0004,
+      "step": 12433
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.9839259620068192e-07,
+      "loss": 0.0003,
+      "step": 12434
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 1.9836824159766194e-07,
+      "loss": 0.0003,
+      "step": 12435
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.9834388699464197e-07,
+      "loss": 0.0004,
+      "step": 12436
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.98319532391622e-07,
+      "loss": 0.0002,
+      "step": 12437
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.9829517778860202e-07,
+      "loss": 1.9617,
+      "step": 12438
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.9827082318558204e-07,
+      "loss": 0.0003,
+      "step": 12439
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.982464685825621e-07,
+      "loss": 0.0004,
+      "step": 12440
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.9822211397954215e-07,
+      "loss": 0.0003,
+      "step": 12441
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.9819775937652217e-07,
+      "loss": 0.0004,
+      "step": 12442
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.981734047735022e-07,
+      "loss": 0.0003,
+      "step": 12443
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.9814905017048222e-07,
+      "loss": 0.0002,
+      "step": 12444
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 1.9812469556746225e-07,
+      "loss": 0.0003,
+      "step": 12445
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.9810034096444227e-07,
+      "loss": 0.0003,
+      "step": 12446
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.980759863614223e-07,
+      "loss": 0.0002,
+      "step": 12447
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.9805163175840235e-07,
+      "loss": 0.0004,
+      "step": 12448
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.9802727715538237e-07,
+      "loss": 0.0003,
+      "step": 12449
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.980029225523624e-07,
+      "loss": 0.0002,
+      "step": 12450
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.9797856794934242e-07,
+      "loss": 0.0003,
+      "step": 12451
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.9795421334632245e-07,
+      "loss": 0.0003,
+      "step": 12452
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.9792985874330247e-07,
+      "loss": 0.0004,
+      "step": 12453
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.979055041402825e-07,
+      "loss": 0.0008,
+      "step": 12454
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.9788114953726252e-07,
+      "loss": 0.0003,
+      "step": 12455
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 1.9785679493424258e-07,
+      "loss": 0.0004,
+      "step": 12456
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.978324403312226e-07,
+      "loss": 0.0003,
+      "step": 12457
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.9780808572820263e-07,
+      "loss": 0.0004,
+      "step": 12458
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.9778373112518265e-07,
+      "loss": 0.0005,
+      "step": 12459
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.9775937652216268e-07,
+      "loss": 0.0003,
+      "step": 12460
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.977350219191427e-07,
+      "loss": 0.0002,
+      "step": 12461
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.9771066731612273e-07,
+      "loss": 0.0002,
+      "step": 12462
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.9768631271310275e-07,
+      "loss": 0.0003,
+      "step": 12463
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.976619581100828e-07,
+      "loss": 0.0003,
+      "step": 12464
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.9763760350706283e-07,
+      "loss": 0.0004,
+      "step": 12465
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 1.9761324890404285e-07,
+      "loss": 0.0003,
+      "step": 12466
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.9758889430102288e-07,
+      "loss": 0.0003,
+      "step": 12467
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.975645396980029e-07,
+      "loss": 0.0002,
+      "step": 12468
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.9754018509498293e-07,
+      "loss": 0.0003,
+      "step": 12469
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.9751583049196295e-07,
+      "loss": 0.0004,
+      "step": 12470
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.97491475888943e-07,
+      "loss": 0.0002,
+      "step": 12471
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.9746712128592303e-07,
+      "loss": 0.0003,
+      "step": 12472
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.9744276668290306e-07,
+      "loss": 0.0003,
+      "step": 12473
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.9741841207988308e-07,
+      "loss": 0.0005,
+      "step": 12474
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.973940574768631e-07,
+      "loss": 0.0002,
+      "step": 12475
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 1.9736970287384313e-07,
+      "loss": 0.0004,
+      "step": 12476
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.9734534827082316e-07,
+      "loss": 0.0003,
+      "step": 12477
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.9732099366780318e-07,
+      "loss": 0.0003,
+      "step": 12478
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.9729663906478323e-07,
+      "loss": 0.0002,
+      "step": 12479
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.9727228446176329e-07,
+      "loss": 0.0004,
+      "step": 12480
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.972479298587433e-07,
+      "loss": 0.0002,
+      "step": 12481
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.9722357525572334e-07,
+      "loss": 0.0005,
+      "step": 12482
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.9719922065270336e-07,
+      "loss": 0.0003,
+      "step": 12483
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.9717486604968339e-07,
+      "loss": 0.0005,
+      "step": 12484
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.971505114466634e-07,
+      "loss": 0.0002,
+      "step": 12485
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 1.9712615684364344e-07,
+      "loss": 0.0131,
+      "step": 12486
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.971018022406235e-07,
+      "loss": 0.0002,
+      "step": 12487
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9707744763760351e-07,
+      "loss": 0.0002,
+      "step": 12488
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9705309303458354e-07,
+      "loss": 0.0002,
+      "step": 12489
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9702873843156356e-07,
+      "loss": 0.0003,
+      "step": 12490
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.970043838285436e-07,
+      "loss": 0.0002,
+      "step": 12491
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9698002922552361e-07,
+      "loss": 0.0002,
+      "step": 12492
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9695567462250364e-07,
+      "loss": 0.0004,
+      "step": 12493
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9693132001948366e-07,
+      "loss": 0.0005,
+      "step": 12494
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9690696541646372e-07,
+      "loss": 0.0186,
+      "step": 12495
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9688261081344374e-07,
+      "loss": 0.0006,
+      "step": 12496
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 1.9685825621042377e-07,
+      "loss": 0.0004,
+      "step": 12497
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.968339016074038e-07,
+      "loss": 0.0004,
+      "step": 12498
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.9680954700438382e-07,
+      "loss": 0.0003,
+      "step": 12499
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.9678519240136384e-07,
+      "loss": 0.0003,
+      "step": 12500
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.9676083779834387e-07,
+      "loss": 0.0004,
+      "step": 12501
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.967364831953239e-07,
+      "loss": 0.0003,
+      "step": 12502
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.9671212859230394e-07,
+      "loss": 0.0003,
+      "step": 12503
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.9668777398928397e-07,
+      "loss": 0.0003,
+      "step": 12504
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.96663419386264e-07,
+      "loss": 0.0002,
+      "step": 12505
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.9663906478324402e-07,
+      "loss": 0.0004,
+      "step": 12506
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 1.9661471018022404e-07,
+      "loss": 0.0003,
+      "step": 12507
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.9659035557720407e-07,
+      "loss": 0.0004,
+      "step": 12508
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.965660009741841e-07,
+      "loss": 0.0004,
+      "step": 12509
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.9654164637116412e-07,
+      "loss": 0.0007,
+      "step": 12510
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.9651729176814417e-07,
+      "loss": 0.0042,
+      "step": 12511
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.964929371651242e-07,
+      "loss": 0.0002,
+      "step": 12512
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.9646858256210422e-07,
+      "loss": 0.0004,
+      "step": 12513
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.9644422795908425e-07,
+      "loss": 0.0004,
+      "step": 12514
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.9641987335606427e-07,
+      "loss": 0.0004,
+      "step": 12515
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.963955187530443e-07,
+      "loss": 0.0002,
+      "step": 12516
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 1.9637116415002432e-07,
+      "loss": 0.0003,
+      "step": 12517
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.963468095470044e-07,
+      "loss": 0.0003,
+      "step": 12518
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9632245494398443e-07,
+      "loss": 0.0002,
+      "step": 12519
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9629810034096445e-07,
+      "loss": 0.0004,
+      "step": 12520
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9627374573794448e-07,
+      "loss": 0.0005,
+      "step": 12521
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.962493911349245e-07,
+      "loss": 0.0005,
+      "step": 12522
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9622503653190453e-07,
+      "loss": 0.0003,
+      "step": 12523
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9620068192888455e-07,
+      "loss": 0.0003,
+      "step": 12524
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9617632732586458e-07,
+      "loss": 0.0003,
+      "step": 12525
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9615197272284463e-07,
+      "loss": 0.0002,
+      "step": 12526
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9612761811982465e-07,
+      "loss": 0.0003,
+      "step": 12527
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 1.9610326351680468e-07,
+      "loss": 0.0003,
+      "step": 12528
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.960789089137847e-07,
+      "loss": 0.0004,
+      "step": 12529
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.9605455431076473e-07,
+      "loss": 0.0004,
+      "step": 12530
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.9603019970774475e-07,
+      "loss": 0.0002,
+      "step": 12531
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.9600584510472478e-07,
+      "loss": 0.0003,
+      "step": 12532
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.959814905017048e-07,
+      "loss": 0.0003,
+      "step": 12533
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.9595713589868486e-07,
+      "loss": 0.0004,
+      "step": 12534
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.9593278129566488e-07,
+      "loss": 0.0004,
+      "step": 12535
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.959084266926449e-07,
+      "loss": 0.0003,
+      "step": 12536
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.9588407208962493e-07,
+      "loss": 0.0002,
+      "step": 12537
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 1.9585971748660496e-07,
+      "loss": 0.0002,
+      "step": 12538
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.9583536288358498e-07,
+      "loss": 0.0002,
+      "step": 12539
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.95811008280565e-07,
+      "loss": 0.0004,
+      "step": 12540
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.9578665367754503e-07,
+      "loss": 0.0003,
+      "step": 12541
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.9576229907452508e-07,
+      "loss": 0.0002,
+      "step": 12542
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.957379444715051e-07,
+      "loss": 0.0004,
+      "step": 12543
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.9571358986848513e-07,
+      "loss": 0.0003,
+      "step": 12544
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.9568923526546516e-07,
+      "loss": 0.0002,
+      "step": 12545
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.9566488066244519e-07,
+      "loss": 0.0003,
+      "step": 12546
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.956405260594252e-07,
+      "loss": 0.0003,
+      "step": 12547
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 1.9561617145640524e-07,
+      "loss": 0.5246,
+      "step": 12548
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.9559181685338526e-07,
+      "loss": 0.0002,
+      "step": 12549
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.955674622503653e-07,
+      "loss": 0.0003,
+      "step": 12550
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.9554310764734534e-07,
+      "loss": 0.0003,
+      "step": 12551
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.9551875304432536e-07,
+      "loss": 0.0002,
+      "step": 12552
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.954943984413054e-07,
+      "loss": 0.0002,
+      "step": 12553
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.954700438382854e-07,
+      "loss": 0.0002,
+      "step": 12554
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.9544568923526544e-07,
+      "loss": 0.0004,
+      "step": 12555
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.954213346322455e-07,
+      "loss": 0.0003,
+      "step": 12556
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.9539698002922551e-07,
+      "loss": 0.0004,
+      "step": 12557
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 1.9537262542620557e-07,
+      "loss": 0.0002,
+      "step": 12558
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.953482708231856e-07,
+      "loss": 0.0002,
+      "step": 12559
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.9532391622016562e-07,
+      "loss": 0.0003,
+      "step": 12560
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.9529956161714564e-07,
+      "loss": 0.0004,
+      "step": 12561
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.9527520701412567e-07,
+      "loss": 0.0005,
+      "step": 12562
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.952508524111057e-07,
+      "loss": 0.0003,
+      "step": 12563
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.9522649780808572e-07,
+      "loss": 0.0002,
+      "step": 12564
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.9520214320506574e-07,
+      "loss": 0.0002,
+      "step": 12565
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.951777886020458e-07,
+      "loss": 0.0003,
+      "step": 12566
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.9515343399902582e-07,
+      "loss": 0.0003,
+      "step": 12567
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.9512907939600584e-07,
+      "loss": 0.0003,
+      "step": 12568
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 1.9510472479298587e-07,
+      "loss": 0.0002,
+      "step": 12569
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.950803701899659e-07,
+      "loss": 0.0004,
+      "step": 12570
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.9505601558694592e-07,
+      "loss": 0.0002,
+      "step": 12571
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.9503166098392595e-07,
+      "loss": 0.0004,
+      "step": 12572
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.95007306380906e-07,
+      "loss": 0.0003,
+      "step": 12573
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.9498295177788602e-07,
+      "loss": 0.0006,
+      "step": 12574
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.9495859717486605e-07,
+      "loss": 0.0002,
+      "step": 12575
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.9493424257184607e-07,
+      "loss": 0.0004,
+      "step": 12576
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.949098879688261e-07,
+      "loss": 0.0003,
+      "step": 12577
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.9488553336580612e-07,
+      "loss": 0.0023,
+      "step": 12578
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 1.9486117876278615e-07,
+      "loss": 0.0001,
+      "step": 12579
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.9483682415976617e-07,
+      "loss": 0.0003,
+      "step": 12580
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.9481246955674622e-07,
+      "loss": 0.0003,
+      "step": 12581
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.9478811495372625e-07,
+      "loss": 0.0005,
+      "step": 12582
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.9476376035070627e-07,
+      "loss": 0.0002,
+      "step": 12583
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.947394057476863e-07,
+      "loss": 0.0003,
+      "step": 12584
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.9471505114466633e-07,
+      "loss": 0.0003,
+      "step": 12585
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.9469069654164635e-07,
+      "loss": 0.0002,
+      "step": 12586
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.9466634193862638e-07,
+      "loss": 0.0007,
+      "step": 12587
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.946419873356064e-07,
+      "loss": 0.0002,
+      "step": 12588
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 1.9461763273258645e-07,
+      "loss": 0.0002,
+      "step": 12589
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.9459327812956648e-07,
+      "loss": 0.0003,
+      "step": 12590
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.945689235265465e-07,
+      "loss": 0.0004,
+      "step": 12591
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.9454456892352653e-07,
+      "loss": 0.0003,
+      "step": 12592
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.9452021432050655e-07,
+      "loss": 0.0003,
+      "step": 12593
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.944958597174866e-07,
+      "loss": 0.0004,
+      "step": 12594
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.9447150511446663e-07,
+      "loss": 0.0003,
+      "step": 12595
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.9444715051144666e-07,
+      "loss": 0.0002,
+      "step": 12596
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.944227959084267e-07,
+      "loss": 0.0002,
+      "step": 12597
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.9439844130540673e-07,
+      "loss": 0.0002,
+      "step": 12598
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.9437408670238676e-07,
+      "loss": 0.0002,
+      "step": 12599
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 1.9434973209936678e-07,
+      "loss": 0.0002,
+      "step": 12600
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.943253774963468e-07,
+      "loss": 0.0003,
+      "step": 12601
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.9430102289332683e-07,
+      "loss": 0.0003,
+      "step": 12602
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.9427666829030686e-07,
+      "loss": 0.0002,
+      "step": 12603
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.9425231368728688e-07,
+      "loss": 0.0002,
+      "step": 12604
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.9422795908426693e-07,
+      "loss": 0.0002,
+      "step": 12605
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.9420360448124696e-07,
+      "loss": 0.0003,
+      "step": 12606
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.9417924987822698e-07,
+      "loss": 1.4666,
+      "step": 12607
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.94154895275207e-07,
+      "loss": 0.0005,
+      "step": 12608
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.9413054067218704e-07,
+      "loss": 0.0002,
+      "step": 12609
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 1.9410618606916706e-07,
+      "loss": 0.0002,
+      "step": 12610
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.9408183146614709e-07,
+      "loss": 0.0004,
+      "step": 12611
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.940574768631271e-07,
+      "loss": 0.0002,
+      "step": 12612
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.9403312226010716e-07,
+      "loss": 0.0002,
+      "step": 12613
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.940087676570872e-07,
+      "loss": 0.0003,
+      "step": 12614
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.939844130540672e-07,
+      "loss": 0.0003,
+      "step": 12615
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.9396005845104724e-07,
+      "loss": 0.0003,
+      "step": 12616
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.9393570384802726e-07,
+      "loss": 0.0008,
+      "step": 12617
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.939113492450073e-07,
+      "loss": 0.0003,
+      "step": 12618
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.9388699464198731e-07,
+      "loss": 0.0003,
+      "step": 12619
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 1.9386264003896734e-07,
+      "loss": 0.0005,
+      "step": 12620
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.938382854359474e-07,
+      "loss": 0.0002,
+      "step": 12621
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.9381393083292742e-07,
+      "loss": 0.0002,
+      "step": 12622
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.9378957622990744e-07,
+      "loss": 0.0004,
+      "step": 12623
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.9376522162688747e-07,
+      "loss": 0.0003,
+      "step": 12624
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.937408670238675e-07,
+      "loss": 0.0006,
+      "step": 12625
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.9371651242084752e-07,
+      "loss": 0.0003,
+      "step": 12626
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.9369215781782754e-07,
+      "loss": 0.0005,
+      "step": 12627
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.9366780321480757e-07,
+      "loss": 0.0003,
+      "step": 12628
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.9364344861178762e-07,
+      "loss": 0.0003,
+      "step": 12629
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 1.9361909400876764e-07,
+      "loss": 0.0003,
+      "step": 12630
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.9359473940574767e-07,
+      "loss": 0.0002,
+      "step": 12631
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.935703848027277e-07,
+      "loss": 0.0002,
+      "step": 12632
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.9354603019970775e-07,
+      "loss": 0.0002,
+      "step": 12633
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.9352167559668777e-07,
+      "loss": 0.0005,
+      "step": 12634
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.934973209936678e-07,
+      "loss": 0.0005,
+      "step": 12635
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.9347296639064785e-07,
+      "loss": 0.0002,
+      "step": 12636
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.9344861178762787e-07,
+      "loss": 0.0004,
+      "step": 12637
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.934242571846079e-07,
+      "loss": 0.0004,
+      "step": 12638
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.9339990258158792e-07,
+      "loss": 0.0032,
+      "step": 12639
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.9337554797856795e-07,
+      "loss": 0.0002,
+      "step": 12640
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 1.9335119337554797e-07,
+      "loss": 0.0003,
+      "step": 12641
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.93326838772528e-07,
+      "loss": 0.0004,
+      "step": 12642
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.9330248416950802e-07,
+      "loss": 0.0004,
+      "step": 12643
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.9327812956648807e-07,
+      "loss": 0.0005,
+      "step": 12644
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.932537749634681e-07,
+      "loss": 0.0004,
+      "step": 12645
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.9322942036044813e-07,
+      "loss": 0.0002,
+      "step": 12646
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.9320506575742815e-07,
+      "loss": 0.0003,
+      "step": 12647
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.9318071115440818e-07,
+      "loss": 0.0003,
+      "step": 12648
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.931563565513882e-07,
+      "loss": 0.0008,
+      "step": 12649
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.9313200194836823e-07,
+      "loss": 0.0002,
+      "step": 12650
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 1.9310764734534825e-07,
+      "loss": 0.0003,
+      "step": 12651
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.930832927423283e-07,
+      "loss": 0.0003,
+      "step": 12652
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.9305893813930833e-07,
+      "loss": 0.0003,
+      "step": 12653
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.9303458353628835e-07,
+      "loss": 0.0003,
+      "step": 12654
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.9301022893326838e-07,
+      "loss": 0.0003,
+      "step": 12655
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.929858743302484e-07,
+      "loss": 0.0002,
+      "step": 12656
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.9296151972722843e-07,
+      "loss": 0.0003,
+      "step": 12657
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.9293716512420845e-07,
+      "loss": 0.0003,
+      "step": 12658
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.9291281052118848e-07,
+      "loss": 0.0004,
+      "step": 12659
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.9288845591816853e-07,
+      "loss": 0.0002,
+      "step": 12660
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 1.9286410131514856e-07,
+      "loss": 0.0004,
+      "step": 12661
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.9283974671212858e-07,
+      "loss": 0.0006,
+      "step": 12662
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.928153921091086e-07,
+      "loss": 0.0002,
+      "step": 12663
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.9279103750608863e-07,
+      "loss": 0.0002,
+      "step": 12664
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.9276668290306866e-07,
+      "loss": 0.0003,
+      "step": 12665
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.9274232830004868e-07,
+      "loss": 0.0003,
+      "step": 12666
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.927179736970287e-07,
+      "loss": 0.0003,
+      "step": 12667
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.9269361909400876e-07,
+      "loss": 0.0001,
+      "step": 12668
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.9266926449098878e-07,
+      "loss": 0.0002,
+      "step": 12669
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.926449098879688e-07,
+      "loss": 0.0003,
+      "step": 12670
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.9262055528494886e-07,
+      "loss": 0.0002,
+      "step": 12671
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 1.9259620068192889e-07,
+      "loss": 0.0008,
+      "step": 12672
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.925718460789089e-07,
+      "loss": 0.0003,
+      "step": 12673
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.9254749147588894e-07,
+      "loss": 0.0003,
+      "step": 12674
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.92523136872869e-07,
+      "loss": 0.0003,
+      "step": 12675
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.92498782269849e-07,
+      "loss": 0.0003,
+      "step": 12676
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.9247442766682904e-07,
+      "loss": 0.0003,
+      "step": 12677
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.9245007306380906e-07,
+      "loss": 0.0003,
+      "step": 12678
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.924257184607891e-07,
+      "loss": 0.0003,
+      "step": 12679
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.924013638577691e-07,
+      "loss": 0.0002,
+      "step": 12680
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.9237700925474914e-07,
+      "loss": 0.0004,
+      "step": 12681
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 1.9235265465172916e-07,
+      "loss": 0.0003,
+      "step": 12682
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.9232830004870922e-07,
+      "loss": 0.0004,
+      "step": 12683
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.9230394544568924e-07,
+      "loss": 0.0003,
+      "step": 12684
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.9227959084266927e-07,
+      "loss": 0.0003,
+      "step": 12685
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.922552362396493e-07,
+      "loss": 0.0002,
+      "step": 12686
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.9223088163662932e-07,
+      "loss": 0.0003,
+      "step": 12687
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.9220652703360934e-07,
+      "loss": 0.0004,
+      "step": 12688
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.9218217243058937e-07,
+      "loss": 2.5715,
+      "step": 12689
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.921578178275694e-07,
+      "loss": 0.0008,
+      "step": 12690
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.9213346322454944e-07,
+      "loss": 0.0002,
+      "step": 12691
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 1.9210910862152947e-07,
+      "loss": 0.0002,
+      "step": 12692
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.920847540185095e-07,
+      "loss": 0.0002,
+      "step": 12693
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.9206039941548952e-07,
+      "loss": 0.0003,
+      "step": 12694
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.9203604481246954e-07,
+      "loss": 0.0004,
+      "step": 12695
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.9201169020944957e-07,
+      "loss": 0.0003,
+      "step": 12696
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.919873356064296e-07,
+      "loss": 0.0002,
+      "step": 12697
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.9196298100340962e-07,
+      "loss": 0.0002,
+      "step": 12698
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.9193862640038967e-07,
+      "loss": 0.0002,
+      "step": 12699
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.919142717973697e-07,
+      "loss": 0.0002,
+      "step": 12700
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.9188991719434972e-07,
+      "loss": 0.0003,
+      "step": 12701
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.9186556259132975e-07,
+      "loss": 0.0006,
+      "step": 12702
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 1.9184120798830977e-07,
+      "loss": 0.0002,
+      "step": 12703
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.918168533852898e-07,
+      "loss": 0.0002,
+      "step": 12704
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.9179249878226982e-07,
+      "loss": 0.0004,
+      "step": 12705
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.9176814417924985e-07,
+      "loss": 0.0005,
+      "step": 12706
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.917437895762299e-07,
+      "loss": 0.0003,
+      "step": 12707
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.9171943497320992e-07,
+      "loss": 0.0002,
+      "step": 12708
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.9169508037018998e-07,
+      "loss": 0.0002,
+      "step": 12709
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.9167072576717e-07,
+      "loss": 0.0002,
+      "step": 12710
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.9164637116415003e-07,
+      "loss": 0.0003,
+      "step": 12711
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.9162201656113005e-07,
+      "loss": 0.0002,
+      "step": 12712
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 1.9159766195811008e-07,
+      "loss": 0.0002,
+      "step": 12713
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.915733073550901e-07,
+      "loss": 0.0005,
+      "step": 12714
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.9154895275207015e-07,
+      "loss": 0.0003,
+      "step": 12715
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.9152459814905018e-07,
+      "loss": 0.0004,
+      "step": 12716
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.915002435460302e-07,
+      "loss": 0.0003,
+      "step": 12717
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.9147588894301023e-07,
+      "loss": 0.0002,
+      "step": 12718
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.9145153433999025e-07,
+      "loss": 0.0002,
+      "step": 12719
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.9142717973697028e-07,
+      "loss": 0.0002,
+      "step": 12720
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.914028251339503e-07,
+      "loss": 0.0004,
+      "step": 12721
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.9137847053093033e-07,
+      "loss": 0.0003,
+      "step": 12722
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 1.9135411592791038e-07,
+      "loss": 0.0003,
+      "step": 12723
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.913297613248904e-07,
+      "loss": 0.0002,
+      "step": 12724
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.9130540672187043e-07,
+      "loss": 0.0002,
+      "step": 12725
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.9128105211885046e-07,
+      "loss": 0.0003,
+      "step": 12726
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.9125669751583048e-07,
+      "loss": 0.0002,
+      "step": 12727
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.912323429128105e-07,
+      "loss": 0.001,
+      "step": 12728
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.9120798830979053e-07,
+      "loss": 0.0003,
+      "step": 12729
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.9118363370677056e-07,
+      "loss": 0.001,
+      "step": 12730
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.911592791037506e-07,
+      "loss": 0.0002,
+      "step": 12731
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.9113492450073063e-07,
+      "loss": 0.0002,
+      "step": 12732
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 1.9111056989771066e-07,
+      "loss": 0.0004,
+      "step": 12733
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9108621529469068e-07,
+      "loss": 0.0003,
+      "step": 12734
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.910618606916707e-07,
+      "loss": 0.0003,
+      "step": 12735
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9103750608865073e-07,
+      "loss": 0.0004,
+      "step": 12736
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9101315148563076e-07,
+      "loss": 0.0002,
+      "step": 12737
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9098879688261078e-07,
+      "loss": 0.0003,
+      "step": 12738
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9096444227959084e-07,
+      "loss": 0.0004,
+      "step": 12739
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9094008767657086e-07,
+      "loss": 0.0003,
+      "step": 12740
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9091573307355089e-07,
+      "loss": 0.0003,
+      "step": 12741
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.908913784705309e-07,
+      "loss": 0.0004,
+      "step": 12742
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9086702386751094e-07,
+      "loss": 0.0003,
+      "step": 12743
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 1.9084266926449096e-07,
+      "loss": 0.0004,
+      "step": 12744
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.90818314661471e-07,
+      "loss": 0.0002,
+      "step": 12745
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.9079396005845104e-07,
+      "loss": 0.0003,
+      "step": 12746
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.9076960545543106e-07,
+      "loss": 0.0005,
+      "step": 12747
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.9074525085241112e-07,
+      "loss": 0.0004,
+      "step": 12748
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.9072089624939114e-07,
+      "loss": 0.0005,
+      "step": 12749
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.9069654164637117e-07,
+      "loss": 0.0002,
+      "step": 12750
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.906721870433512e-07,
+      "loss": 0.0005,
+      "step": 12751
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.9064783244033122e-07,
+      "loss": 0.0002,
+      "step": 12752
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.9062347783731124e-07,
+      "loss": 0.0002,
+      "step": 12753
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 1.905991232342913e-07,
+      "loss": 0.0003,
+      "step": 12754
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9057476863127132e-07,
+      "loss": 0.0001,
+      "step": 12755
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9055041402825134e-07,
+      "loss": 0.0003,
+      "step": 12756
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9052605942523137e-07,
+      "loss": 0.0005,
+      "step": 12757
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.905017048222114e-07,
+      "loss": 0.0005,
+      "step": 12758
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9047735021919142e-07,
+      "loss": 0.0002,
+      "step": 12759
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9045299561617144e-07,
+      "loss": 0.0003,
+      "step": 12760
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9042864101315147e-07,
+      "loss": 0.0002,
+      "step": 12761
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9040428641013152e-07,
+      "loss": 0.0003,
+      "step": 12762
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9037993180711155e-07,
+      "loss": 0.0003,
+      "step": 12763
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 1.9035557720409157e-07,
+      "loss": 0.0003,
+      "step": 12764
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.903312226010716e-07,
+      "loss": 0.0005,
+      "step": 12765
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.9030686799805162e-07,
+      "loss": 0.0003,
+      "step": 12766
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.9028251339503165e-07,
+      "loss": 0.0004,
+      "step": 12767
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.9025815879201167e-07,
+      "loss": 0.0003,
+      "step": 12768
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.902338041889917e-07,
+      "loss": 0.0002,
+      "step": 12769
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.9020944958597175e-07,
+      "loss": 0.0002,
+      "step": 12770
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.9018509498295177e-07,
+      "loss": 0.0003,
+      "step": 12771
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.901607403799318e-07,
+      "loss": 0.0003,
+      "step": 12772
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.9013638577691182e-07,
+      "loss": 0.0004,
+      "step": 12773
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.9011203117389185e-07,
+      "loss": 0.0003,
+      "step": 12774
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 1.9008767657087187e-07,
+      "loss": 0.0005,
+      "step": 12775
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.900633219678519e-07,
+      "loss": 1.3439,
+      "step": 12776
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.9003896736483192e-07,
+      "loss": 0.0003,
+      "step": 12777
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.9001461276181198e-07,
+      "loss": 0.0007,
+      "step": 12778
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.89990258158792e-07,
+      "loss": 0.0002,
+      "step": 12779
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.8996590355577203e-07,
+      "loss": 0.0002,
+      "step": 12780
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.8994154895275205e-07,
+      "loss": 0.0002,
+      "step": 12781
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.8991719434973208e-07,
+      "loss": 0.0005,
+      "step": 12782
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.898928397467121e-07,
+      "loss": 0.0003,
+      "step": 12783
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.8986848514369213e-07,
+      "loss": 0.0003,
+      "step": 12784
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 1.8984413054067215e-07,
+      "loss": 0.0005,
+      "step": 12785
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8981977593765223e-07,
+      "loss": 0.0003,
+      "step": 12786
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8979542133463226e-07,
+      "loss": 0.0002,
+      "step": 12787
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8977106673161228e-07,
+      "loss": 0.0002,
+      "step": 12788
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.897467121285923e-07,
+      "loss": 0.0002,
+      "step": 12789
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8972235752557233e-07,
+      "loss": 0.0002,
+      "step": 12790
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8969800292255236e-07,
+      "loss": 0.0002,
+      "step": 12791
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8967364831953238e-07,
+      "loss": 0.0005,
+      "step": 12792
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8964929371651243e-07,
+      "loss": 0.0004,
+      "step": 12793
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8962493911349246e-07,
+      "loss": 0.0003,
+      "step": 12794
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 1.8960058451047248e-07,
+      "loss": 0.0004,
+      "step": 12795
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.895762299074525e-07,
+      "loss": 0.0003,
+      "step": 12796
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.8955187530443253e-07,
+      "loss": 0.0003,
+      "step": 12797
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.8952752070141256e-07,
+      "loss": 0.0002,
+      "step": 12798
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.8950316609839258e-07,
+      "loss": 0.0001,
+      "step": 12799
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.894788114953726e-07,
+      "loss": 0.0003,
+      "step": 12800
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.8945445689235266e-07,
+      "loss": 0.0002,
+      "step": 12801
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.8943010228933269e-07,
+      "loss": 0.0005,
+      "step": 12802
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.894057476863127e-07,
+      "loss": 0.0004,
+      "step": 12803
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.8938139308329274e-07,
+      "loss": 0.0003,
+      "step": 12804
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 1.8935703848027276e-07,
+      "loss": 0.0004,
+      "step": 12805
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.893326838772528e-07,
+      "loss": 0.0003,
+      "step": 12806
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.893083292742328e-07,
+      "loss": 0.0004,
+      "step": 12807
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.8928397467121284e-07,
+      "loss": 0.0003,
+      "step": 12808
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.892596200681929e-07,
+      "loss": 0.0003,
+      "step": 12809
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.8923526546517291e-07,
+      "loss": 0.0007,
+      "step": 12810
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.8921091086215294e-07,
+      "loss": 0.0014,
+      "step": 12811
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.8918655625913296e-07,
+      "loss": 0.0003,
+      "step": 12812
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.89162201656113e-07,
+      "loss": 0.0002,
+      "step": 12813
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.8913784705309301e-07,
+      "loss": 0.0002,
+      "step": 12814
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.8911349245007304e-07,
+      "loss": 0.0005,
+      "step": 12815
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 1.8908913784705306e-07,
+      "loss": 0.0002,
+      "step": 12816
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.8906478324403312e-07,
+      "loss": 0.0002,
+      "step": 12817
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.8904042864101314e-07,
+      "loss": 0.0003,
+      "step": 12818
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.8901607403799317e-07,
+      "loss": 0.0003,
+      "step": 12819
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.889917194349732e-07,
+      "loss": 0.0004,
+      "step": 12820
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.8896736483195322e-07,
+      "loss": 0.0003,
+      "step": 12821
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.8894301022893324e-07,
+      "loss": 0.0002,
+      "step": 12822
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.8891865562591327e-07,
+      "loss": 0.0003,
+      "step": 12823
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.8889430102289332e-07,
+      "loss": 0.0005,
+      "step": 12824
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.8886994641987337e-07,
+      "loss": 0.0002,
+      "step": 12825
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 1.888455918168534e-07,
+      "loss": 0.0004,
+      "step": 12826
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.8882123721383342e-07,
+      "loss": 0.0002,
+      "step": 12827
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.8879688261081345e-07,
+      "loss": 0.0002,
+      "step": 12828
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.8877252800779347e-07,
+      "loss": 0.0004,
+      "step": 12829
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.887481734047735e-07,
+      "loss": 0.0003,
+      "step": 12830
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.8872381880175352e-07,
+      "loss": 0.0005,
+      "step": 12831
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.8869946419873355e-07,
+      "loss": 0.0005,
+      "step": 12832
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.886751095957136e-07,
+      "loss": 0.0002,
+      "step": 12833
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.8865075499269362e-07,
+      "loss": 0.0004,
+      "step": 12834
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.8862640038967365e-07,
+      "loss": 0.0003,
+      "step": 12835
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 1.8860204578665367e-07,
+      "loss": 0.0003,
+      "step": 12836
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.885776911836337e-07,
+      "loss": 0.0003,
+      "step": 12837
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.8855333658061372e-07,
+      "loss": 0.0004,
+      "step": 12838
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.8852898197759375e-07,
+      "loss": 0.0002,
+      "step": 12839
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.885046273745738e-07,
+      "loss": 0.0003,
+      "step": 12840
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.8848027277155383e-07,
+      "loss": 0.0003,
+      "step": 12841
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.8845591816853385e-07,
+      "loss": 0.0003,
+      "step": 12842
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.8843156356551388e-07,
+      "loss": 0.0003,
+      "step": 12843
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.884072089624939e-07,
+      "loss": 0.0003,
+      "step": 12844
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.8838285435947393e-07,
+      "loss": 0.0005,
+      "step": 12845
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.8835849975645395e-07,
+      "loss": 0.0003,
+      "step": 12846
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 1.8833414515343398e-07,
+      "loss": 0.0004,
+      "step": 12847
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.8830979055041403e-07,
+      "loss": 0.0003,
+      "step": 12848
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.8828543594739405e-07,
+      "loss": 0.0004,
+      "step": 12849
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.8826108134437408e-07,
+      "loss": 0.0002,
+      "step": 12850
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.882367267413541e-07,
+      "loss": 0.0003,
+      "step": 12851
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.8821237213833413e-07,
+      "loss": 0.0004,
+      "step": 12852
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.8818801753531415e-07,
+      "loss": 0.0005,
+      "step": 12853
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.8816366293229418e-07,
+      "loss": 0.0004,
+      "step": 12854
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.881393083292742e-07,
+      "loss": 0.0003,
+      "step": 12855
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.8811495372625426e-07,
+      "loss": 0.0004,
+      "step": 12856
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 1.8809059912323428e-07,
+      "loss": 0.0003,
+      "step": 12857
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.880662445202143e-07,
+      "loss": 0.0003,
+      "step": 12858
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.8804188991719433e-07,
+      "loss": 0.0007,
+      "step": 12859
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.8801753531417436e-07,
+      "loss": 0.0002,
+      "step": 12860
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.8799318071115438e-07,
+      "loss": 0.0003,
+      "step": 12861
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.879688261081344e-07,
+      "loss": 0.0004,
+      "step": 12862
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.8794447150511446e-07,
+      "loss": 0.0002,
+      "step": 12863
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.879201169020945e-07,
+      "loss": 0.0003,
+      "step": 12864
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.8789576229907454e-07,
+      "loss": 0.0002,
+      "step": 12865
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.8787140769605456e-07,
+      "loss": 0.0003,
+      "step": 12866
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 1.8784705309303459e-07,
+      "loss": 0.0002,
+      "step": 12867
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.878226984900146e-07,
+      "loss": 0.0004,
+      "step": 12868
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.8779834388699464e-07,
+      "loss": 0.0002,
+      "step": 12869
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.8777398928397466e-07,
+      "loss": 0.0003,
+      "step": 12870
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.877496346809547e-07,
+      "loss": 0.0007,
+      "step": 12871
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.8772528007793474e-07,
+      "loss": 0.0003,
+      "step": 12872
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.8770092547491476e-07,
+      "loss": 0.0002,
+      "step": 12873
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.876765708718948e-07,
+      "loss": 0.0002,
+      "step": 12874
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.8765221626887481e-07,
+      "loss": 0.0003,
+      "step": 12875
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.8762786166585484e-07,
+      "loss": 0.0003,
+      "step": 12876
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 1.8760350706283486e-07,
+      "loss": 0.0005,
+      "step": 12877
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.875791524598149e-07,
+      "loss": 0.0003,
+      "step": 12878
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.8755479785679492e-07,
+      "loss": 0.0004,
+      "step": 12879
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.8753044325377497e-07,
+      "loss": 0.0003,
+      "step": 12880
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.87506088650755e-07,
+      "loss": 0.0002,
+      "step": 12881
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.8748173404773502e-07,
+      "loss": 0.0004,
+      "step": 12882
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.8745737944471504e-07,
+      "loss": 0.0003,
+      "step": 12883
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.8743302484169507e-07,
+      "loss": 0.0003,
+      "step": 12884
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.874086702386751e-07,
+      "loss": 0.0003,
+      "step": 12885
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.8738431563565512e-07,
+      "loss": 0.0004,
+      "step": 12886
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.8735996103263514e-07,
+      "loss": 0.0003,
+      "step": 12887
+    },
+    {
+      "epoch": 12.52,
+      "learning_rate": 1.873356064296152e-07,
+      "loss": 0.0003,
+      "step": 12888
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8731125182659522e-07,
+      "loss": 0.0004,
+      "step": 12889
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8728689722357524e-07,
+      "loss": 0.0004,
+      "step": 12890
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8726254262055527e-07,
+      "loss": 0.0002,
+      "step": 12891
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.872381880175353e-07,
+      "loss": 0.0004,
+      "step": 12892
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8721383341451532e-07,
+      "loss": 0.0004,
+      "step": 12893
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8718947881149535e-07,
+      "loss": 0.0002,
+      "step": 12894
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8716512420847537e-07,
+      "loss": 0.0004,
+      "step": 12895
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8714076960545542e-07,
+      "loss": 0.0002,
+      "step": 12896
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8711641500243545e-07,
+      "loss": 0.0003,
+      "step": 12897
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 1.8709206039941547e-07,
+      "loss": 0.0003,
+      "step": 12898
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.870677057963955e-07,
+      "loss": 0.0002,
+      "step": 12899
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.8704335119337552e-07,
+      "loss": 0.0003,
+      "step": 12900
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.8701899659035557e-07,
+      "loss": 0.0001,
+      "step": 12901
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.869946419873356e-07,
+      "loss": 0.0006,
+      "step": 12902
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.8697028738431565e-07,
+      "loss": 0.0003,
+      "step": 12903
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.8694593278129568e-07,
+      "loss": 0.0003,
+      "step": 12904
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.869215781782757e-07,
+      "loss": 0.0003,
+      "step": 12905
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.8689722357525573e-07,
+      "loss": 0.0003,
+      "step": 12906
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.8687286897223575e-07,
+      "loss": 0.0002,
+      "step": 12907
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 1.8684851436921578e-07,
+      "loss": 0.0004,
+      "step": 12908
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.868241597661958e-07,
+      "loss": 0.0003,
+      "step": 12909
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.8679980516317583e-07,
+      "loss": 0.0002,
+      "step": 12910
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.8677545056015588e-07,
+      "loss": 0.0002,
+      "step": 12911
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.867510959571359e-07,
+      "loss": 0.0005,
+      "step": 12912
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.8672674135411593e-07,
+      "loss": 0.0007,
+      "step": 12913
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.8670238675109595e-07,
+      "loss": 0.0004,
+      "step": 12914
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.8667803214807598e-07,
+      "loss": 0.0005,
+      "step": 12915
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.86653677545056e-07,
+      "loss": 0.0003,
+      "step": 12916
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.8662932294203603e-07,
+      "loss": 0.0003,
+      "step": 12917
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.8660496833901606e-07,
+      "loss": 0.0005,
+      "step": 12918
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 1.865806137359961e-07,
+      "loss": 0.0003,
+      "step": 12919
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8655625913297613e-07,
+      "loss": 0.0002,
+      "step": 12920
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8653190452995616e-07,
+      "loss": 0.0003,
+      "step": 12921
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8650754992693618e-07,
+      "loss": 0.0005,
+      "step": 12922
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.864831953239162e-07,
+      "loss": 0.0002,
+      "step": 12923
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8645884072089623e-07,
+      "loss": 0.0002,
+      "step": 12924
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8643448611787626e-07,
+      "loss": 0.0003,
+      "step": 12925
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8641013151485628e-07,
+      "loss": 0.0002,
+      "step": 12926
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8638577691183633e-07,
+      "loss": 0.0006,
+      "step": 12927
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8636142230881636e-07,
+      "loss": 0.0003,
+      "step": 12928
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 1.8633706770579639e-07,
+      "loss": 0.0002,
+      "step": 12929
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.863127131027764e-07,
+      "loss": 0.0002,
+      "step": 12930
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.8628835849975644e-07,
+      "loss": 0.0001,
+      "step": 12931
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.8626400389673646e-07,
+      "loss": 0.0003,
+      "step": 12932
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.8623964929371649e-07,
+      "loss": 0.0002,
+      "step": 12933
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.862152946906965e-07,
+      "loss": 0.0003,
+      "step": 12934
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.8619094008767656e-07,
+      "loss": 0.0002,
+      "step": 12935
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.861665854846566e-07,
+      "loss": 0.0002,
+      "step": 12936
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.861422308816366e-07,
+      "loss": 0.0008,
+      "step": 12937
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.8611787627861664e-07,
+      "loss": 0.0003,
+      "step": 12938
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 1.860935216755967e-07,
+      "loss": 0.0003,
+      "step": 12939
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.8606916707257671e-07,
+      "loss": 0.0004,
+      "step": 12940
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.8604481246955674e-07,
+      "loss": 0.0004,
+      "step": 12941
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.860204578665368e-07,
+      "loss": 0.0005,
+      "step": 12942
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.8599610326351682e-07,
+      "loss": 0.0004,
+      "step": 12943
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.8597174866049684e-07,
+      "loss": 0.0004,
+      "step": 12944
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.8594739405747687e-07,
+      "loss": 0.0003,
+      "step": 12945
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.859230394544569e-07,
+      "loss": 0.0003,
+      "step": 12946
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.8589868485143692e-07,
+      "loss": 0.0006,
+      "step": 12947
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.8587433024841694e-07,
+      "loss": 0.0005,
+      "step": 12948
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 1.8584997564539697e-07,
+      "loss": 0.0003,
+      "step": 12949
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.8582562104237702e-07,
+      "loss": 0.0002,
+      "step": 12950
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.8580126643935704e-07,
+      "loss": 0.0002,
+      "step": 12951
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.8577691183633707e-07,
+      "loss": 0.0005,
+      "step": 12952
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.857525572333171e-07,
+      "loss": 0.0002,
+      "step": 12953
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.8572820263029712e-07,
+      "loss": 0.0005,
+      "step": 12954
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.8570384802727715e-07,
+      "loss": 0.0004,
+      "step": 12955
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.8567949342425717e-07,
+      "loss": 0.0007,
+      "step": 12956
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.856551388212372e-07,
+      "loss": 0.0003,
+      "step": 12957
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.8563078421821725e-07,
+      "loss": 0.0003,
+      "step": 12958
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.8560642961519727e-07,
+      "loss": 0.0003,
+      "step": 12959
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 1.855820750121773e-07,
+      "loss": 0.0003,
+      "step": 12960
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.8555772040915732e-07,
+      "loss": 0.0002,
+      "step": 12961
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.8553336580613735e-07,
+      "loss": 0.0003,
+      "step": 12962
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.8550901120311737e-07,
+      "loss": 0.0004,
+      "step": 12963
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.854846566000974e-07,
+      "loss": 0.0003,
+      "step": 12964
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.8546030199707742e-07,
+      "loss": 0.0003,
+      "step": 12965
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.8543594739405747e-07,
+      "loss": 0.0002,
+      "step": 12966
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.854115927910375e-07,
+      "loss": 0.0002,
+      "step": 12967
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.8538723818801753e-07,
+      "loss": 0.0005,
+      "step": 12968
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.8536288358499755e-07,
+      "loss": 0.0003,
+      "step": 12969
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 1.8533852898197758e-07,
+      "loss": 0.0002,
+      "step": 12970
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.853141743789576e-07,
+      "loss": 0.0002,
+      "step": 12971
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.8528981977593763e-07,
+      "loss": 0.0003,
+      "step": 12972
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.8526546517291765e-07,
+      "loss": 0.0004,
+      "step": 12973
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.852411105698977e-07,
+      "loss": 0.0003,
+      "step": 12974
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.8521675596687773e-07,
+      "loss": 0.0002,
+      "step": 12975
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.8519240136385775e-07,
+      "loss": 0.0002,
+      "step": 12976
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.8516804676083778e-07,
+      "loss": 0.0003,
+      "step": 12977
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.8514369215781783e-07,
+      "loss": 0.0002,
+      "step": 12978
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.8511933755479786e-07,
+      "loss": 0.0002,
+      "step": 12979
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 1.8509498295177788e-07,
+      "loss": 0.0008,
+      "step": 12980
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.850706283487579e-07,
+      "loss": 0.0006,
+      "step": 12981
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.8504627374573796e-07,
+      "loss": 0.0015,
+      "step": 12982
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.8502191914271798e-07,
+      "loss": 0.0003,
+      "step": 12983
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.84997564539698e-07,
+      "loss": 0.0003,
+      "step": 12984
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.8497320993667803e-07,
+      "loss": 0.0003,
+      "step": 12985
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.8494885533365806e-07,
+      "loss": 0.0003,
+      "step": 12986
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.8492450073063808e-07,
+      "loss": 0.0002,
+      "step": 12987
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.849001461276181e-07,
+      "loss": 0.0003,
+      "step": 12988
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.8487579152459813e-07,
+      "loss": 0.0002,
+      "step": 12989
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.8485143692157818e-07,
+      "loss": 0.0003,
+      "step": 12990
+    },
+    {
+      "epoch": 12.62,
+      "learning_rate": 1.848270823185582e-07,
+      "loss": 0.0003,
+      "step": 12991
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.8480272771553824e-07,
+      "loss": 0.0005,
+      "step": 12992
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.8477837311251826e-07,
+      "loss": 0.0003,
+      "step": 12993
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.8475401850949829e-07,
+      "loss": 0.0003,
+      "step": 12994
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.847296639064783e-07,
+      "loss": 0.0005,
+      "step": 12995
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.8470530930345834e-07,
+      "loss": 0.0003,
+      "step": 12996
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.8468095470043836e-07,
+      "loss": 0.0003,
+      "step": 12997
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.846566000974184e-07,
+      "loss": 0.0005,
+      "step": 12998
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.8463224549439844e-07,
+      "loss": 0.0004,
+      "step": 12999
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.8460789089137846e-07,
+      "loss": 0.0002,
+      "step": 13000
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 1.845835362883585e-07,
+      "loss": 0.0004,
+      "step": 13001
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.8455918168533851e-07,
+      "loss": 0.0002,
+      "step": 13002
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.8453482708231854e-07,
+      "loss": 0.0002,
+      "step": 13003
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.8451047247929856e-07,
+      "loss": 0.0002,
+      "step": 13004
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.844861178762786e-07,
+      "loss": 0.0003,
+      "step": 13005
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.8446176327325864e-07,
+      "loss": 0.0003,
+      "step": 13006
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.8443740867023867e-07,
+      "loss": 0.0004,
+      "step": 13007
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.844130540672187e-07,
+      "loss": 0.0003,
+      "step": 13008
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.8438869946419872e-07,
+      "loss": 0.0003,
+      "step": 13009
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.8436434486117874e-07,
+      "loss": 0.0002,
+      "step": 13010
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 1.8433999025815877e-07,
+      "loss": 0.0003,
+      "step": 13011
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.843156356551388e-07,
+      "loss": 0.0004,
+      "step": 13012
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.8429128105211884e-07,
+      "loss": 0.0003,
+      "step": 13013
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.8426692644909887e-07,
+      "loss": 0.0003,
+      "step": 13014
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.842425718460789e-07,
+      "loss": 0.0004,
+      "step": 13015
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.8421821724305895e-07,
+      "loss": 0.0003,
+      "step": 13016
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.8419386264003897e-07,
+      "loss": 0.0002,
+      "step": 13017
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.84169508037019e-07,
+      "loss": 0.0003,
+      "step": 13018
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.8414515343399902e-07,
+      "loss": 0.0004,
+      "step": 13019
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.8412079883097905e-07,
+      "loss": 0.0005,
+      "step": 13020
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 1.840964442279591e-07,
+      "loss": 0.0003,
+      "step": 13021
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8407208962493912e-07,
+      "loss": 0.0004,
+      "step": 13022
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8404773502191915e-07,
+      "loss": 1.5393,
+      "step": 13023
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8402338041889917e-07,
+      "loss": 0.0003,
+      "step": 13024
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.839990258158792e-07,
+      "loss": 0.0003,
+      "step": 13025
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8397467121285922e-07,
+      "loss": 0.0002,
+      "step": 13026
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8395031660983925e-07,
+      "loss": 0.0005,
+      "step": 13027
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8392596200681927e-07,
+      "loss": 0.0004,
+      "step": 13028
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8390160740379933e-07,
+      "loss": 0.0003,
+      "step": 13029
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8387725280077935e-07,
+      "loss": 0.0004,
+      "step": 13030
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.8385289819775938e-07,
+      "loss": 0.0003,
+      "step": 13031
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 1.838285435947394e-07,
+      "loss": 0.0002,
+      "step": 13032
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.8380418899171943e-07,
+      "loss": 0.0003,
+      "step": 13033
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.8377983438869945e-07,
+      "loss": 0.0001,
+      "step": 13034
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.8375547978567948e-07,
+      "loss": 0.0003,
+      "step": 13035
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.837311251826595e-07,
+      "loss": 0.0003,
+      "step": 13036
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.8370677057963955e-07,
+      "loss": 0.0002,
+      "step": 13037
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.8368241597661958e-07,
+      "loss": 0.0003,
+      "step": 13038
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.836580613735996e-07,
+      "loss": 0.0002,
+      "step": 13039
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.8363370677057963e-07,
+      "loss": 0.0003,
+      "step": 13040
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.8360935216755965e-07,
+      "loss": 0.0002,
+      "step": 13041
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 1.8358499756453968e-07,
+      "loss": 0.0005,
+      "step": 13042
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.835606429615197e-07,
+      "loss": 0.0003,
+      "step": 13043
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.8353628835849973e-07,
+      "loss": 0.0001,
+      "step": 13044
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.8351193375547978e-07,
+      "loss": 0.0003,
+      "step": 13045
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.834875791524598e-07,
+      "loss": 0.0003,
+      "step": 13046
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.8346322454943983e-07,
+      "loss": 0.0003,
+      "step": 13047
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.8343886994641986e-07,
+      "loss": 0.0004,
+      "step": 13048
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.8341451534339988e-07,
+      "loss": 0.0003,
+      "step": 13049
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.833901607403799e-07,
+      "loss": 0.0004,
+      "step": 13050
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.8336580613735993e-07,
+      "loss": 0.0002,
+      "step": 13051
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 1.8334145153433996e-07,
+      "loss": 0.0012,
+      "step": 13052
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.8331709693132e-07,
+      "loss": 0.0003,
+      "step": 13053
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.8329274232830006e-07,
+      "loss": 0.0003,
+      "step": 13054
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.8326838772528009e-07,
+      "loss": 0.0003,
+      "step": 13055
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.832440331222601e-07,
+      "loss": 0.0002,
+      "step": 13056
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.8321967851924014e-07,
+      "loss": 0.0003,
+      "step": 13057
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.8319532391622016e-07,
+      "loss": 0.0004,
+      "step": 13058
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.8317096931320019e-07,
+      "loss": 0.0004,
+      "step": 13059
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.8314661471018024e-07,
+      "loss": 0.0004,
+      "step": 13060
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.8312226010716026e-07,
+      "loss": 0.0002,
+      "step": 13061
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.830979055041403e-07,
+      "loss": 0.0004,
+      "step": 13062
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 1.830735509011203e-07,
+      "loss": 0.0003,
+      "step": 13063
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.8304919629810034e-07,
+      "loss": 0.0002,
+      "step": 13064
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.8302484169508036e-07,
+      "loss": 0.0004,
+      "step": 13065
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.830004870920604e-07,
+      "loss": 0.0004,
+      "step": 13066
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.8297613248904041e-07,
+      "loss": 0.0002,
+      "step": 13067
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.8295177788602047e-07,
+      "loss": 0.0002,
+      "step": 13068
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.829274232830005e-07,
+      "loss": 0.0014,
+      "step": 13069
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.8290306867998052e-07,
+      "loss": 0.0003,
+      "step": 13070
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.8287871407696054e-07,
+      "loss": 0.0003,
+      "step": 13071
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.8285435947394057e-07,
+      "loss": 0.0006,
+      "step": 13072
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 1.828300048709206e-07,
+      "loss": 0.0004,
+      "step": 13073
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.8280565026790062e-07,
+      "loss": 0.0003,
+      "step": 13074
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.8278129566488064e-07,
+      "loss": 0.0009,
+      "step": 13075
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.827569410618607e-07,
+      "loss": 0.0005,
+      "step": 13076
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.8273258645884072e-07,
+      "loss": 0.0003,
+      "step": 13077
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.8270823185582074e-07,
+      "loss": 0.0004,
+      "step": 13078
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.8268387725280077e-07,
+      "loss": 0.0003,
+      "step": 13079
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.826595226497808e-07,
+      "loss": 0.0002,
+      "step": 13080
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.8263516804676082e-07,
+      "loss": 0.0002,
+      "step": 13081
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.8261081344374084e-07,
+      "loss": 0.0002,
+      "step": 13082
+    },
+    {
+      "epoch": 12.71,
+      "learning_rate": 1.8258645884072087e-07,
+      "loss": 0.0002,
+      "step": 13083
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.8256210423770092e-07,
+      "loss": 0.0002,
+      "step": 13084
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.8253774963468095e-07,
+      "loss": 0.0003,
+      "step": 13085
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.8251339503166097e-07,
+      "loss": 0.0002,
+      "step": 13086
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.82489040428641e-07,
+      "loss": 0.0005,
+      "step": 13087
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.8246468582562102e-07,
+      "loss": 0.0004,
+      "step": 13088
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.8244033122260105e-07,
+      "loss": 0.0003,
+      "step": 13089
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.8241597661958107e-07,
+      "loss": 0.0007,
+      "step": 13090
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.823916220165611e-07,
+      "loss": 0.0003,
+      "step": 13091
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.8236726741354118e-07,
+      "loss": 0.0003,
+      "step": 13092
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.823429128105212e-07,
+      "loss": 0.0002,
+      "step": 13093
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 1.8231855820750123e-07,
+      "loss": 0.0003,
+      "step": 13094
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.8229420360448125e-07,
+      "loss": 0.0003,
+      "step": 13095
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.8226984900146128e-07,
+      "loss": 0.0002,
+      "step": 13096
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.822454943984413e-07,
+      "loss": 0.0003,
+      "step": 13097
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.8222113979542133e-07,
+      "loss": 0.0005,
+      "step": 13098
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.8219678519240135e-07,
+      "loss": 0.0003,
+      "step": 13099
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.821724305893814e-07,
+      "loss": 0.0003,
+      "step": 13100
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.8214807598636143e-07,
+      "loss": 0.0003,
+      "step": 13101
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.8212372138334145e-07,
+      "loss": 0.0005,
+      "step": 13102
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.8209936678032148e-07,
+      "loss": 0.0005,
+      "step": 13103
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 1.820750121773015e-07,
+      "loss": 0.0005,
+      "step": 13104
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8205065757428153e-07,
+      "loss": 0.0002,
+      "step": 13105
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8202630297126155e-07,
+      "loss": 0.0004,
+      "step": 13106
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8200194836824158e-07,
+      "loss": 0.0003,
+      "step": 13107
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8197759376522163e-07,
+      "loss": 0.0002,
+      "step": 13108
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8195323916220166e-07,
+      "loss": 0.0005,
+      "step": 13109
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8192888455918168e-07,
+      "loss": 0.0002,
+      "step": 13110
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.819045299561617e-07,
+      "loss": 0.0002,
+      "step": 13111
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8188017535314173e-07,
+      "loss": 0.0002,
+      "step": 13112
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8185582075012176e-07,
+      "loss": 0.0002,
+      "step": 13113
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 1.8183146614710178e-07,
+      "loss": 0.0004,
+      "step": 13114
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.8180711154408183e-07,
+      "loss": 0.0003,
+      "step": 13115
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.8178275694106186e-07,
+      "loss": 0.0004,
+      "step": 13116
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.8175840233804188e-07,
+      "loss": 0.0003,
+      "step": 13117
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.817340477350219e-07,
+      "loss": 0.0002,
+      "step": 13118
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.8170969313200193e-07,
+      "loss": 0.0003,
+      "step": 13119
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.8168533852898196e-07,
+      "loss": 0.0112,
+      "step": 13120
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.8166098392596198e-07,
+      "loss": 0.0004,
+      "step": 13121
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.81636629322942e-07,
+      "loss": 0.0004,
+      "step": 13122
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.8161227471992206e-07,
+      "loss": 0.0004,
+      "step": 13123
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 1.8158792011690209e-07,
+      "loss": 0.0002,
+      "step": 13124
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.815635655138821e-07,
+      "loss": 0.0003,
+      "step": 13125
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.8153921091086214e-07,
+      "loss": 0.0002,
+      "step": 13126
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.8151485630784216e-07,
+      "loss": 0.0005,
+      "step": 13127
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.814905017048222e-07,
+      "loss": 0.0002,
+      "step": 13128
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.814661471018022e-07,
+      "loss": 0.0005,
+      "step": 13129
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.8144179249878224e-07,
+      "loss": 0.0003,
+      "step": 13130
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.8141743789576232e-07,
+      "loss": 0.0006,
+      "step": 13131
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.8139308329274234e-07,
+      "loss": 0.0003,
+      "step": 13132
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.8136872868972237e-07,
+      "loss": 0.0002,
+      "step": 13133
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.813443740867024e-07,
+      "loss": 1.8467,
+      "step": 13134
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 1.8132001948368242e-07,
+      "loss": 0.0003,
+      "step": 13135
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.8129566488066244e-07,
+      "loss": 0.0002,
+      "step": 13136
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.8127131027764247e-07,
+      "loss": 0.0005,
+      "step": 13137
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.812469556746225e-07,
+      "loss": 0.0006,
+      "step": 13138
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.8122260107160254e-07,
+      "loss": 0.0002,
+      "step": 13139
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.8119824646858257e-07,
+      "loss": 0.0002,
+      "step": 13140
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.811738918655626e-07,
+      "loss": 0.0004,
+      "step": 13141
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.8114953726254262e-07,
+      "loss": 0.0002,
+      "step": 13142
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.8112518265952264e-07,
+      "loss": 0.0002,
+      "step": 13143
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.8110082805650267e-07,
+      "loss": 0.0004,
+      "step": 13144
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 1.810764734534827e-07,
+      "loss": 0.0002,
+      "step": 13145
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.8105211885046272e-07,
+      "loss": 0.0003,
+      "step": 13146
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.8102776424744277e-07,
+      "loss": 0.0003,
+      "step": 13147
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.810034096444228e-07,
+      "loss": 0.0002,
+      "step": 13148
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.8097905504140282e-07,
+      "loss": 0.0003,
+      "step": 13149
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.8095470043838285e-07,
+      "loss": 0.0003,
+      "step": 13150
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.8093034583536287e-07,
+      "loss": 0.0004,
+      "step": 13151
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.809059912323429e-07,
+      "loss": 0.0003,
+      "step": 13152
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.8088163662932292e-07,
+      "loss": 0.0002,
+      "step": 13153
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.8085728202630295e-07,
+      "loss": 0.0002,
+      "step": 13154
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 1.80832927423283e-07,
+      "loss": 0.0003,
+      "step": 13155
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8080857282026302e-07,
+      "loss": 0.0004,
+      "step": 13156
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8078421821724305e-07,
+      "loss": 0.0003,
+      "step": 13157
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8075986361422307e-07,
+      "loss": 0.0004,
+      "step": 13158
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.807355090112031e-07,
+      "loss": 0.0005,
+      "step": 13159
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8071115440818312e-07,
+      "loss": 0.0004,
+      "step": 13160
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8068679980516315e-07,
+      "loss": 0.0002,
+      "step": 13161
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8066244520214318e-07,
+      "loss": 0.0003,
+      "step": 13162
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8063809059912323e-07,
+      "loss": 0.0003,
+      "step": 13163
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8061373599610325e-07,
+      "loss": 0.0005,
+      "step": 13164
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.8058938139308328e-07,
+      "loss": 0.0002,
+      "step": 13165
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 1.805650267900633e-07,
+      "loss": 0.0002,
+      "step": 13166
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.8054067218704333e-07,
+      "loss": 0.0004,
+      "step": 13167
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.8051631758402335e-07,
+      "loss": 0.0002,
+      "step": 13168
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.804919629810034e-07,
+      "loss": 0.0004,
+      "step": 13169
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.8046760837798346e-07,
+      "loss": 0.0003,
+      "step": 13170
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.8044325377496348e-07,
+      "loss": 0.0003,
+      "step": 13171
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.804188991719435e-07,
+      "loss": 0.0004,
+      "step": 13172
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.8039454456892353e-07,
+      "loss": 0.0004,
+      "step": 13173
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.8037018996590356e-07,
+      "loss": 0.0002,
+      "step": 13174
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.8034583536288358e-07,
+      "loss": 0.0004,
+      "step": 13175
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 1.803214807598636e-07,
+      "loss": 0.0004,
+      "step": 13176
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.8029712615684363e-07,
+      "loss": 0.0018,
+      "step": 13177
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.8027277155382368e-07,
+      "loss": 0.0005,
+      "step": 13178
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.802484169508037e-07,
+      "loss": 0.0003,
+      "step": 13179
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.8022406234778373e-07,
+      "loss": 0.0003,
+      "step": 13180
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.8019970774476376e-07,
+      "loss": 0.0003,
+      "step": 13181
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.8017535314174378e-07,
+      "loss": 0.0003,
+      "step": 13182
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.801509985387238e-07,
+      "loss": 0.0003,
+      "step": 13183
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.8012664393570383e-07,
+      "loss": 0.0004,
+      "step": 13184
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.8010228933268386e-07,
+      "loss": 0.0006,
+      "step": 13185
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 1.800779347296639e-07,
+      "loss": 0.0003,
+      "step": 13186
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.8005358012664394e-07,
+      "loss": 0.0003,
+      "step": 13187
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.8002922552362396e-07,
+      "loss": 0.0005,
+      "step": 13188
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.80004870920604e-07,
+      "loss": 0.0003,
+      "step": 13189
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.79980516317584e-07,
+      "loss": 0.0003,
+      "step": 13190
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.7995616171456404e-07,
+      "loss": 0.0004,
+      "step": 13191
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.7993180711154406e-07,
+      "loss": 0.0006,
+      "step": 13192
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.799074525085241e-07,
+      "loss": 0.0004,
+      "step": 13193
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.7988309790550414e-07,
+      "loss": 0.0004,
+      "step": 13194
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.7985874330248416e-07,
+      "loss": 0.0002,
+      "step": 13195
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 1.798343886994642e-07,
+      "loss": 0.0005,
+      "step": 13196
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7981003409644421e-07,
+      "loss": 0.0004,
+      "step": 13197
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7978567949342424e-07,
+      "loss": 0.0003,
+      "step": 13198
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7976132489040426e-07,
+      "loss": 2.0649,
+      "step": 13199
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.797369702873843e-07,
+      "loss": 0.0004,
+      "step": 13200
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7971261568436432e-07,
+      "loss": 0.0003,
+      "step": 13201
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7968826108134437e-07,
+      "loss": 0.0005,
+      "step": 13202
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.796639064783244e-07,
+      "loss": 0.0001,
+      "step": 13203
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7963955187530442e-07,
+      "loss": 0.0003,
+      "step": 13204
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7961519727228444e-07,
+      "loss": 0.0002,
+      "step": 13205
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7959084266926447e-07,
+      "loss": 0.0003,
+      "step": 13206
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 1.7956648806624452e-07,
+      "loss": 0.0004,
+      "step": 13207
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.7954213346322454e-07,
+      "loss": 0.0004,
+      "step": 13208
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.795177788602046e-07,
+      "loss": 0.0002,
+      "step": 13209
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.7949342425718462e-07,
+      "loss": 0.0016,
+      "step": 13210
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.7946906965416465e-07,
+      "loss": 0.0002,
+      "step": 13211
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.7944471505114467e-07,
+      "loss": 0.0003,
+      "step": 13212
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.794203604481247e-07,
+      "loss": 0.0005,
+      "step": 13213
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.7939600584510472e-07,
+      "loss": 0.0006,
+      "step": 13214
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.7937165124208475e-07,
+      "loss": 0.0002,
+      "step": 13215
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.7934729663906477e-07,
+      "loss": 0.0003,
+      "step": 13216
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 1.7932294203604482e-07,
+      "loss": 0.0003,
+      "step": 13217
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.7929858743302485e-07,
+      "loss": 0.0005,
+      "step": 13218
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.7927423283000487e-07,
+      "loss": 0.0002,
+      "step": 13219
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.792498782269849e-07,
+      "loss": 0.0002,
+      "step": 13220
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.7922552362396492e-07,
+      "loss": 0.0002,
+      "step": 13221
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.7920116902094495e-07,
+      "loss": 0.0004,
+      "step": 13222
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.7917681441792497e-07,
+      "loss": 0.0002,
+      "step": 13223
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.79152459814905e-07,
+      "loss": 0.0002,
+      "step": 13224
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.7912810521188505e-07,
+      "loss": 0.0005,
+      "step": 13225
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.7910375060886508e-07,
+      "loss": 0.0003,
+      "step": 13226
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 1.790793960058451e-07,
+      "loss": 0.0002,
+      "step": 13227
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.7905504140282513e-07,
+      "loss": 0.0002,
+      "step": 13228
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.7903068679980515e-07,
+      "loss": 0.0003,
+      "step": 13229
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.7900633219678518e-07,
+      "loss": 0.0006,
+      "step": 13230
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.789819775937652e-07,
+      "loss": 0.0003,
+      "step": 13231
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.7895762299074523e-07,
+      "loss": 0.0002,
+      "step": 13232
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.7893326838772528e-07,
+      "loss": 0.0003,
+      "step": 13233
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.789089137847053e-07,
+      "loss": 0.0005,
+      "step": 13234
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.7888455918168533e-07,
+      "loss": 0.0003,
+      "step": 13235
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.7886020457866535e-07,
+      "loss": 0.0002,
+      "step": 13236
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.7883584997564538e-07,
+      "loss": 0.0004,
+      "step": 13237
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 1.788114953726254e-07,
+      "loss": 0.0003,
+      "step": 13238
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.7878714076960543e-07,
+      "loss": 0.0002,
+      "step": 13239
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.7876278616658546e-07,
+      "loss": 0.0003,
+      "step": 13240
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.787384315635655e-07,
+      "loss": 0.0005,
+      "step": 13241
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.7871407696054553e-07,
+      "loss": 0.0002,
+      "step": 13242
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.7868972235752556e-07,
+      "loss": 0.0002,
+      "step": 13243
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.7866536775450558e-07,
+      "loss": 0.0003,
+      "step": 13244
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.786410131514856e-07,
+      "loss": 0.0004,
+      "step": 13245
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.7861665854846566e-07,
+      "loss": 0.0003,
+      "step": 13246
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.7859230394544568e-07,
+      "loss": 0.0003,
+      "step": 13247
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 1.785679493424257e-07,
+      "loss": 0.0004,
+      "step": 13248
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.7854359473940576e-07,
+      "loss": 0.0003,
+      "step": 13249
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.7851924013638579e-07,
+      "loss": 0.0004,
+      "step": 13250
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.784948855333658e-07,
+      "loss": 0.0004,
+      "step": 13251
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.7847053093034584e-07,
+      "loss": 0.0003,
+      "step": 13252
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.7844617632732586e-07,
+      "loss": 0.0003,
+      "step": 13253
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.784218217243059e-07,
+      "loss": 0.0003,
+      "step": 13254
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.783974671212859e-07,
+      "loss": 0.0002,
+      "step": 13255
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.7837311251826594e-07,
+      "loss": 0.0002,
+      "step": 13256
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.78348757915246e-07,
+      "loss": 0.0003,
+      "step": 13257
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 1.7832440331222601e-07,
+      "loss": 0.0002,
+      "step": 13258
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.7830004870920604e-07,
+      "loss": 0.0003,
+      "step": 13259
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.7827569410618606e-07,
+      "loss": 0.0003,
+      "step": 13260
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.782513395031661e-07,
+      "loss": 0.0002,
+      "step": 13261
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.7822698490014612e-07,
+      "loss": 0.0004,
+      "step": 13262
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.7820263029712614e-07,
+      "loss": 0.0002,
+      "step": 13263
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.7817827569410617e-07,
+      "loss": 0.0003,
+      "step": 13264
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.7815392109108622e-07,
+      "loss": 0.0002,
+      "step": 13265
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.7812956648806624e-07,
+      "loss": 0.0003,
+      "step": 13266
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.7810521188504627e-07,
+      "loss": 0.0004,
+      "step": 13267
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 1.780808572820263e-07,
+      "loss": 0.0003,
+      "step": 13268
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.7805650267900632e-07,
+      "loss": 0.0004,
+      "step": 13269
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.7803214807598634e-07,
+      "loss": 0.0002,
+      "step": 13270
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.7800779347296637e-07,
+      "loss": 0.0003,
+      "step": 13271
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.779834388699464e-07,
+      "loss": 0.0003,
+      "step": 13272
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.7795908426692644e-07,
+      "loss": 0.0003,
+      "step": 13273
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.7793472966390647e-07,
+      "loss": 0.0002,
+      "step": 13274
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.779103750608865e-07,
+      "loss": 0.0003,
+      "step": 13275
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.7788602045786652e-07,
+      "loss": 0.0002,
+      "step": 13276
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.7786166585484655e-07,
+      "loss": 0.0002,
+      "step": 13277
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.7783731125182657e-07,
+      "loss": 0.5166,
+      "step": 13278
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 1.778129566488066e-07,
+      "loss": 0.0003,
+      "step": 13279
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.7778860204578665e-07,
+      "loss": 0.0003,
+      "step": 13280
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.7776424744276667e-07,
+      "loss": 0.0003,
+      "step": 13281
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.777398928397467e-07,
+      "loss": 0.0003,
+      "step": 13282
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.7771553823672672e-07,
+      "loss": 0.0004,
+      "step": 13283
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.7769118363370677e-07,
+      "loss": 0.0002,
+      "step": 13284
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.776668290306868e-07,
+      "loss": 0.0004,
+      "step": 13285
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.7764247442766682e-07,
+      "loss": 0.0003,
+      "step": 13286
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.7761811982464685e-07,
+      "loss": 0.0002,
+      "step": 13287
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.775937652216269e-07,
+      "loss": 1.9388,
+      "step": 13288
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 1.7756941061860693e-07,
+      "loss": 0.0003,
+      "step": 13289
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.7754505601558695e-07,
+      "loss": 0.0003,
+      "step": 13290
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.7752070141256698e-07,
+      "loss": 0.0003,
+      "step": 13291
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.77496346809547e-07,
+      "loss": 0.0002,
+      "step": 13292
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.7747199220652703e-07,
+      "loss": 0.0003,
+      "step": 13293
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.7744763760350705e-07,
+      "loss": 0.0002,
+      "step": 13294
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.7742328300048708e-07,
+      "loss": 0.0002,
+      "step": 13295
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.7739892839746713e-07,
+      "loss": 0.0003,
+      "step": 13296
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.7737457379444715e-07,
+      "loss": 0.0003,
+      "step": 13297
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.7735021919142718e-07,
+      "loss": 0.0004,
+      "step": 13298
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 1.773258645884072e-07,
+      "loss": 0.0003,
+      "step": 13299
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.7730150998538723e-07,
+      "loss": 0.0003,
+      "step": 13300
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.7727715538236726e-07,
+      "loss": 0.0003,
+      "step": 13301
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.7725280077934728e-07,
+      "loss": 0.0002,
+      "step": 13302
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.772284461763273e-07,
+      "loss": 0.0002,
+      "step": 13303
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.7720409157330736e-07,
+      "loss": 0.0003,
+      "step": 13304
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.7717973697028738e-07,
+      "loss": 0.0003,
+      "step": 13305
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.771553823672674e-07,
+      "loss": 0.0005,
+      "step": 13306
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.7713102776424743e-07,
+      "loss": 0.0004,
+      "step": 13307
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.7710667316122746e-07,
+      "loss": 0.0002,
+      "step": 13308
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.7708231855820748e-07,
+      "loss": 0.0002,
+      "step": 13309
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 1.770579639551875e-07,
+      "loss": 0.0004,
+      "step": 13310
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.7703360935216753e-07,
+      "loss": 0.0004,
+      "step": 13311
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.7700925474914759e-07,
+      "loss": 0.0005,
+      "step": 13312
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.769849001461276e-07,
+      "loss": 0.0003,
+      "step": 13313
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.7696054554310764e-07,
+      "loss": 0.0002,
+      "step": 13314
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.7693619094008766e-07,
+      "loss": 0.0004,
+      "step": 13315
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.7691183633706769e-07,
+      "loss": 0.0003,
+      "step": 13316
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.768874817340477e-07,
+      "loss": 0.0004,
+      "step": 13317
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.7686312713102774e-07,
+      "loss": 0.0005,
+      "step": 13318
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.7683877252800776e-07,
+      "loss": 0.0003,
+      "step": 13319
+    },
+    {
+      "epoch": 12.94,
+      "learning_rate": 1.768144179249878e-07,
+      "loss": 0.0004,
+      "step": 13320
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.7679006332196784e-07,
+      "loss": 0.0002,
+      "step": 13321
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.767657087189479e-07,
+      "loss": 0.0004,
+      "step": 13322
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.7674135411592791e-07,
+      "loss": 0.0003,
+      "step": 13323
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.7671699951290794e-07,
+      "loss": 0.0003,
+      "step": 13324
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.7669264490988797e-07,
+      "loss": 0.0004,
+      "step": 13325
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.76668290306868e-07,
+      "loss": 0.0003,
+      "step": 13326
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.7664393570384804e-07,
+      "loss": 0.0003,
+      "step": 13327
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.7661958110082807e-07,
+      "loss": 0.0003,
+      "step": 13328
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.765952264978081e-07,
+      "loss": 0.0003,
+      "step": 13329
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 1.7657087189478812e-07,
+      "loss": 0.0003,
+      "step": 13330
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.7654651729176814e-07,
+      "loss": 0.0002,
+      "step": 13331
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.7652216268874817e-07,
+      "loss": 0.0002,
+      "step": 13332
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.764978080857282e-07,
+      "loss": 0.0002,
+      "step": 13333
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.7647345348270822e-07,
+      "loss": 0.0003,
+      "step": 13334
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.7644909887968827e-07,
+      "loss": 0.0002,
+      "step": 13335
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.764247442766683e-07,
+      "loss": 0.0004,
+      "step": 13336
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.7640038967364832e-07,
+      "loss": 0.0004,
+      "step": 13337
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.7637603507062835e-07,
+      "loss": 0.0003,
+      "step": 13338
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.7635168046760837e-07,
+      "loss": 0.0004,
+      "step": 13339
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 1.763273258645884e-07,
+      "loss": 0.0002,
+      "step": 13340
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7630297126156842e-07,
+      "loss": 0.0002,
+      "step": 13341
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7627861665854845e-07,
+      "loss": 0.0005,
+      "step": 13342
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.762542620555285e-07,
+      "loss": 0.0002,
+      "step": 13343
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7622990745250852e-07,
+      "loss": 0.0002,
+      "step": 13344
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7620555284948855e-07,
+      "loss": 0.0004,
+      "step": 13345
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7618119824646857e-07,
+      "loss": 0.0002,
+      "step": 13346
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.761568436434486e-07,
+      "loss": 0.0004,
+      "step": 13347
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7613248904042862e-07,
+      "loss": 0.0001,
+      "step": 13348
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7610813443740865e-07,
+      "loss": 0.0005,
+      "step": 13349
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7608377983438867e-07,
+      "loss": 0.0002,
+      "step": 13350
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 1.7605942523136873e-07,
+      "loss": 0.0003,
+      "step": 13351
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.7603507062834875e-07,
+      "loss": 0.0003,
+      "step": 13352
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.7601071602532878e-07,
+      "loss": 0.0004,
+      "step": 13353
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.759863614223088e-07,
+      "loss": 0.0002,
+      "step": 13354
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.7596200681928883e-07,
+      "loss": 0.0002,
+      "step": 13355
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.7593765221626885e-07,
+      "loss": 0.0003,
+      "step": 13356
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.7591329761324888e-07,
+      "loss": 0.0004,
+      "step": 13357
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.758889430102289e-07,
+      "loss": 0.0003,
+      "step": 13358
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.7586458840720895e-07,
+      "loss": 0.0006,
+      "step": 13359
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.7584023380418898e-07,
+      "loss": 0.0004,
+      "step": 13360
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 1.7581587920116903e-07,
+      "loss": 0.0003,
+      "step": 13361
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.7579152459814906e-07,
+      "loss": 0.0002,
+      "step": 13362
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.7576716999512908e-07,
+      "loss": 0.0003,
+      "step": 13363
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.757428153921091e-07,
+      "loss": 0.0003,
+      "step": 13364
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.7571846078908913e-07,
+      "loss": 0.0004,
+      "step": 13365
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.7569410618606916e-07,
+      "loss": 0.0003,
+      "step": 13366
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.756697515830492e-07,
+      "loss": 0.0001,
+      "step": 13367
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.7564539698002923e-07,
+      "loss": 0.0003,
+      "step": 13368
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.7562104237700926e-07,
+      "loss": 0.0003,
+      "step": 13369
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.7559668777398928e-07,
+      "loss": 0.0004,
+      "step": 13370
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 1.755723331709693e-07,
+      "loss": 0.0007,
+      "step": 13371
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7554797856794933e-07,
+      "loss": 0.0003,
+      "step": 13372
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7552362396492936e-07,
+      "loss": 0.0006,
+      "step": 13373
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7549926936190938e-07,
+      "loss": 0.0002,
+      "step": 13374
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7547491475888944e-07,
+      "loss": 0.0003,
+      "step": 13375
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7545056015586946e-07,
+      "loss": 0.0004,
+      "step": 13376
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7542620555284949e-07,
+      "loss": 0.0004,
+      "step": 13377
+    },
+    {
+      "epoch": 13.0,
+      "eval_accuracy": 0.9980554205153136,
+      "eval_loss": 0.015082407742738724,
+      "eval_runtime": 2.7835,
+      "eval_samples_per_second": 738.995,
+      "eval_steps_per_second": 11.856,
+      "step": 13377
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.754018509498295e-07,
+      "loss": 0.0003,
+      "step": 13378
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7537749634680954e-07,
+      "loss": 0.0003,
+      "step": 13379
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7535314174378956e-07,
+      "loss": 0.0003,
+      "step": 13380
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7532878714076959e-07,
+      "loss": 0.0004,
+      "step": 13381
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 1.7530443253774964e-07,
+      "loss": 0.0004,
+      "step": 13382
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.7528007793472966e-07,
+      "loss": 0.0003,
+      "step": 13383
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.752557233317097e-07,
+      "loss": 0.0003,
+      "step": 13384
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.7523136872868971e-07,
+      "loss": 0.0003,
+      "step": 13385
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.7520701412566974e-07,
+      "loss": 0.0005,
+      "step": 13386
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.7518265952264976e-07,
+      "loss": 0.0003,
+      "step": 13387
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.751583049196298e-07,
+      "loss": 0.0004,
+      "step": 13388
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.7513395031660981e-07,
+      "loss": 0.0086,
+      "step": 13389
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.7510959571358987e-07,
+      "loss": 0.0003,
+      "step": 13390
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.750852411105699e-07,
+      "loss": 0.0004,
+      "step": 13391
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 1.7506088650754992e-07,
+      "loss": 0.0002,
+      "step": 13392
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.7503653190452994e-07,
+      "loss": 0.0008,
+      "step": 13393
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.7501217730150997e-07,
+      "loss": 0.0003,
+      "step": 13394
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.7498782269849e-07,
+      "loss": 0.0003,
+      "step": 13395
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.7496346809547002e-07,
+      "loss": 0.0003,
+      "step": 13396
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.7493911349245004e-07,
+      "loss": 0.0002,
+      "step": 13397
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.749147588894301e-07,
+      "loss": 0.0005,
+      "step": 13398
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.7489040428641015e-07,
+      "loss": 0.0002,
+      "step": 13399
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.7486604968339017e-07,
+      "loss": 0.0002,
+      "step": 13400
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.748416950803702e-07,
+      "loss": 0.0003,
+      "step": 13401
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 1.7481734047735022e-07,
+      "loss": 0.0002,
+      "step": 13402
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.7479298587433025e-07,
+      "loss": 0.0003,
+      "step": 13403
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.7476863127131027e-07,
+      "loss": 0.0003,
+      "step": 13404
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.747442766682903e-07,
+      "loss": 1.662,
+      "step": 13405
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.7471992206527035e-07,
+      "loss": 0.0002,
+      "step": 13406
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.7469556746225037e-07,
+      "loss": 0.0003,
+      "step": 13407
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.746712128592304e-07,
+      "loss": 0.0002,
+      "step": 13408
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.7464685825621042e-07,
+      "loss": 0.0005,
+      "step": 13409
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.7462250365319045e-07,
+      "loss": 0.0003,
+      "step": 13410
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.7459814905017047e-07,
+      "loss": 0.0004,
+      "step": 13411
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.745737944471505e-07,
+      "loss": 0.0003,
+      "step": 13412
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 1.7454943984413052e-07,
+      "loss": 0.0004,
+      "step": 13413
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.7452508524111058e-07,
+      "loss": 0.0004,
+      "step": 13414
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.745007306380906e-07,
+      "loss": 0.0007,
+      "step": 13415
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.7447637603507063e-07,
+      "loss": 0.0005,
+      "step": 13416
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.7445202143205065e-07,
+      "loss": 0.0002,
+      "step": 13417
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.7442766682903068e-07,
+      "loss": 0.3083,
+      "step": 13418
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.744033122260107e-07,
+      "loss": 0.0004,
+      "step": 13419
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.7437895762299073e-07,
+      "loss": 0.0003,
+      "step": 13420
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.7435460301997075e-07,
+      "loss": 0.0003,
+      "step": 13421
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.743302484169508e-07,
+      "loss": 0.0001,
+      "step": 13422
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 1.7430589381393083e-07,
+      "loss": 0.0004,
+      "step": 13423
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.7428153921091085e-07,
+      "loss": 0.0002,
+      "step": 13424
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.7425718460789088e-07,
+      "loss": 0.0004,
+      "step": 13425
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.742328300048709e-07,
+      "loss": 0.0003,
+      "step": 13426
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.7420847540185093e-07,
+      "loss": 0.0003,
+      "step": 13427
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.7418412079883095e-07,
+      "loss": 0.0003,
+      "step": 13428
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.7415976619581098e-07,
+      "loss": 0.0003,
+      "step": 13429
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.7413541159279103e-07,
+      "loss": 0.0006,
+      "step": 13430
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.7411105698977106e-07,
+      "loss": 0.1635,
+      "step": 13431
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.7408670238675108e-07,
+      "loss": 0.0002,
+      "step": 13432
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 1.740623477837311e-07,
+      "loss": 0.0003,
+      "step": 13433
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.7403799318071113e-07,
+      "loss": 0.0001,
+      "step": 13434
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.7401363857769116e-07,
+      "loss": 0.0002,
+      "step": 13435
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.7398928397467118e-07,
+      "loss": 0.0003,
+      "step": 13436
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.7396492937165126e-07,
+      "loss": 0.0002,
+      "step": 13437
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.7394057476863129e-07,
+      "loss": 0.0003,
+      "step": 13438
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.739162201656113e-07,
+      "loss": 0.0002,
+      "step": 13439
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.7389186556259134e-07,
+      "loss": 0.0002,
+      "step": 13440
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.7386751095957136e-07,
+      "loss": 0.0005,
+      "step": 13441
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.7384315635655139e-07,
+      "loss": 0.0003,
+      "step": 13442
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 1.738188017535314e-07,
+      "loss": 0.0002,
+      "step": 13443
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.7379444715051144e-07,
+      "loss": 0.0003,
+      "step": 13444
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.737700925474915e-07,
+      "loss": 0.0005,
+      "step": 13445
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.737457379444715e-07,
+      "loss": 0.0003,
+      "step": 13446
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.7372138334145154e-07,
+      "loss": 0.0003,
+      "step": 13447
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.7369702873843156e-07,
+      "loss": 0.0004,
+      "step": 13448
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.736726741354116e-07,
+      "loss": 0.0004,
+      "step": 13449
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.7364831953239161e-07,
+      "loss": 0.0002,
+      "step": 13450
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.7362396492937164e-07,
+      "loss": 0.0003,
+      "step": 13451
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.7359961032635166e-07,
+      "loss": 0.0003,
+      "step": 13452
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.7357525572333172e-07,
+      "loss": 0.0002,
+      "step": 13453
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 1.7355090112031174e-07,
+      "loss": 0.0003,
+      "step": 13454
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.7352654651729177e-07,
+      "loss": 0.0003,
+      "step": 13455
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.735021919142718e-07,
+      "loss": 0.0004,
+      "step": 13456
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.7347783731125182e-07,
+      "loss": 0.0002,
+      "step": 13457
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.7345348270823184e-07,
+      "loss": 0.0003,
+      "step": 13458
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.7342912810521187e-07,
+      "loss": 0.0003,
+      "step": 13459
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.734047735021919e-07,
+      "loss": 0.0004,
+      "step": 13460
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.7338041889917194e-07,
+      "loss": 0.0002,
+      "step": 13461
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.7335606429615197e-07,
+      "loss": 0.0003,
+      "step": 13462
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.73331709693132e-07,
+      "loss": 0.0003,
+      "step": 13463
+    },
+    {
+      "epoch": 13.08,
+      "learning_rate": 1.7330735509011202e-07,
+      "loss": 0.0003,
+      "step": 13464
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.7328300048709204e-07,
+      "loss": 0.0004,
+      "step": 13465
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.7325864588407207e-07,
+      "loss": 0.0003,
+      "step": 13466
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.732342912810521e-07,
+      "loss": 0.0003,
+      "step": 13467
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.7320993667803212e-07,
+      "loss": 0.0002,
+      "step": 13468
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.7318558207501217e-07,
+      "loss": 0.0003,
+      "step": 13469
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.731612274719922e-07,
+      "loss": 0.0004,
+      "step": 13470
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.7313687286897222e-07,
+      "loss": 0.0004,
+      "step": 13471
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.7311251826595225e-07,
+      "loss": 0.0002,
+      "step": 13472
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.7308816366293227e-07,
+      "loss": 0.0003,
+      "step": 13473
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 1.730638090599123e-07,
+      "loss": 0.0002,
+      "step": 13474
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7303945445689235e-07,
+      "loss": 0.0002,
+      "step": 13475
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7301509985387237e-07,
+      "loss": 0.0003,
+      "step": 13476
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7299074525085243e-07,
+      "loss": 0.0003,
+      "step": 13477
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7296639064783245e-07,
+      "loss": 0.0022,
+      "step": 13478
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7294203604481248e-07,
+      "loss": 0.0003,
+      "step": 13479
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.729176814417925e-07,
+      "loss": 0.0004,
+      "step": 13480
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7289332683877253e-07,
+      "loss": 0.0003,
+      "step": 13481
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7286897223575255e-07,
+      "loss": 0.0003,
+      "step": 13482
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7284461763273258e-07,
+      "loss": 0.0002,
+      "step": 13483
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7282026302971263e-07,
+      "loss": 0.0002,
+      "step": 13484
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 1.7279590842669265e-07,
+      "loss": 0.0004,
+      "step": 13485
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.7277155382367268e-07,
+      "loss": 0.0003,
+      "step": 13486
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.727471992206527e-07,
+      "loss": 0.0003,
+      "step": 13487
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.7272284461763273e-07,
+      "loss": 0.0003,
+      "step": 13488
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.7269849001461275e-07,
+      "loss": 0.0009,
+      "step": 13489
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.7267413541159278e-07,
+      "loss": 0.0002,
+      "step": 13490
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.726497808085728e-07,
+      "loss": 0.0003,
+      "step": 13491
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.7262542620555286e-07,
+      "loss": 0.0002,
+      "step": 13492
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.7260107160253288e-07,
+      "loss": 0.0003,
+      "step": 13493
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.725767169995129e-07,
+      "loss": 0.0005,
+      "step": 13494
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 1.7255236239649293e-07,
+      "loss": 0.0002,
+      "step": 13495
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.7252800779347296e-07,
+      "loss": 0.0005,
+      "step": 13496
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.7250365319045298e-07,
+      "loss": 0.0002,
+      "step": 13497
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.72479298587433e-07,
+      "loss": 0.0003,
+      "step": 13498
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.7245494398441303e-07,
+      "loss": 0.0002,
+      "step": 13499
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.7243058938139308e-07,
+      "loss": 0.0003,
+      "step": 13500
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.724062347783731e-07,
+      "loss": 0.0002,
+      "step": 13501
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.7238188017535313e-07,
+      "loss": 0.0004,
+      "step": 13502
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.7235752557233316e-07,
+      "loss": 0.0003,
+      "step": 13503
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.7233317096931318e-07,
+      "loss": 0.0003,
+      "step": 13504
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 1.723088163662932e-07,
+      "loss": 0.0002,
+      "step": 13505
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.7228446176327323e-07,
+      "loss": 0.0004,
+      "step": 13506
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.7226010716025326e-07,
+      "loss": 0.0002,
+      "step": 13507
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.722357525572333e-07,
+      "loss": 0.0002,
+      "step": 13508
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.7221139795421334e-07,
+      "loss": 0.0004,
+      "step": 13509
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.7218704335119336e-07,
+      "loss": 0.0002,
+      "step": 13510
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.721626887481734e-07,
+      "loss": 0.0002,
+      "step": 13511
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.721383341451534e-07,
+      "loss": 0.0002,
+      "step": 13512
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.7211397954213344e-07,
+      "loss": 0.0002,
+      "step": 13513
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.720896249391135e-07,
+      "loss": 0.0004,
+      "step": 13514
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 1.7206527033609351e-07,
+      "loss": 0.0003,
+      "step": 13515
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.7204091573307357e-07,
+      "loss": 0.0002,
+      "step": 13516
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.720165611300536e-07,
+      "loss": 0.0002,
+      "step": 13517
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.7199220652703362e-07,
+      "loss": 0.0004,
+      "step": 13518
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.7196785192401364e-07,
+      "loss": 0.0001,
+      "step": 13519
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.7194349732099367e-07,
+      "loss": 0.0003,
+      "step": 13520
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.719191427179737e-07,
+      "loss": 0.0002,
+      "step": 13521
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.7189478811495372e-07,
+      "loss": 0.0001,
+      "step": 13522
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.7187043351193374e-07,
+      "loss": 0.0002,
+      "step": 13523
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.718460789089138e-07,
+      "loss": 0.0002,
+      "step": 13524
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.7182172430589382e-07,
+      "loss": 0.0005,
+      "step": 13525
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 1.7179736970287384e-07,
+      "loss": 0.0003,
+      "step": 13526
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.7177301509985387e-07,
+      "loss": 0.0003,
+      "step": 13527
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.717486604968339e-07,
+      "loss": 0.0002,
+      "step": 13528
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.7172430589381392e-07,
+      "loss": 0.0003,
+      "step": 13529
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.7169995129079394e-07,
+      "loss": 0.0002,
+      "step": 13530
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.7167559668777397e-07,
+      "loss": 0.0004,
+      "step": 13531
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.7165124208475402e-07,
+      "loss": 0.0002,
+      "step": 13532
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.7162688748173405e-07,
+      "loss": 0.0002,
+      "step": 13533
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.7160253287871407e-07,
+      "loss": 0.0004,
+      "step": 13534
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.715781782756941e-07,
+      "loss": 0.0003,
+      "step": 13535
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 1.7155382367267412e-07,
+      "loss": 0.0002,
+      "step": 13536
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.7152946906965415e-07,
+      "loss": 0.0003,
+      "step": 13537
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.7150511446663417e-07,
+      "loss": 0.0003,
+      "step": 13538
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.714807598636142e-07,
+      "loss": 0.0026,
+      "step": 13539
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.7145640526059425e-07,
+      "loss": 0.0002,
+      "step": 13540
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.7143205065757427e-07,
+      "loss": 0.0004,
+      "step": 13541
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.714076960545543e-07,
+      "loss": 0.0005,
+      "step": 13542
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.7138334145153432e-07,
+      "loss": 0.0003,
+      "step": 13543
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.7135898684851435e-07,
+      "loss": 0.0004,
+      "step": 13544
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.7133463224549438e-07,
+      "loss": 0.0002,
+      "step": 13545
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 1.713102776424744e-07,
+      "loss": 0.0002,
+      "step": 13546
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.7128592303945443e-07,
+      "loss": 0.0074,
+      "step": 13547
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.7126156843643448e-07,
+      "loss": 0.0003,
+      "step": 13548
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.712372138334145e-07,
+      "loss": 0.0003,
+      "step": 13549
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.7121285923039453e-07,
+      "loss": 0.0005,
+      "step": 13550
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.7118850462737455e-07,
+      "loss": 0.0004,
+      "step": 13551
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.711641500243546e-07,
+      "loss": 0.0004,
+      "step": 13552
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.7113979542133463e-07,
+      "loss": 0.0005,
+      "step": 13553
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.7111544081831465e-07,
+      "loss": 0.0003,
+      "step": 13554
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.710910862152947e-07,
+      "loss": 0.0003,
+      "step": 13555
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.7106673161227473e-07,
+      "loss": 0.0003,
+      "step": 13556
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 1.7104237700925476e-07,
+      "loss": 0.0002,
+      "step": 13557
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.7101802240623478e-07,
+      "loss": 0.0003,
+      "step": 13558
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.709936678032148e-07,
+      "loss": 0.0004,
+      "step": 13559
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.7096931320019483e-07,
+      "loss": 0.0004,
+      "step": 13560
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.7094495859717486e-07,
+      "loss": 0.0002,
+      "step": 13561
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.7092060399415488e-07,
+      "loss": 0.0004,
+      "step": 13562
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.7089624939113493e-07,
+      "loss": 0.0002,
+      "step": 13563
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.7087189478811496e-07,
+      "loss": 0.0002,
+      "step": 13564
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.7084754018509498e-07,
+      "loss": 0.0003,
+      "step": 13565
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.70823185582075e-07,
+      "loss": 0.0002,
+      "step": 13566
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 1.7079883097905503e-07,
+      "loss": 0.0003,
+      "step": 13567
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.7077447637603506e-07,
+      "loss": 0.0005,
+      "step": 13568
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.7075012177301508e-07,
+      "loss": 0.0005,
+      "step": 13569
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.707257671699951e-07,
+      "loss": 0.0002,
+      "step": 13570
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.7070141256697516e-07,
+      "loss": 0.0002,
+      "step": 13571
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.706770579639552e-07,
+      "loss": 0.0004,
+      "step": 13572
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.706527033609352e-07,
+      "loss": 0.0002,
+      "step": 13573
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.7062834875791524e-07,
+      "loss": 0.0002,
+      "step": 13574
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.7060399415489526e-07,
+      "loss": 0.0002,
+      "step": 13575
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.705796395518753e-07,
+      "loss": 0.0004,
+      "step": 13576
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 1.705552849488553e-07,
+      "loss": 0.0002,
+      "step": 13577
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.7053093034583534e-07,
+      "loss": 0.0004,
+      "step": 13578
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.705065757428154e-07,
+      "loss": 0.0003,
+      "step": 13579
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.7048222113979541e-07,
+      "loss": 0.0003,
+      "step": 13580
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.7045786653677544e-07,
+      "loss": 0.0002,
+      "step": 13581
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.7043351193375546e-07,
+      "loss": 0.0002,
+      "step": 13582
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.704091573307355e-07,
+      "loss": 0.0004,
+      "step": 13583
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.7038480272771552e-07,
+      "loss": 0.0002,
+      "step": 13584
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.7036044812469554e-07,
+      "loss": 0.0003,
+      "step": 13585
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.7033609352167557e-07,
+      "loss": 0.0004,
+      "step": 13586
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 1.7031173891865562e-07,
+      "loss": 0.0002,
+      "step": 13587
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7028738431563564e-07,
+      "loss": 0.0002,
+      "step": 13588
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7026302971261567e-07,
+      "loss": 0.0003,
+      "step": 13589
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7023867510959572e-07,
+      "loss": 0.0002,
+      "step": 13590
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7021432050657574e-07,
+      "loss": 0.0004,
+      "step": 13591
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7018996590355577e-07,
+      "loss": 0.0002,
+      "step": 13592
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.701656113005358e-07,
+      "loss": 0.0003,
+      "step": 13593
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7014125669751585e-07,
+      "loss": 0.0003,
+      "step": 13594
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7011690209449587e-07,
+      "loss": 0.0002,
+      "step": 13595
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.700925474914759e-07,
+      "loss": 0.0001,
+      "step": 13596
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7006819288845592e-07,
+      "loss": 0.0004,
+      "step": 13597
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 1.7004383828543595e-07,
+      "loss": 0.0002,
+      "step": 13598
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.7001948368241597e-07,
+      "loss": 0.0002,
+      "step": 13599
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.69995129079396e-07,
+      "loss": 0.0003,
+      "step": 13600
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.6997077447637602e-07,
+      "loss": 0.0003,
+      "step": 13601
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.6994641987335607e-07,
+      "loss": 0.0002,
+      "step": 13602
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.699220652703361e-07,
+      "loss": 0.0002,
+      "step": 13603
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.6989771066731612e-07,
+      "loss": 0.0002,
+      "step": 13604
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.6987335606429615e-07,
+      "loss": 0.0002,
+      "step": 13605
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.6984900146127617e-07,
+      "loss": 0.0002,
+      "step": 13606
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.698246468582562e-07,
+      "loss": 0.0002,
+      "step": 13607
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 1.6980029225523623e-07,
+      "loss": 0.0004,
+      "step": 13608
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.6977593765221625e-07,
+      "loss": 0.0003,
+      "step": 13609
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.697515830491963e-07,
+      "loss": 0.0003,
+      "step": 13610
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.6972722844617633e-07,
+      "loss": 0.0003,
+      "step": 13611
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.6970287384315635e-07,
+      "loss": 0.0004,
+      "step": 13612
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.6967851924013638e-07,
+      "loss": 0.0006,
+      "step": 13613
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.696541646371164e-07,
+      "loss": 0.0002,
+      "step": 13614
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.6962981003409643e-07,
+      "loss": 0.0002,
+      "step": 13615
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.6960545543107645e-07,
+      "loss": 0.0003,
+      "step": 13616
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.6958110082805648e-07,
+      "loss": 0.0005,
+      "step": 13617
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 1.6955674622503653e-07,
+      "loss": 0.0004,
+      "step": 13618
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.6953239162201655e-07,
+      "loss": 0.0004,
+      "step": 13619
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.6950803701899658e-07,
+      "loss": 0.0003,
+      "step": 13620
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.694836824159766e-07,
+      "loss": 0.0003,
+      "step": 13621
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.6945932781295663e-07,
+      "loss": 0.0004,
+      "step": 13622
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.6943497320993666e-07,
+      "loss": 0.0003,
+      "step": 13623
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.6941061860691668e-07,
+      "loss": 0.0002,
+      "step": 13624
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.693862640038967e-07,
+      "loss": 0.0003,
+      "step": 13625
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.6936190940087676e-07,
+      "loss": 0.0002,
+      "step": 13626
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.6933755479785678e-07,
+      "loss": 0.0005,
+      "step": 13627
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.693132001948368e-07,
+      "loss": 0.0004,
+      "step": 13628
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 1.6928884559181686e-07,
+      "loss": 0.0002,
+      "step": 13629
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.6926449098879688e-07,
+      "loss": 0.0004,
+      "step": 13630
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.692401363857769e-07,
+      "loss": 0.0003,
+      "step": 13631
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.6921578178275693e-07,
+      "loss": 0.0005,
+      "step": 13632
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.6919142717973696e-07,
+      "loss": 0.0002,
+      "step": 13633
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.69167072576717e-07,
+      "loss": 0.0002,
+      "step": 13634
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.6914271797369704e-07,
+      "loss": 0.0003,
+      "step": 13635
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.6911836337067706e-07,
+      "loss": 0.0002,
+      "step": 13636
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.690940087676571e-07,
+      "loss": 0.0003,
+      "step": 13637
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.690696541646371e-07,
+      "loss": 0.0003,
+      "step": 13638
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 1.6904529956161714e-07,
+      "loss": 0.0002,
+      "step": 13639
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.6902094495859716e-07,
+      "loss": 0.0003,
+      "step": 13640
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.689965903555772e-07,
+      "loss": 0.0002,
+      "step": 13641
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.6897223575255724e-07,
+      "loss": 0.0005,
+      "step": 13642
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.6894788114953726e-07,
+      "loss": 0.0002,
+      "step": 13643
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.689235265465173e-07,
+      "loss": 0.0003,
+      "step": 13644
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.6889917194349732e-07,
+      "loss": 0.0004,
+      "step": 13645
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.6887481734047734e-07,
+      "loss": 0.0003,
+      "step": 13646
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.6885046273745737e-07,
+      "loss": 0.0002,
+      "step": 13647
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.688261081344374e-07,
+      "loss": 0.0002,
+      "step": 13648
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 1.6880175353141744e-07,
+      "loss": 0.0002,
+      "step": 13649
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.6877739892839747e-07,
+      "loss": 0.0002,
+      "step": 13650
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.687530443253775e-07,
+      "loss": 0.0003,
+      "step": 13651
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.6872868972235752e-07,
+      "loss": 0.0003,
+      "step": 13652
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.6870433511933754e-07,
+      "loss": 0.0003,
+      "step": 13653
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.6867998051631757e-07,
+      "loss": 0.0003,
+      "step": 13654
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.686556259132976e-07,
+      "loss": 0.0003,
+      "step": 13655
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.6863127131027762e-07,
+      "loss": 0.0002,
+      "step": 13656
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.6860691670725767e-07,
+      "loss": 1.6474,
+      "step": 13657
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.685825621042377e-07,
+      "loss": 0.0002,
+      "step": 13658
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 1.6855820750121772e-07,
+      "loss": 0.0003,
+      "step": 13659
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.6853385289819775e-07,
+      "loss": 0.0003,
+      "step": 13660
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.6850949829517777e-07,
+      "loss": 0.0003,
+      "step": 13661
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.684851436921578e-07,
+      "loss": 0.0002,
+      "step": 13662
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.6846078908913782e-07,
+      "loss": 0.0003,
+      "step": 13663
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.6843643448611785e-07,
+      "loss": 0.0228,
+      "step": 13664
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.684120798830979e-07,
+      "loss": 0.0002,
+      "step": 13665
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.6838772528007792e-07,
+      "loss": 0.0002,
+      "step": 13666
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.6836337067705797e-07,
+      "loss": 0.0002,
+      "step": 13667
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.68339016074038e-07,
+      "loss": 0.0002,
+      "step": 13668
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.6831466147101802e-07,
+      "loss": 0.0003,
+      "step": 13669
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 1.6829030686799805e-07,
+      "loss": 0.0003,
+      "step": 13670
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.6826595226497808e-07,
+      "loss": 0.0002,
+      "step": 13671
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.682415976619581e-07,
+      "loss": 0.0002,
+      "step": 13672
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.6821724305893815e-07,
+      "loss": 0.0003,
+      "step": 13673
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.6819288845591818e-07,
+      "loss": 0.0005,
+      "step": 13674
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.681685338528982e-07,
+      "loss": 0.0002,
+      "step": 13675
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.6814417924987823e-07,
+      "loss": 0.0002,
+      "step": 13676
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.6811982464685825e-07,
+      "loss": 0.0002,
+      "step": 13677
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.6809547004383828e-07,
+      "loss": 0.0003,
+      "step": 13678
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.680711154408183e-07,
+      "loss": 0.0005,
+      "step": 13679
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 1.6804676083779833e-07,
+      "loss": 0.0004,
+      "step": 13680
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.6802240623477838e-07,
+      "loss": 0.0002,
+      "step": 13681
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.679980516317584e-07,
+      "loss": 0.0002,
+      "step": 13682
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.6797369702873843e-07,
+      "loss": 0.0002,
+      "step": 13683
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.6794934242571846e-07,
+      "loss": 0.0003,
+      "step": 13684
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.6792498782269848e-07,
+      "loss": 0.0002,
+      "step": 13685
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.679006332196785e-07,
+      "loss": 0.0002,
+      "step": 13686
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.6787627861665853e-07,
+      "loss": 0.0002,
+      "step": 13687
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.6785192401363856e-07,
+      "loss": 0.0002,
+      "step": 13688
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.678275694106186e-07,
+      "loss": 0.0003,
+      "step": 13689
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 1.6780321480759863e-07,
+      "loss": 0.0003,
+      "step": 13690
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6777886020457866e-07,
+      "loss": 0.0005,
+      "step": 13691
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6775450560155868e-07,
+      "loss": 0.0004,
+      "step": 13692
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.677301509985387e-07,
+      "loss": 0.0002,
+      "step": 13693
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6770579639551873e-07,
+      "loss": 0.0001,
+      "step": 13694
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6768144179249876e-07,
+      "loss": 0.0003,
+      "step": 13695
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6765708718947878e-07,
+      "loss": 0.0002,
+      "step": 13696
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6763273258645884e-07,
+      "loss": 0.0006,
+      "step": 13697
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6760837798343886e-07,
+      "loss": 0.0003,
+      "step": 13698
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6758402338041889e-07,
+      "loss": 0.0003,
+      "step": 13699
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.675596687773989e-07,
+      "loss": 0.0002,
+      "step": 13700
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 1.6753531417437894e-07,
+      "loss": 0.0003,
+      "step": 13701
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.6751095957135896e-07,
+      "loss": 0.0002,
+      "step": 13702
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.6748660496833899e-07,
+      "loss": 0.0004,
+      "step": 13703
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.67462250365319e-07,
+      "loss": 0.0003,
+      "step": 13704
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.674378957622991e-07,
+      "loss": 0.0004,
+      "step": 13705
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.6741354115927911e-07,
+      "loss": 0.0002,
+      "step": 13706
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.6738918655625914e-07,
+      "loss": 0.0002,
+      "step": 13707
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.6736483195323917e-07,
+      "loss": 0.0007,
+      "step": 13708
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.673404773502192e-07,
+      "loss": 0.0003,
+      "step": 13709
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.6731612274719922e-07,
+      "loss": 0.3738,
+      "step": 13710
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 1.6729176814417924e-07,
+      "loss": 0.0003,
+      "step": 13711
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.672674135411593e-07,
+      "loss": 0.0002,
+      "step": 13712
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.6724305893813932e-07,
+      "loss": 0.0003,
+      "step": 13713
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.6721870433511934e-07,
+      "loss": 0.0003,
+      "step": 13714
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.6719434973209937e-07,
+      "loss": 0.0002,
+      "step": 13715
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.671699951290794e-07,
+      "loss": 0.0003,
+      "step": 13716
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.6714564052605942e-07,
+      "loss": 0.0003,
+      "step": 13717
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.6712128592303944e-07,
+      "loss": 0.0004,
+      "step": 13718
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.6709693132001947e-07,
+      "loss": 0.0003,
+      "step": 13719
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.6707257671699952e-07,
+      "loss": 0.0003,
+      "step": 13720
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 1.6704822211397955e-07,
+      "loss": 0.0003,
+      "step": 13721
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.6702386751095957e-07,
+      "loss": 0.0003,
+      "step": 13722
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.669995129079396e-07,
+      "loss": 0.0003,
+      "step": 13723
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.6697515830491962e-07,
+      "loss": 0.0003,
+      "step": 13724
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.6695080370189965e-07,
+      "loss": 0.0003,
+      "step": 13725
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.6692644909887967e-07,
+      "loss": 0.0002,
+      "step": 13726
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.669020944958597e-07,
+      "loss": 0.0003,
+      "step": 13727
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.6687773989283975e-07,
+      "loss": 0.0003,
+      "step": 13728
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.6685338528981977e-07,
+      "loss": 0.0003,
+      "step": 13729
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.668290306867998e-07,
+      "loss": 0.0003,
+      "step": 13730
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.6680467608377982e-07,
+      "loss": 0.0003,
+      "step": 13731
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 1.6678032148075985e-07,
+      "loss": 0.0004,
+      "step": 13732
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.6675596687773987e-07,
+      "loss": 0.0003,
+      "step": 13733
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.667316122747199e-07,
+      "loss": 0.0003,
+      "step": 13734
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.6670725767169992e-07,
+      "loss": 0.0003,
+      "step": 13735
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.6668290306867998e-07,
+      "loss": 0.0004,
+      "step": 13736
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.6665854846566e-07,
+      "loss": 0.0002,
+      "step": 13737
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.6663419386264003e-07,
+      "loss": 0.1902,
+      "step": 13738
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.6660983925962005e-07,
+      "loss": 0.0004,
+      "step": 13739
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.6658548465660008e-07,
+      "loss": 0.0002,
+      "step": 13740
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.665611300535801e-07,
+      "loss": 0.0002,
+      "step": 13741
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 1.6653677545056013e-07,
+      "loss": 0.0002,
+      "step": 13742
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6651242084754015e-07,
+      "loss": 0.0004,
+      "step": 13743
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6648806624452023e-07,
+      "loss": 0.0002,
+      "step": 13744
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6646371164150026e-07,
+      "loss": 0.0003,
+      "step": 13745
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6643935703848028e-07,
+      "loss": 0.0003,
+      "step": 13746
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.664150024354603e-07,
+      "loss": 0.0003,
+      "step": 13747
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6639064783244033e-07,
+      "loss": 0.0004,
+      "step": 13748
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6636629322942036e-07,
+      "loss": 0.0005,
+      "step": 13749
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6634193862640038e-07,
+      "loss": 0.0003,
+      "step": 13750
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6631758402338043e-07,
+      "loss": 0.0003,
+      "step": 13751
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 1.6629322942036046e-07,
+      "loss": 0.0003,
+      "step": 13752
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.6626887481734048e-07,
+      "loss": 0.0003,
+      "step": 13753
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.662445202143205e-07,
+      "loss": 0.0003,
+      "step": 13754
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.6622016561130053e-07,
+      "loss": 0.0002,
+      "step": 13755
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.6619581100828056e-07,
+      "loss": 0.0002,
+      "step": 13756
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.6617145640526058e-07,
+      "loss": 0.0002,
+      "step": 13757
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.661471018022406e-07,
+      "loss": 0.0002,
+      "step": 13758
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.6612274719922066e-07,
+      "loss": 0.0002,
+      "step": 13759
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.6609839259620069e-07,
+      "loss": 0.0002,
+      "step": 13760
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.660740379931807e-07,
+      "loss": 0.0003,
+      "step": 13761
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 1.6604968339016074e-07,
+      "loss": 0.0002,
+      "step": 13762
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.6602532878714076e-07,
+      "loss": 0.0014,
+      "step": 13763
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.6600097418412079e-07,
+      "loss": 0.0005,
+      "step": 13764
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.659766195811008e-07,
+      "loss": 0.0002,
+      "step": 13765
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.6595226497808084e-07,
+      "loss": 0.0003,
+      "step": 13766
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.659279103750609e-07,
+      "loss": 0.0002,
+      "step": 13767
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.6590355577204091e-07,
+      "loss": 0.0003,
+      "step": 13768
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.6587920116902094e-07,
+      "loss": 0.0002,
+      "step": 13769
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.6585484656600096e-07,
+      "loss": 0.0004,
+      "step": 13770
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.65830491962981e-07,
+      "loss": 0.0003,
+      "step": 13771
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.6580613735996101e-07,
+      "loss": 0.0007,
+      "step": 13772
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 1.6578178275694104e-07,
+      "loss": 0.0532,
+      "step": 13773
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6575742815392106e-07,
+      "loss": 0.0004,
+      "step": 13774
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6573307355090112e-07,
+      "loss": 0.0005,
+      "step": 13775
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6570871894788114e-07,
+      "loss": 0.0004,
+      "step": 13776
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6568436434486117e-07,
+      "loss": 0.0004,
+      "step": 13777
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.656600097418412e-07,
+      "loss": 0.0005,
+      "step": 13778
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6563565513882122e-07,
+      "loss": 0.0002,
+      "step": 13779
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6561130053580124e-07,
+      "loss": 0.0002,
+      "step": 13780
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6558694593278127e-07,
+      "loss": 0.0003,
+      "step": 13781
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6556259132976132e-07,
+      "loss": 0.0002,
+      "step": 13782
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 1.6553823672674137e-07,
+      "loss": 0.0003,
+      "step": 13783
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.655138821237214e-07,
+      "loss": 0.0004,
+      "step": 13784
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.6548952752070142e-07,
+      "loss": 0.0003,
+      "step": 13785
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.6546517291768145e-07,
+      "loss": 0.0003,
+      "step": 13786
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.6544081831466147e-07,
+      "loss": 0.0002,
+      "step": 13787
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.654164637116415e-07,
+      "loss": 0.0002,
+      "step": 13788
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.6539210910862152e-07,
+      "loss": 0.0005,
+      "step": 13789
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.6536775450560155e-07,
+      "loss": 0.0003,
+      "step": 13790
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.653433999025816e-07,
+      "loss": 0.0003,
+      "step": 13791
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.6531904529956162e-07,
+      "loss": 0.0004,
+      "step": 13792
+    },
+    {
+      "epoch": 13.4,
+      "learning_rate": 1.6529469069654165e-07,
+      "loss": 0.0005,
+      "step": 13793
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6527033609352167e-07,
+      "loss": 0.0003,
+      "step": 13794
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.652459814905017e-07,
+      "loss": 0.0002,
+      "step": 13795
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6522162688748172e-07,
+      "loss": 0.0003,
+      "step": 13796
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6519727228446175e-07,
+      "loss": 0.0004,
+      "step": 13797
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6517291768144177e-07,
+      "loss": 0.0003,
+      "step": 13798
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6514856307842183e-07,
+      "loss": 0.0004,
+      "step": 13799
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6512420847540185e-07,
+      "loss": 0.0005,
+      "step": 13800
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6509985387238188e-07,
+      "loss": 0.0002,
+      "step": 13801
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.650754992693619e-07,
+      "loss": 0.0003,
+      "step": 13802
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6505114466634193e-07,
+      "loss": 0.0003,
+      "step": 13803
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 1.6502679006332195e-07,
+      "loss": 0.0005,
+      "step": 13804
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.6500243546030198e-07,
+      "loss": 0.0002,
+      "step": 13805
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.64978080857282e-07,
+      "loss": 0.0003,
+      "step": 13806
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.6495372625426205e-07,
+      "loss": 0.0005,
+      "step": 13807
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.6492937165124208e-07,
+      "loss": 0.0003,
+      "step": 13808
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.649050170482221e-07,
+      "loss": 0.0002,
+      "step": 13809
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.6488066244520213e-07,
+      "loss": 0.0003,
+      "step": 13810
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.6485630784218215e-07,
+      "loss": 0.0004,
+      "step": 13811
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.6483195323916218e-07,
+      "loss": 0.0002,
+      "step": 13812
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.648075986361422e-07,
+      "loss": 0.0002,
+      "step": 13813
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 1.6478324403312223e-07,
+      "loss": 0.0003,
+      "step": 13814
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.6475888943010228e-07,
+      "loss": 0.0002,
+      "step": 13815
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.647345348270823e-07,
+      "loss": 0.0002,
+      "step": 13816
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.6471018022406233e-07,
+      "loss": 0.0002,
+      "step": 13817
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.6468582562104236e-07,
+      "loss": 0.0003,
+      "step": 13818
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.6466147101802238e-07,
+      "loss": 0.0002,
+      "step": 13819
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.6463711641500243e-07,
+      "loss": 0.0003,
+      "step": 13820
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.6461276181198246e-07,
+      "loss": 0.0016,
+      "step": 13821
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.645884072089625e-07,
+      "loss": 0.0003,
+      "step": 13822
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.6456405260594254e-07,
+      "loss": 0.0003,
+      "step": 13823
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 1.6453969800292256e-07,
+      "loss": 0.0006,
+      "step": 13824
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.6451534339990259e-07,
+      "loss": 0.0003,
+      "step": 13825
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.644909887968826e-07,
+      "loss": 0.0003,
+      "step": 13826
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.6446663419386264e-07,
+      "loss": 0.0002,
+      "step": 13827
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.6444227959084266e-07,
+      "loss": 0.0004,
+      "step": 13828
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.6441792498782269e-07,
+      "loss": 0.0002,
+      "step": 13829
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.6439357038480274e-07,
+      "loss": 0.0004,
+      "step": 13830
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.6436921578178276e-07,
+      "loss": 0.0003,
+      "step": 13831
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.643448611787628e-07,
+      "loss": 0.0002,
+      "step": 13832
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.6432050657574281e-07,
+      "loss": 0.0003,
+      "step": 13833
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 1.6429615197272284e-07,
+      "loss": 0.0003,
+      "step": 13834
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.6427179736970286e-07,
+      "loss": 0.0002,
+      "step": 13835
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.642474427666829e-07,
+      "loss": 0.0004,
+      "step": 13836
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.6422308816366291e-07,
+      "loss": 0.0002,
+      "step": 13837
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.6419873356064297e-07,
+      "loss": 0.0002,
+      "step": 13838
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.64174378957623e-07,
+      "loss": 0.0003,
+      "step": 13839
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.6415002435460302e-07,
+      "loss": 0.0003,
+      "step": 13840
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.6412566975158304e-07,
+      "loss": 0.0003,
+      "step": 13841
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.6410131514856307e-07,
+      "loss": 0.0003,
+      "step": 13842
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.640769605455431e-07,
+      "loss": 0.0008,
+      "step": 13843
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.6405260594252312e-07,
+      "loss": 0.0001,
+      "step": 13844
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 1.6402825133950314e-07,
+      "loss": 0.0003,
+      "step": 13845
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.640038967364832e-07,
+      "loss": 0.0003,
+      "step": 13846
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.6397954213346322e-07,
+      "loss": 0.0003,
+      "step": 13847
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.6395518753044324e-07,
+      "loss": 0.0003,
+      "step": 13848
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.6393083292742327e-07,
+      "loss": 0.0001,
+      "step": 13849
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.639064783244033e-07,
+      "loss": 0.0002,
+      "step": 13850
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.6388212372138332e-07,
+      "loss": 0.0005,
+      "step": 13851
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.6385776911836334e-07,
+      "loss": 0.0003,
+      "step": 13852
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.6383341451534337e-07,
+      "loss": 0.0002,
+      "step": 13853
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.6380905991232342e-07,
+      "loss": 0.0002,
+      "step": 13854
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 1.6378470530930345e-07,
+      "loss": 0.0003,
+      "step": 13855
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.6376035070628347e-07,
+      "loss": 0.0004,
+      "step": 13856
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.637359961032635e-07,
+      "loss": 0.0002,
+      "step": 13857
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.6371164150024355e-07,
+      "loss": 0.0002,
+      "step": 13858
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.6368728689722357e-07,
+      "loss": 0.0003,
+      "step": 13859
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.636629322942036e-07,
+      "loss": 0.0003,
+      "step": 13860
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.6363857769118365e-07,
+      "loss": 0.0003,
+      "step": 13861
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.6361422308816368e-07,
+      "loss": 0.0002,
+      "step": 13862
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.635898684851437e-07,
+      "loss": 0.0003,
+      "step": 13863
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.6356551388212373e-07,
+      "loss": 0.0004,
+      "step": 13864
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 1.6354115927910375e-07,
+      "loss": 0.0004,
+      "step": 13865
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.6351680467608378e-07,
+      "loss": 0.0003,
+      "step": 13866
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.634924500730638e-07,
+      "loss": 0.0002,
+      "step": 13867
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.6346809547004383e-07,
+      "loss": 0.0002,
+      "step": 13868
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.6344374086702388e-07,
+      "loss": 0.0002,
+      "step": 13869
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.634193862640039e-07,
+      "loss": 0.0003,
+      "step": 13870
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.6339503166098393e-07,
+      "loss": 0.0002,
+      "step": 13871
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.6337067705796395e-07,
+      "loss": 0.0002,
+      "step": 13872
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.6334632245494398e-07,
+      "loss": 0.0002,
+      "step": 13873
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.63321967851924e-07,
+      "loss": 0.0003,
+      "step": 13874
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.6329761324890403e-07,
+      "loss": 0.0003,
+      "step": 13875
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 1.6327325864588405e-07,
+      "loss": 0.0002,
+      "step": 13876
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.632489040428641e-07,
+      "loss": 0.0004,
+      "step": 13877
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.6322454943984413e-07,
+      "loss": 0.0003,
+      "step": 13878
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.6320019483682416e-07,
+      "loss": 0.0003,
+      "step": 13879
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.6317584023380418e-07,
+      "loss": 0.0003,
+      "step": 13880
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.631514856307842e-07,
+      "loss": 0.0003,
+      "step": 13881
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.6312713102776423e-07,
+      "loss": 0.0003,
+      "step": 13882
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.6310277642474426e-07,
+      "loss": 0.0003,
+      "step": 13883
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.6307842182172428e-07,
+      "loss": 0.0006,
+      "step": 13884
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.6305406721870433e-07,
+      "loss": 0.0002,
+      "step": 13885
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 1.6302971261568436e-07,
+      "loss": 0.0002,
+      "step": 13886
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.6300535801266438e-07,
+      "loss": 0.0002,
+      "step": 13887
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.629810034096444e-07,
+      "loss": 0.0002,
+      "step": 13888
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.6295664880662443e-07,
+      "loss": 0.0003,
+      "step": 13889
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.6293229420360446e-07,
+      "loss": 0.0003,
+      "step": 13890
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.6290793960058449e-07,
+      "loss": 0.0003,
+      "step": 13891
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.628835849975645e-07,
+      "loss": 0.0003,
+      "step": 13892
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.6285923039454456e-07,
+      "loss": 0.0002,
+      "step": 13893
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.628348757915246e-07,
+      "loss": 0.0007,
+      "step": 13894
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.628105211885046e-07,
+      "loss": 0.0002,
+      "step": 13895
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 1.6278616658548464e-07,
+      "loss": 0.0002,
+      "step": 13896
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.627618119824647e-07,
+      "loss": 0.0004,
+      "step": 13897
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.6273745737944471e-07,
+      "loss": 0.0003,
+      "step": 13898
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.6271310277642474e-07,
+      "loss": 0.0002,
+      "step": 13899
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.6268874817340476e-07,
+      "loss": 0.0003,
+      "step": 13900
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.6266439357038482e-07,
+      "loss": 0.0004,
+      "step": 13901
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.6264003896736484e-07,
+      "loss": 0.0003,
+      "step": 13902
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.6261568436434487e-07,
+      "loss": 0.0006,
+      "step": 13903
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.625913297613249e-07,
+      "loss": 0.0004,
+      "step": 13904
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.6256697515830492e-07,
+      "loss": 0.0002,
+      "step": 13905
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 1.6254262055528494e-07,
+      "loss": 0.0002,
+      "step": 13906
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.6251826595226497e-07,
+      "loss": 0.0004,
+      "step": 13907
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.62493911349245e-07,
+      "loss": 0.0002,
+      "step": 13908
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.6246955674622504e-07,
+      "loss": 0.0003,
+      "step": 13909
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.6244520214320507e-07,
+      "loss": 0.0002,
+      "step": 13910
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.624208475401851e-07,
+      "loss": 0.0002,
+      "step": 13911
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.6239649293716512e-07,
+      "loss": 0.001,
+      "step": 13912
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.6237213833414514e-07,
+      "loss": 0.0001,
+      "step": 13913
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.6234778373112517e-07,
+      "loss": 0.0003,
+      "step": 13914
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.623234291281052e-07,
+      "loss": 0.0002,
+      "step": 13915
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.6229907452508522e-07,
+      "loss": 0.0002,
+      "step": 13916
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 1.6227471992206527e-07,
+      "loss": 0.0004,
+      "step": 13917
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.622503653190453e-07,
+      "loss": 0.0003,
+      "step": 13918
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.6222601071602532e-07,
+      "loss": 0.0004,
+      "step": 13919
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.6220165611300535e-07,
+      "loss": 0.0006,
+      "step": 13920
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.6217730150998537e-07,
+      "loss": 0.0002,
+      "step": 13921
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.621529469069654e-07,
+      "loss": 0.0003,
+      "step": 13922
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.6212859230394542e-07,
+      "loss": 0.0002,
+      "step": 13923
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.6210423770092547e-07,
+      "loss": 0.0003,
+      "step": 13924
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.620798830979055e-07,
+      "loss": 0.0001,
+      "step": 13925
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.6205552849488552e-07,
+      "loss": 0.0003,
+      "step": 13926
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 1.6203117389186555e-07,
+      "loss": 0.0002,
+      "step": 13927
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.6200681928884558e-07,
+      "loss": 0.0003,
+      "step": 13928
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.619824646858256e-07,
+      "loss": 0.0002,
+      "step": 13929
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.6195811008280563e-07,
+      "loss": 0.0002,
+      "step": 13930
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.6193375547978565e-07,
+      "loss": 0.0002,
+      "step": 13931
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.619094008767657e-07,
+      "loss": 0.0001,
+      "step": 13932
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.6188504627374573e-07,
+      "loss": 0.0002,
+      "step": 13933
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.6186069167072575e-07,
+      "loss": 0.0002,
+      "step": 13934
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.618363370677058e-07,
+      "loss": 0.0002,
+      "step": 13935
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.6181198246468583e-07,
+      "loss": 0.0002,
+      "step": 13936
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 1.6178762786166585e-07,
+      "loss": 0.0003,
+      "step": 13937
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.6176327325864588e-07,
+      "loss": 0.0003,
+      "step": 13938
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.617389186556259e-07,
+      "loss": 0.1315,
+      "step": 13939
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.6171456405260596e-07,
+      "loss": 0.0003,
+      "step": 13940
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.6169020944958598e-07,
+      "loss": 0.0005,
+      "step": 13941
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.61665854846566e-07,
+      "loss": 0.0001,
+      "step": 13942
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.6164150024354603e-07,
+      "loss": 0.0002,
+      "step": 13943
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.6161714564052606e-07,
+      "loss": 0.0003,
+      "step": 13944
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.6159279103750608e-07,
+      "loss": 0.0002,
+      "step": 13945
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.615684364344861e-07,
+      "loss": 0.0005,
+      "step": 13946
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.6154408183146613e-07,
+      "loss": 0.0003,
+      "step": 13947
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 1.6151972722844618e-07,
+      "loss": 0.0004,
+      "step": 13948
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.614953726254262e-07,
+      "loss": 0.0003,
+      "step": 13949
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.6147101802240623e-07,
+      "loss": 0.0003,
+      "step": 13950
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.6144666341938626e-07,
+      "loss": 0.0003,
+      "step": 13951
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.6142230881636628e-07,
+      "loss": 0.0003,
+      "step": 13952
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.613979542133463e-07,
+      "loss": 0.0002,
+      "step": 13953
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.6137359961032634e-07,
+      "loss": 0.0004,
+      "step": 13954
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.6134924500730636e-07,
+      "loss": 0.0002,
+      "step": 13955
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.613248904042864e-07,
+      "loss": 0.0003,
+      "step": 13956
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.6130053580126644e-07,
+      "loss": 0.0005,
+      "step": 13957
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 1.6127618119824646e-07,
+      "loss": 0.0002,
+      "step": 13958
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.612518265952265e-07,
+      "loss": 0.0002,
+      "step": 13959
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.612274719922065e-07,
+      "loss": 0.0002,
+      "step": 13960
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.6120311738918654e-07,
+      "loss": 0.0003,
+      "step": 13961
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.6117876278616656e-07,
+      "loss": 0.0002,
+      "step": 13962
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.611544081831466e-07,
+      "loss": 0.0002,
+      "step": 13963
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.6113005358012664e-07,
+      "loss": 0.0002,
+      "step": 13964
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.6110569897710666e-07,
+      "loss": 0.0002,
+      "step": 13965
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.610813443740867e-07,
+      "loss": 0.0002,
+      "step": 13966
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.6105698977106672e-07,
+      "loss": 0.0003,
+      "step": 13967
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 1.6103263516804674e-07,
+      "loss": 0.0003,
+      "step": 13968
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.6100828056502677e-07,
+      "loss": 0.0002,
+      "step": 13969
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.609839259620068e-07,
+      "loss": 0.0003,
+      "step": 13970
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.6095957135898682e-07,
+      "loss": 0.0005,
+      "step": 13971
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.6093521675596687e-07,
+      "loss": 0.0002,
+      "step": 13972
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.6091086215294692e-07,
+      "loss": 0.0002,
+      "step": 13973
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.6088650754992694e-07,
+      "loss": 0.0008,
+      "step": 13974
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.6086215294690697e-07,
+      "loss": 0.0002,
+      "step": 13975
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.60837798343887e-07,
+      "loss": 0.0003,
+      "step": 13976
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.6081344374086702e-07,
+      "loss": 0.0002,
+      "step": 13977
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 1.6078908913784705e-07,
+      "loss": 0.0002,
+      "step": 13978
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.607647345348271e-07,
+      "loss": 0.0003,
+      "step": 13979
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6074037993180712e-07,
+      "loss": 0.0002,
+      "step": 13980
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6071602532878715e-07,
+      "loss": 0.0003,
+      "step": 13981
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6069167072576717e-07,
+      "loss": 0.0003,
+      "step": 13982
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.606673161227472e-07,
+      "loss": 0.0002,
+      "step": 13983
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6064296151972722e-07,
+      "loss": 0.0003,
+      "step": 13984
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6061860691670725e-07,
+      "loss": 0.0003,
+      "step": 13985
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6059425231368727e-07,
+      "loss": 0.0003,
+      "step": 13986
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6056989771066732e-07,
+      "loss": 0.0004,
+      "step": 13987
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6054554310764735e-07,
+      "loss": 0.0003,
+      "step": 13988
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 1.6052118850462737e-07,
+      "loss": 0.0002,
+      "step": 13989
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.604968339016074e-07,
+      "loss": 0.0004,
+      "step": 13990
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.6047247929858743e-07,
+      "loss": 0.0004,
+      "step": 13991
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.6044812469556745e-07,
+      "loss": 0.0002,
+      "step": 13992
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.6042377009254748e-07,
+      "loss": 0.0004,
+      "step": 13993
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.603994154895275e-07,
+      "loss": 0.0002,
+      "step": 13994
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.6037506088650755e-07,
+      "loss": 0.0002,
+      "step": 13995
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.6035070628348758e-07,
+      "loss": 0.0002,
+      "step": 13996
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.603263516804676e-07,
+      "loss": 0.0004,
+      "step": 13997
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.6030199707744763e-07,
+      "loss": 0.0002,
+      "step": 13998
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 1.6027764247442765e-07,
+      "loss": 0.0002,
+      "step": 13999
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.6025328787140768e-07,
+      "loss": 0.0002,
+      "step": 14000
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.602289332683877e-07,
+      "loss": 0.0002,
+      "step": 14001
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.6020457866536773e-07,
+      "loss": 0.0003,
+      "step": 14002
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.6018022406234778e-07,
+      "loss": 0.0003,
+      "step": 14003
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.601558694593278e-07,
+      "loss": 0.0003,
+      "step": 14004
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.6013151485630783e-07,
+      "loss": 0.0002,
+      "step": 14005
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.6010716025328786e-07,
+      "loss": 0.0002,
+      "step": 14006
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.6008280565026788e-07,
+      "loss": 0.0002,
+      "step": 14007
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.600584510472479e-07,
+      "loss": 0.0002,
+      "step": 14008
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 1.6003409644422793e-07,
+      "loss": 0.0003,
+      "step": 14009
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.6000974184120796e-07,
+      "loss": 0.0005,
+      "step": 14010
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.59985387238188e-07,
+      "loss": 0.0003,
+      "step": 14011
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.5996103263516806e-07,
+      "loss": 0.0001,
+      "step": 14012
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.5993667803214808e-07,
+      "loss": 0.0003,
+      "step": 14013
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.599123234291281e-07,
+      "loss": 0.0002,
+      "step": 14014
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.5988796882610813e-07,
+      "loss": 0.0003,
+      "step": 14015
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.5986361422308816e-07,
+      "loss": 0.0004,
+      "step": 14016
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.5983925962006819e-07,
+      "loss": 0.0002,
+      "step": 14017
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.5981490501704824e-07,
+      "loss": 0.0002,
+      "step": 14018
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.5979055041402826e-07,
+      "loss": 0.0002,
+      "step": 14019
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 1.597661958110083e-07,
+      "loss": 0.0004,
+      "step": 14020
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.597418412079883e-07,
+      "loss": 0.0002,
+      "step": 14021
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.5971748660496834e-07,
+      "loss": 0.0001,
+      "step": 14022
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.5969313200194836e-07,
+      "loss": 0.0004,
+      "step": 14023
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.596687773989284e-07,
+      "loss": 0.0002,
+      "step": 14024
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.596444227959084e-07,
+      "loss": 0.0003,
+      "step": 14025
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.5962006819288846e-07,
+      "loss": 0.0002,
+      "step": 14026
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.595957135898685e-07,
+      "loss": 0.0002,
+      "step": 14027
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.5957135898684852e-07,
+      "loss": 0.0002,
+      "step": 14028
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.5954700438382854e-07,
+      "loss": 0.0004,
+      "step": 14029
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 1.5952264978080857e-07,
+      "loss": 0.0002,
+      "step": 14030
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.594982951777886e-07,
+      "loss": 0.0002,
+      "step": 14031
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.5947394057476862e-07,
+      "loss": 0.0001,
+      "step": 14032
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.5944958597174864e-07,
+      "loss": 0.0003,
+      "step": 14033
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.594252313687287e-07,
+      "loss": 0.0005,
+      "step": 14034
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.5940087676570872e-07,
+      "loss": 0.0003,
+      "step": 14035
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.5937652216268874e-07,
+      "loss": 0.0002,
+      "step": 14036
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.5935216755966877e-07,
+      "loss": 0.0003,
+      "step": 14037
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.593278129566488e-07,
+      "loss": 0.0002,
+      "step": 14038
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.5930345835362882e-07,
+      "loss": 0.0002,
+      "step": 14039
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 1.5927910375060884e-07,
+      "loss": 0.0003,
+      "step": 14040
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.5925474914758887e-07,
+      "loss": 0.0002,
+      "step": 14041
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.5923039454456892e-07,
+      "loss": 0.0003,
+      "step": 14042
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.5920603994154895e-07,
+      "loss": 0.0003,
+      "step": 14043
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.5918168533852897e-07,
+      "loss": 0.0004,
+      "step": 14044
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.59157330735509e-07,
+      "loss": 0.0002,
+      "step": 14045
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.5913297613248902e-07,
+      "loss": 0.0002,
+      "step": 14046
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.5910862152946905e-07,
+      "loss": 0.0003,
+      "step": 14047
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.5908426692644907e-07,
+      "loss": 0.0004,
+      "step": 14048
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.590599123234291e-07,
+      "loss": 0.0002,
+      "step": 14049
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 1.5903555772040917e-07,
+      "loss": 0.0003,
+      "step": 14050
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.590112031173892e-07,
+      "loss": 0.0003,
+      "step": 14051
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.5898684851436922e-07,
+      "loss": 0.0003,
+      "step": 14052
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.5896249391134925e-07,
+      "loss": 0.0004,
+      "step": 14053
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.5893813930832928e-07,
+      "loss": 0.0003,
+      "step": 14054
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.589137847053093e-07,
+      "loss": 0.0003,
+      "step": 14055
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.5888943010228933e-07,
+      "loss": 0.0002,
+      "step": 14056
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.5886507549926935e-07,
+      "loss": 0.0002,
+      "step": 14057
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.588407208962494e-07,
+      "loss": 0.0003,
+      "step": 14058
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.5881636629322943e-07,
+      "loss": 0.0003,
+      "step": 14059
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.5879201169020945e-07,
+      "loss": 0.0001,
+      "step": 14060
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 1.5876765708718948e-07,
+      "loss": 0.0002,
+      "step": 14061
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.587433024841695e-07,
+      "loss": 0.0002,
+      "step": 14062
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.5871894788114953e-07,
+      "loss": 0.0003,
+      "step": 14063
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.5869459327812955e-07,
+      "loss": 0.0005,
+      "step": 14064
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.5867023867510958e-07,
+      "loss": 0.0004,
+      "step": 14065
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.5864588407208963e-07,
+      "loss": 0.0004,
+      "step": 14066
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.5862152946906966e-07,
+      "loss": 0.0002,
+      "step": 14067
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.5859717486604968e-07,
+      "loss": 0.0002,
+      "step": 14068
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.585728202630297e-07,
+      "loss": 0.0002,
+      "step": 14069
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.5854846566000973e-07,
+      "loss": 0.0002,
+      "step": 14070
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 1.5852411105698976e-07,
+      "loss": 0.0005,
+      "step": 14071
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.5849975645396978e-07,
+      "loss": 0.0002,
+      "step": 14072
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.584754018509498e-07,
+      "loss": 0.0002,
+      "step": 14073
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.5845104724792986e-07,
+      "loss": 0.0003,
+      "step": 14074
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.5842669264490988e-07,
+      "loss": 0.0002,
+      "step": 14075
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.584023380418899e-07,
+      "loss": 0.0003,
+      "step": 14076
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.5837798343886993e-07,
+      "loss": 0.0005,
+      "step": 14077
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.5835362883584996e-07,
+      "loss": 0.0002,
+      "step": 14078
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.5832927423282998e-07,
+      "loss": 0.0002,
+      "step": 14079
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.5830491962981e-07,
+      "loss": 0.0004,
+      "step": 14080
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 1.5828056502679003e-07,
+      "loss": 0.0002,
+      "step": 14081
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.5825621042377009e-07,
+      "loss": 0.0002,
+      "step": 14082
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.582318558207501e-07,
+      "loss": 0.0007,
+      "step": 14083
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.5820750121773014e-07,
+      "loss": 0.0003,
+      "step": 14084
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.5818314661471016e-07,
+      "loss": 0.0002,
+      "step": 14085
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.5815879201169019e-07,
+      "loss": 0.0007,
+      "step": 14086
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.581344374086702e-07,
+      "loss": 0.0003,
+      "step": 14087
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.5811008280565026e-07,
+      "loss": 0.0003,
+      "step": 14088
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.5808572820263031e-07,
+      "loss": 0.0003,
+      "step": 14089
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.5806137359961034e-07,
+      "loss": 0.0003,
+      "step": 14090
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.5803701899659037e-07,
+      "loss": 0.0003,
+      "step": 14091
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 1.580126643935704e-07,
+      "loss": 0.0002,
+      "step": 14092
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.5798830979055042e-07,
+      "loss": 0.0002,
+      "step": 14093
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.5796395518753044e-07,
+      "loss": 0.0002,
+      "step": 14094
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.5793960058451047e-07,
+      "loss": 0.0002,
+      "step": 14095
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.579152459814905e-07,
+      "loss": 0.0001,
+      "step": 14096
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.5789089137847054e-07,
+      "loss": 0.0004,
+      "step": 14097
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.5786653677545057e-07,
+      "loss": 0.0001,
+      "step": 14098
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.578421821724306e-07,
+      "loss": 0.0002,
+      "step": 14099
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.5781782756941062e-07,
+      "loss": 0.0002,
+      "step": 14100
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.5779347296639064e-07,
+      "loss": 0.0004,
+      "step": 14101
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 1.5776911836337067e-07,
+      "loss": 0.0003,
+      "step": 14102
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.577447637603507e-07,
+      "loss": 0.0003,
+      "step": 14103
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.5772040915733072e-07,
+      "loss": 0.0004,
+      "step": 14104
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.5769605455431077e-07,
+      "loss": 0.0003,
+      "step": 14105
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.576716999512908e-07,
+      "loss": 0.0002,
+      "step": 14106
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.5764734534827082e-07,
+      "loss": 0.0005,
+      "step": 14107
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.5762299074525085e-07,
+      "loss": 0.0003,
+      "step": 14108
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.5759863614223087e-07,
+      "loss": 0.006,
+      "step": 14109
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.575742815392109e-07,
+      "loss": 0.0003,
+      "step": 14110
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.5754992693619092e-07,
+      "loss": 0.0001,
+      "step": 14111
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 1.5752557233317095e-07,
+      "loss": 0.0002,
+      "step": 14112
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.57501217730151e-07,
+      "loss": 0.0002,
+      "step": 14113
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5747686312713102e-07,
+      "loss": 0.0001,
+      "step": 14114
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5745250852411105e-07,
+      "loss": 0.0003,
+      "step": 14115
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5742815392109107e-07,
+      "loss": 0.0002,
+      "step": 14116
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.574037993180711e-07,
+      "loss": 0.0005,
+      "step": 14117
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5737944471505112e-07,
+      "loss": 0.0003,
+      "step": 14118
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5735509011203115e-07,
+      "loss": 0.0003,
+      "step": 14119
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5733073550901117e-07,
+      "loss": 0.0002,
+      "step": 14120
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5730638090599123e-07,
+      "loss": 0.0003,
+      "step": 14121
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5728202630297125e-07,
+      "loss": 0.0003,
+      "step": 14122
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 1.5725767169995128e-07,
+      "loss": 0.0001,
+      "step": 14123
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.572333170969313e-07,
+      "loss": 0.0003,
+      "step": 14124
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.5720896249391133e-07,
+      "loss": 0.0002,
+      "step": 14125
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.5718460789089135e-07,
+      "loss": 0.0002,
+      "step": 14126
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.571602532878714e-07,
+      "loss": 2.4484,
+      "step": 14127
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.5713589868485146e-07,
+      "loss": 0.0004,
+      "step": 14128
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.5711154408183148e-07,
+      "loss": 0.0002,
+      "step": 14129
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.570871894788115e-07,
+      "loss": 0.0003,
+      "step": 14130
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.5706283487579153e-07,
+      "loss": 2.7187,
+      "step": 14131
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.5703848027277156e-07,
+      "loss": 0.0004,
+      "step": 14132
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 1.5701412566975158e-07,
+      "loss": 0.0002,
+      "step": 14133
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.569897710667316e-07,
+      "loss": 0.0002,
+      "step": 14134
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.5696541646371163e-07,
+      "loss": 0.0001,
+      "step": 14135
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.5694106186069168e-07,
+      "loss": 0.0003,
+      "step": 14136
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.569167072576717e-07,
+      "loss": 0.0004,
+      "step": 14137
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.5689235265465173e-07,
+      "loss": 0.0003,
+      "step": 14138
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.5686799805163176e-07,
+      "loss": 0.0002,
+      "step": 14139
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.5684364344861178e-07,
+      "loss": 0.0003,
+      "step": 14140
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.568192888455918e-07,
+      "loss": 0.0001,
+      "step": 14141
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.5679493424257183e-07,
+      "loss": 0.0002,
+      "step": 14142
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 1.5677057963955186e-07,
+      "loss": 0.0003,
+      "step": 14143
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.567462250365319e-07,
+      "loss": 0.0002,
+      "step": 14144
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.5672187043351194e-07,
+      "loss": 0.0003,
+      "step": 14145
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.5669751583049196e-07,
+      "loss": 0.0002,
+      "step": 14146
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.5667316122747199e-07,
+      "loss": 0.0004,
+      "step": 14147
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.56648806624452e-07,
+      "loss": 0.0002,
+      "step": 14148
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.5662445202143204e-07,
+      "loss": 0.0003,
+      "step": 14149
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.5660009741841206e-07,
+      "loss": 0.0003,
+      "step": 14150
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.565757428153921e-07,
+      "loss": 0.0002,
+      "step": 14151
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.5655138821237214e-07,
+      "loss": 0.0002,
+      "step": 14152
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 1.5652703360935216e-07,
+      "loss": 0.0003,
+      "step": 14153
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.565026790063322e-07,
+      "loss": 0.0003,
+      "step": 14154
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.5647832440331221e-07,
+      "loss": 0.0002,
+      "step": 14155
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.5645396980029224e-07,
+      "loss": 0.0002,
+      "step": 14156
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.5642961519727226e-07,
+      "loss": 0.0002,
+      "step": 14157
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.564052605942523e-07,
+      "loss": 0.0004,
+      "step": 14158
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.5638090599123231e-07,
+      "loss": 0.0001,
+      "step": 14159
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.5635655138821237e-07,
+      "loss": 0.0004,
+      "step": 14160
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.563321967851924e-07,
+      "loss": 0.0004,
+      "step": 14161
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.5630784218217242e-07,
+      "loss": 0.0003,
+      "step": 14162
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.5628348757915244e-07,
+      "loss": 0.0003,
+      "step": 14163
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 1.5625913297613247e-07,
+      "loss": 0.0002,
+      "step": 14164
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5623477837311252e-07,
+      "loss": 0.0002,
+      "step": 14165
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5621042377009254e-07,
+      "loss": 0.0003,
+      "step": 14166
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5618606916707257e-07,
+      "loss": 0.0003,
+      "step": 14167
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5616171456405262e-07,
+      "loss": 0.0002,
+      "step": 14168
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5613735996103265e-07,
+      "loss": 0.0003,
+      "step": 14169
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5611300535801267e-07,
+      "loss": 0.5222,
+      "step": 14170
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.560886507549927e-07,
+      "loss": 0.0001,
+      "step": 14171
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5606429615197272e-07,
+      "loss": 0.0002,
+      "step": 14172
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5603994154895275e-07,
+      "loss": 0.0002,
+      "step": 14173
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 1.5601558694593277e-07,
+      "loss": 0.0003,
+      "step": 14174
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.559912323429128e-07,
+      "loss": 0.0002,
+      "step": 14175
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.5596687773989285e-07,
+      "loss": 0.0005,
+      "step": 14176
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.5594252313687287e-07,
+      "loss": 0.0004,
+      "step": 14177
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.559181685338529e-07,
+      "loss": 0.0003,
+      "step": 14178
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.5589381393083292e-07,
+      "loss": 0.0004,
+      "step": 14179
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.5586945932781295e-07,
+      "loss": 0.0004,
+      "step": 14180
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.5584510472479297e-07,
+      "loss": 0.0003,
+      "step": 14181
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.55820750121773e-07,
+      "loss": 0.0002,
+      "step": 14182
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.5579639551875302e-07,
+      "loss": 0.0005,
+      "step": 14183
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 1.5577204091573308e-07,
+      "loss": 0.0003,
+      "step": 14184
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.557476863127131e-07,
+      "loss": 0.0003,
+      "step": 14185
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.5572333170969313e-07,
+      "loss": 0.0003,
+      "step": 14186
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.5569897710667315e-07,
+      "loss": 0.0005,
+      "step": 14187
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.5567462250365318e-07,
+      "loss": 0.0002,
+      "step": 14188
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.556502679006332e-07,
+      "loss": 0.0003,
+      "step": 14189
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.5562591329761323e-07,
+      "loss": 0.0003,
+      "step": 14190
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.5560155869459328e-07,
+      "loss": 0.0003,
+      "step": 14191
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.555772040915733e-07,
+      "loss": 0.0005,
+      "step": 14192
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.5555284948855333e-07,
+      "loss": 0.0003,
+      "step": 14193
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.5552849488553335e-07,
+      "loss": 0.0001,
+      "step": 14194
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 1.5550414028251338e-07,
+      "loss": 0.0002,
+      "step": 14195
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.554797856794934e-07,
+      "loss": 0.0005,
+      "step": 14196
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.5545543107647343e-07,
+      "loss": 0.0003,
+      "step": 14197
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.5543107647345345e-07,
+      "loss": 0.0002,
+      "step": 14198
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.554067218704335e-07,
+      "loss": 0.0003,
+      "step": 14199
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.5538236726741353e-07,
+      "loss": 0.0003,
+      "step": 14200
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.5535801266439356e-07,
+      "loss": 0.0002,
+      "step": 14201
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.5533365806137358e-07,
+      "loss": 0.0002,
+      "step": 14202
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.5530930345835363e-07,
+      "loss": 0.0003,
+      "step": 14203
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.5528494885533366e-07,
+      "loss": 0.0002,
+      "step": 14204
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 1.5526059425231368e-07,
+      "loss": 0.0008,
+      "step": 14205
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.552362396492937e-07,
+      "loss": 0.0003,
+      "step": 14206
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.5521188504627376e-07,
+      "loss": 0.0002,
+      "step": 14207
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.5518753044325379e-07,
+      "loss": 0.0002,
+      "step": 14208
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.551631758402338e-07,
+      "loss": 0.0003,
+      "step": 14209
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.5513882123721384e-07,
+      "loss": 0.0002,
+      "step": 14210
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.5511446663419386e-07,
+      "loss": 0.0002,
+      "step": 14211
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.5509011203117389e-07,
+      "loss": 0.0006,
+      "step": 14212
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.550657574281539e-07,
+      "loss": 0.0003,
+      "step": 14213
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.5504140282513394e-07,
+      "loss": 0.0003,
+      "step": 14214
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 1.55017048222114e-07,
+      "loss": 0.0064,
+      "step": 14215
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5499269361909401e-07,
+      "loss": 0.0003,
+      "step": 14216
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5496833901607404e-07,
+      "loss": 0.0007,
+      "step": 14217
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5494398441305406e-07,
+      "loss": 0.0006,
+      "step": 14218
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.549196298100341e-07,
+      "loss": 0.0004,
+      "step": 14219
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5489527520701411e-07,
+      "loss": 0.0004,
+      "step": 14220
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5487092060399414e-07,
+      "loss": 0.0002,
+      "step": 14221
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5484656600097416e-07,
+      "loss": 0.0004,
+      "step": 14222
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5482221139795422e-07,
+      "loss": 0.0003,
+      "step": 14223
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5479785679493424e-07,
+      "loss": 0.0005,
+      "step": 14224
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 1.5477350219191427e-07,
+      "loss": 0.0002,
+      "step": 14225
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.547491475888943e-07,
+      "loss": 0.0005,
+      "step": 14226
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.5472479298587432e-07,
+      "loss": 0.0003,
+      "step": 14227
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.5470043838285434e-07,
+      "loss": 0.0003,
+      "step": 14228
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.5467608377983437e-07,
+      "loss": 0.0002,
+      "step": 14229
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.546517291768144e-07,
+      "loss": 0.0004,
+      "step": 14230
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.5462737457379444e-07,
+      "loss": 0.0003,
+      "step": 14231
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.5460301997077447e-07,
+      "loss": 0.0005,
+      "step": 14232
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.545786653677545e-07,
+      "loss": 0.0003,
+      "step": 14233
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.5455431076473452e-07,
+      "loss": 0.0004,
+      "step": 14234
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.5452995616171454e-07,
+      "loss": 0.0001,
+      "step": 14235
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 1.5450560155869457e-07,
+      "loss": 0.0002,
+      "step": 14236
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.544812469556746e-07,
+      "loss": 0.0006,
+      "step": 14237
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.5445689235265462e-07,
+      "loss": 0.0004,
+      "step": 14238
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.5443253774963467e-07,
+      "loss": 0.0002,
+      "step": 14239
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.544081831466147e-07,
+      "loss": 0.0003,
+      "step": 14240
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.5438382854359475e-07,
+      "loss": 0.0004,
+      "step": 14241
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.5435947394057477e-07,
+      "loss": 0.0003,
+      "step": 14242
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.543351193375548e-07,
+      "loss": 0.0004,
+      "step": 14243
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.5431076473453482e-07,
+      "loss": 0.0004,
+      "step": 14244
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.5428641013151485e-07,
+      "loss": 0.0003,
+      "step": 14245
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 1.542620555284949e-07,
+      "loss": 0.0003,
+      "step": 14246
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5423770092547493e-07,
+      "loss": 0.0003,
+      "step": 14247
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5421334632245495e-07,
+      "loss": 0.0003,
+      "step": 14248
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5418899171943498e-07,
+      "loss": 0.0004,
+      "step": 14249
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.54164637116415e-07,
+      "loss": 0.0002,
+      "step": 14250
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5414028251339503e-07,
+      "loss": 0.0003,
+      "step": 14251
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5411592791037505e-07,
+      "loss": 0.0002,
+      "step": 14252
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5409157330735508e-07,
+      "loss": 0.0003,
+      "step": 14253
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5406721870433513e-07,
+      "loss": 3.3612,
+      "step": 14254
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5404286410131515e-07,
+      "loss": 0.0003,
+      "step": 14255
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 1.5401850949829518e-07,
+      "loss": 0.0002,
+      "step": 14256
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.539941548952752e-07,
+      "loss": 0.0001,
+      "step": 14257
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.5396980029225523e-07,
+      "loss": 0.0002,
+      "step": 14258
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.5394544568923525e-07,
+      "loss": 0.0002,
+      "step": 14259
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.5392109108621528e-07,
+      "loss": 0.0002,
+      "step": 14260
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.538967364831953e-07,
+      "loss": 0.0004,
+      "step": 14261
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.5387238188017536e-07,
+      "loss": 0.0003,
+      "step": 14262
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.5384802727715538e-07,
+      "loss": 0.0001,
+      "step": 14263
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.538236726741354e-07,
+      "loss": 0.0005,
+      "step": 14264
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.5379931807111543e-07,
+      "loss": 0.0003,
+      "step": 14265
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.5377496346809546e-07,
+      "loss": 0.0003,
+      "step": 14266
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 1.5375060886507548e-07,
+      "loss": 0.0002,
+      "step": 14267
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.537262542620555e-07,
+      "loss": 0.0003,
+      "step": 14268
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.5370189965903553e-07,
+      "loss": 1.3767,
+      "step": 14269
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.5367754505601558e-07,
+      "loss": 0.0002,
+      "step": 14270
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.536531904529956e-07,
+      "loss": 0.0002,
+      "step": 14271
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.5362883584997563e-07,
+      "loss": 0.0003,
+      "step": 14272
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.5360448124695566e-07,
+      "loss": 0.0004,
+      "step": 14273
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.5358012664393569e-07,
+      "loss": 0.0003,
+      "step": 14274
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.535557720409157e-07,
+      "loss": 0.0002,
+      "step": 14275
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.5353141743789574e-07,
+      "loss": 0.0002,
+      "step": 14276
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 1.5350706283487576e-07,
+      "loss": 0.0002,
+      "step": 14277
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.534827082318558e-07,
+      "loss": 0.0003,
+      "step": 14278
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.5345835362883584e-07,
+      "loss": 0.0003,
+      "step": 14279
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.534339990258159e-07,
+      "loss": 0.0004,
+      "step": 14280
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.5340964442279591e-07,
+      "loss": 0.0005,
+      "step": 14281
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.5338528981977594e-07,
+      "loss": 0.0002,
+      "step": 14282
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.5336093521675596e-07,
+      "loss": 0.0002,
+      "step": 14283
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.53336580613736e-07,
+      "loss": 0.0005,
+      "step": 14284
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.5331222601071604e-07,
+      "loss": 0.0002,
+      "step": 14285
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.5328787140769607e-07,
+      "loss": 0.0003,
+      "step": 14286
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 1.532635168046761e-07,
+      "loss": 0.0002,
+      "step": 14287
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5323916220165612e-07,
+      "loss": 0.0002,
+      "step": 14288
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5321480759863614e-07,
+      "loss": 0.0002,
+      "step": 14289
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5319045299561617e-07,
+      "loss": 0.0003,
+      "step": 14290
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.531660983925962e-07,
+      "loss": 0.0002,
+      "step": 14291
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5314174378957622e-07,
+      "loss": 0.0003,
+      "step": 14292
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5311738918655627e-07,
+      "loss": 0.0029,
+      "step": 14293
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.530930345835363e-07,
+      "loss": 0.0002,
+      "step": 14294
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5306867998051632e-07,
+      "loss": 0.0002,
+      "step": 14295
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5304432537749634e-07,
+      "loss": 0.0003,
+      "step": 14296
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 1.5301997077447637e-07,
+      "loss": 0.0003,
+      "step": 14297
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.529956161714564e-07,
+      "loss": 0.0002,
+      "step": 14298
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5297126156843642e-07,
+      "loss": 0.0001,
+      "step": 14299
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5294690696541645e-07,
+      "loss": 0.0002,
+      "step": 14300
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.529225523623965e-07,
+      "loss": 0.0003,
+      "step": 14301
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5289819775937652e-07,
+      "loss": 0.0004,
+      "step": 14302
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5287384315635655e-07,
+      "loss": 0.0003,
+      "step": 14303
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5284948855333657e-07,
+      "loss": 0.0003,
+      "step": 14304
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.528251339503166e-07,
+      "loss": 0.0002,
+      "step": 14305
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5280077934729662e-07,
+      "loss": 0.0005,
+      "step": 14306
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5277642474427665e-07,
+      "loss": 0.0002,
+      "step": 14307
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 1.5275207014125667e-07,
+      "loss": 0.0003,
+      "step": 14308
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.5272771553823672e-07,
+      "loss": 0.0002,
+      "step": 14309
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.5270336093521675e-07,
+      "loss": 0.0002,
+      "step": 14310
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.5267900633219678e-07,
+      "loss": 0.0002,
+      "step": 14311
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.526546517291768e-07,
+      "loss": 0.0004,
+      "step": 14312
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.5263029712615683e-07,
+      "loss": 0.0004,
+      "step": 14313
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.5260594252313685e-07,
+      "loss": 0.0002,
+      "step": 14314
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.5258158792011688e-07,
+      "loss": 0.0003,
+      "step": 14315
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.525572333170969e-07,
+      "loss": 0.0002,
+      "step": 14316
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.5253287871407695e-07,
+      "loss": 0.0003,
+      "step": 14317
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 1.52508524111057e-07,
+      "loss": 0.0002,
+      "step": 14318
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5248416950803703e-07,
+      "loss": 0.0002,
+      "step": 14319
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5245981490501705e-07,
+      "loss": 0.0003,
+      "step": 14320
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5243546030199708e-07,
+      "loss": 0.0004,
+      "step": 14321
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.524111056989771e-07,
+      "loss": 0.0002,
+      "step": 14322
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5238675109595713e-07,
+      "loss": 0.0002,
+      "step": 14323
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5236239649293716e-07,
+      "loss": 0.0002,
+      "step": 14324
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.523380418899172e-07,
+      "loss": 0.0002,
+      "step": 14325
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5231368728689723e-07,
+      "loss": 0.0008,
+      "step": 14326
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5228933268387726e-07,
+      "loss": 0.0002,
+      "step": 14327
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 1.5226497808085728e-07,
+      "loss": 0.0003,
+      "step": 14328
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.522406234778373e-07,
+      "loss": 0.0003,
+      "step": 14329
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5221626887481733e-07,
+      "loss": 0.0003,
+      "step": 14330
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5219191427179736e-07,
+      "loss": 0.0002,
+      "step": 14331
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5216755966877738e-07,
+      "loss": 0.0002,
+      "step": 14332
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5214320506575743e-07,
+      "loss": 0.0003,
+      "step": 14333
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5211885046273746e-07,
+      "loss": 0.0004,
+      "step": 14334
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5209449585971748e-07,
+      "loss": 0.0004,
+      "step": 14335
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.520701412566975e-07,
+      "loss": 0.0003,
+      "step": 14336
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5204578665367754e-07,
+      "loss": 0.0002,
+      "step": 14337
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5202143205065756e-07,
+      "loss": 0.0002,
+      "step": 14338
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 1.5199707744763759e-07,
+      "loss": 0.0002,
+      "step": 14339
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.519727228446176e-07,
+      "loss": 0.0002,
+      "step": 14340
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.5194836824159766e-07,
+      "loss": 0.0003,
+      "step": 14341
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.519240136385777e-07,
+      "loss": 0.0002,
+      "step": 14342
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.518996590355577e-07,
+      "loss": 0.0003,
+      "step": 14343
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.5187530443253774e-07,
+      "loss": 0.0003,
+      "step": 14344
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.5185094982951776e-07,
+      "loss": 0.0002,
+      "step": 14345
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.518265952264978e-07,
+      "loss": 0.0003,
+      "step": 14346
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.5180224062347781e-07,
+      "loss": 0.0004,
+      "step": 14347
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.5177788602045784e-07,
+      "loss": 0.0004,
+      "step": 14348
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 1.517535314174379e-07,
+      "loss": 0.0004,
+      "step": 14349
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5172917681441792e-07,
+      "loss": 0.0003,
+      "step": 14350
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5170482221139794e-07,
+      "loss": 0.0004,
+      "step": 14351
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5168046760837797e-07,
+      "loss": 0.0003,
+      "step": 14352
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.51656113005358e-07,
+      "loss": 0.0002,
+      "step": 14353
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5163175840233802e-07,
+      "loss": 0.0004,
+      "step": 14354
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5160740379931804e-07,
+      "loss": 0.0002,
+      "step": 14355
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5158304919629812e-07,
+      "loss": 0.0003,
+      "step": 14356
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5155869459327814e-07,
+      "loss": 0.0005,
+      "step": 14357
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.5153433999025817e-07,
+      "loss": 0.0003,
+      "step": 14358
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 1.515099853872382e-07,
+      "loss": 0.0003,
+      "step": 14359
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5148563078421822e-07,
+      "loss": 0.0002,
+      "step": 14360
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5146127618119825e-07,
+      "loss": 0.0002,
+      "step": 14361
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5143692157817827e-07,
+      "loss": 0.0002,
+      "step": 14362
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.514125669751583e-07,
+      "loss": 0.0003,
+      "step": 14363
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5138821237213835e-07,
+      "loss": 0.0004,
+      "step": 14364
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5136385776911837e-07,
+      "loss": 0.0003,
+      "step": 14365
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.513395031660984e-07,
+      "loss": 0.0003,
+      "step": 14366
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5131514856307842e-07,
+      "loss": 0.0002,
+      "step": 14367
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5129079396005845e-07,
+      "loss": 0.0004,
+      "step": 14368
+    },
+    {
+      "epoch": 13.96,
+      "learning_rate": 1.5126643935703847e-07,
+      "loss": 0.0002,
+      "step": 14369
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.512420847540185e-07,
+      "loss": 0.0004,
+      "step": 14370
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5121773015099852e-07,
+      "loss": 0.0003,
+      "step": 14371
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5119337554797857e-07,
+      "loss": 0.0002,
+      "step": 14372
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.511690209449586e-07,
+      "loss": 0.0002,
+      "step": 14373
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5114466634193863e-07,
+      "loss": 0.0004,
+      "step": 14374
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5112031173891865e-07,
+      "loss": 0.0002,
+      "step": 14375
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5109595713589868e-07,
+      "loss": 0.0002,
+      "step": 14376
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.510716025328787e-07,
+      "loss": 0.0002,
+      "step": 14377
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5104724792985873e-07,
+      "loss": 0.0003,
+      "step": 14378
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.5102289332683875e-07,
+      "loss": 0.0002,
+      "step": 14379
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 1.509985387238188e-07,
+      "loss": 0.0002,
+      "step": 14380
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5097418412079883e-07,
+      "loss": 0.0002,
+      "step": 14381
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5094982951777885e-07,
+      "loss": 0.0002,
+      "step": 14382
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5092547491475888e-07,
+      "loss": 0.0003,
+      "step": 14383
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.509011203117389e-07,
+      "loss": 0.0001,
+      "step": 14384
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5087676570871893e-07,
+      "loss": 0.0002,
+      "step": 14385
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5085241110569895e-07,
+      "loss": 0.0004,
+      "step": 14386
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5082805650267898e-07,
+      "loss": 0.0003,
+      "step": 14387
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5080370189965903e-07,
+      "loss": 0.0005,
+      "step": 14388
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5077934729663906e-07,
+      "loss": 0.0003,
+      "step": 14389
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 1.5075499269361908e-07,
+      "loss": 0.0003,
+      "step": 14390
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.507306380905991e-07,
+      "loss": 0.0004,
+      "step": 14391
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5070628348757913e-07,
+      "loss": 0.0003,
+      "step": 14392
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5068192888455916e-07,
+      "loss": 0.0013,
+      "step": 14393
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5065757428153918e-07,
+      "loss": 0.0001,
+      "step": 14394
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5063321967851926e-07,
+      "loss": 0.0002,
+      "step": 14395
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5060886507549928e-07,
+      "loss": 0.0003,
+      "step": 14396
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.505845104724793e-07,
+      "loss": 0.0002,
+      "step": 14397
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5056015586945934e-07,
+      "loss": 0.0003,
+      "step": 14398
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5053580126643936e-07,
+      "loss": 0.0002,
+      "step": 14399
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 1.5051144666341939e-07,
+      "loss": 0.0004,
+      "step": 14400
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.504870920603994e-07,
+      "loss": 0.0002,
+      "step": 14401
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.5046273745737944e-07,
+      "loss": 0.0004,
+      "step": 14402
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.504383828543595e-07,
+      "loss": 0.0002,
+      "step": 14403
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.504140282513395e-07,
+      "loss": 0.0002,
+      "step": 14404
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.5038967364831954e-07,
+      "loss": 0.0001,
+      "step": 14405
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.5036531904529956e-07,
+      "loss": 0.0006,
+      "step": 14406
+    },
+    {
+      "epoch": 14.0,
+      "eval_accuracy": 0.9990277102576568,
+      "eval_loss": 0.014307250268757343,
+      "eval_runtime": 2.7864,
+      "eval_samples_per_second": 738.236,
+      "eval_steps_per_second": 11.843,
+      "step": 14406
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.503409644422796e-07,
+      "loss": 0.0002,
+      "step": 14407
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.503166098392596e-07,
+      "loss": 0.0004,
+      "step": 14408
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.5029225523623964e-07,
+      "loss": 0.0002,
+      "step": 14409
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.5026790063321966e-07,
+      "loss": 0.0003,
+      "step": 14410
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 1.5024354603019972e-07,
+      "loss": 0.0003,
+      "step": 14411
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5021919142717974e-07,
+      "loss": 0.0003,
+      "step": 14412
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5019483682415977e-07,
+      "loss": 0.0003,
+      "step": 14413
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.501704822211398e-07,
+      "loss": 0.0002,
+      "step": 14414
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5014612761811982e-07,
+      "loss": 0.0003,
+      "step": 14415
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5012177301509984e-07,
+      "loss": 0.0002,
+      "step": 14416
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5009741841207987e-07,
+      "loss": 0.0009,
+      "step": 14417
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.500730638090599e-07,
+      "loss": 0.0003,
+      "step": 14418
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5004870920603994e-07,
+      "loss": 0.0005,
+      "step": 14419
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5002435460301997e-07,
+      "loss": 0.0002,
+      "step": 14420
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 1.5e-07,
+      "loss": 0.0005,
+      "step": 14421
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.4997564539698002e-07,
+      "loss": 0.0005,
+      "step": 14422
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.4995129079396004e-07,
+      "loss": 0.0003,
+      "step": 14423
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.4992693619094007e-07,
+      "loss": 0.0004,
+      "step": 14424
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.499025815879201e-07,
+      "loss": 0.0002,
+      "step": 14425
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.4987822698490012e-07,
+      "loss": 0.0003,
+      "step": 14426
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.4985387238188017e-07,
+      "loss": 0.0003,
+      "step": 14427
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.498295177788602e-07,
+      "loss": 0.0003,
+      "step": 14428
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.4980516317584022e-07,
+      "loss": 0.0002,
+      "step": 14429
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.4978080857282025e-07,
+      "loss": 0.0003,
+      "step": 14430
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 1.4975645396980027e-07,
+      "loss": 0.0003,
+      "step": 14431
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.497320993667803e-07,
+      "loss": 0.0002,
+      "step": 14432
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4970774476376035e-07,
+      "loss": 0.0005,
+      "step": 14433
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4968339016074037e-07,
+      "loss": 0.0005,
+      "step": 14434
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4965903555772042e-07,
+      "loss": 0.0002,
+      "step": 14435
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4963468095470045e-07,
+      "loss": 0.0002,
+      "step": 14436
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4961032635168048e-07,
+      "loss": 0.0003,
+      "step": 14437
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.495859717486605e-07,
+      "loss": 0.0003,
+      "step": 14438
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4956161714564053e-07,
+      "loss": 0.0002,
+      "step": 14439
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4953726254262055e-07,
+      "loss": 0.0003,
+      "step": 14440
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.4951290793960058e-07,
+      "loss": 0.0003,
+      "step": 14441
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.494885533365806e-07,
+      "loss": 0.0003,
+      "step": 14442
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.4946419873356065e-07,
+      "loss": 0.0002,
+      "step": 14443
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.4943984413054068e-07,
+      "loss": 0.0003,
+      "step": 14444
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.494154895275207e-07,
+      "loss": 0.0003,
+      "step": 14445
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.4939113492450073e-07,
+      "loss": 0.0003,
+      "step": 14446
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.4936678032148075e-07,
+      "loss": 0.0002,
+      "step": 14447
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.4934242571846078e-07,
+      "loss": 0.0004,
+      "step": 14448
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.493180711154408e-07,
+      "loss": 0.0003,
+      "step": 14449
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.4929371651242083e-07,
+      "loss": 0.0004,
+      "step": 14450
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.4926936190940088e-07,
+      "loss": 0.0003,
+      "step": 14451
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 1.492450073063809e-07,
+      "loss": 0.0002,
+      "step": 14452
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.4922065270336093e-07,
+      "loss": 0.0002,
+      "step": 14453
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.4919629810034096e-07,
+      "loss": 0.0002,
+      "step": 14454
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.4917194349732098e-07,
+      "loss": 0.0002,
+      "step": 14455
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.49147588894301e-07,
+      "loss": 0.0002,
+      "step": 14456
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.4912323429128103e-07,
+      "loss": 0.0004,
+      "step": 14457
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.4909887968826108e-07,
+      "loss": 0.0002,
+      "step": 14458
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.490745250852411e-07,
+      "loss": 0.0002,
+      "step": 14459
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.4905017048222113e-07,
+      "loss": 0.0001,
+      "step": 14460
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.4902581587920116e-07,
+      "loss": 0.0003,
+      "step": 14461
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 1.4900146127618118e-07,
+      "loss": 0.0003,
+      "step": 14462
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.489771066731612e-07,
+      "loss": 0.0002,
+      "step": 14463
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.4895275207014123e-07,
+      "loss": 0.0002,
+      "step": 14464
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.4892839746712126e-07,
+      "loss": 0.0003,
+      "step": 14465
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.489040428641013e-07,
+      "loss": 0.0002,
+      "step": 14466
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.4887968826108134e-07,
+      "loss": 0.0002,
+      "step": 14467
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.4885533365806136e-07,
+      "loss": 0.0002,
+      "step": 14468
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.4883097905504139e-07,
+      "loss": 0.0002,
+      "step": 14469
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.488066244520214e-07,
+      "loss": 0.0001,
+      "step": 14470
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.4878226984900146e-07,
+      "loss": 0.0002,
+      "step": 14471
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 1.487579152459815e-07,
+      "loss": 0.0002,
+      "step": 14472
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.4873356064296151e-07,
+      "loss": 0.0006,
+      "step": 14473
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.4870920603994157e-07,
+      "loss": 0.0002,
+      "step": 14474
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.486848514369216e-07,
+      "loss": 0.0003,
+      "step": 14475
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.4866049683390162e-07,
+      "loss": 0.0003,
+      "step": 14476
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.4863614223088164e-07,
+      "loss": 0.0002,
+      "step": 14477
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.4861178762786167e-07,
+      "loss": 0.0002,
+      "step": 14478
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.485874330248417e-07,
+      "loss": 0.0001,
+      "step": 14479
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.4856307842182172e-07,
+      "loss": 0.0003,
+      "step": 14480
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.4853872381880174e-07,
+      "loss": 0.0003,
+      "step": 14481
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.485143692157818e-07,
+      "loss": 0.0004,
+      "step": 14482
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 1.4849001461276182e-07,
+      "loss": 0.0003,
+      "step": 14483
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4846566000974184e-07,
+      "loss": 0.0004,
+      "step": 14484
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4844130540672187e-07,
+      "loss": 0.0003,
+      "step": 14485
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.484169508037019e-07,
+      "loss": 0.0004,
+      "step": 14486
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4839259620068192e-07,
+      "loss": 0.0002,
+      "step": 14487
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4836824159766194e-07,
+      "loss": 0.0002,
+      "step": 14488
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4834388699464197e-07,
+      "loss": 0.0002,
+      "step": 14489
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4831953239162202e-07,
+      "loss": 0.0003,
+      "step": 14490
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4829517778860205e-07,
+      "loss": 0.0002,
+      "step": 14491
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.4827082318558207e-07,
+      "loss": 0.0002,
+      "step": 14492
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.482464685825621e-07,
+      "loss": 0.0003,
+      "step": 14493
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.4822211397954212e-07,
+      "loss": 0.0002,
+      "step": 14494
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.4819775937652215e-07,
+      "loss": 0.0002,
+      "step": 14495
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.4817340477350217e-07,
+      "loss": 0.0003,
+      "step": 14496
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.481490501704822e-07,
+      "loss": 0.0002,
+      "step": 14497
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.4812469556746225e-07,
+      "loss": 0.0004,
+      "step": 14498
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.4810034096444227e-07,
+      "loss": 0.0004,
+      "step": 14499
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.480759863614223e-07,
+      "loss": 0.0002,
+      "step": 14500
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.4805163175840232e-07,
+      "loss": 0.0002,
+      "step": 14501
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.4802727715538235e-07,
+      "loss": 0.0004,
+      "step": 14502
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 1.4800292255236237e-07,
+      "loss": 0.0005,
+      "step": 14503
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.479785679493424e-07,
+      "loss": 0.0002,
+      "step": 14504
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.4795421334632242e-07,
+      "loss": 0.0005,
+      "step": 14505
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.4792985874330248e-07,
+      "loss": 0.0003,
+      "step": 14506
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.479055041402825e-07,
+      "loss": 0.0003,
+      "step": 14507
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.4788114953726253e-07,
+      "loss": 0.0002,
+      "step": 14508
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.4785679493424255e-07,
+      "loss": 0.0002,
+      "step": 14509
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.478324403312226e-07,
+      "loss": 0.0004,
+      "step": 14510
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.4780808572820263e-07,
+      "loss": 0.0003,
+      "step": 14511
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.4778373112518265e-07,
+      "loss": 0.0004,
+      "step": 14512
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.477593765221627e-07,
+      "loss": 0.0002,
+      "step": 14513
+    },
+    {
+      "epoch": 14.1,
+      "learning_rate": 1.4773502191914273e-07,
+      "loss": 0.0003,
+      "step": 14514
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4771066731612276e-07,
+      "loss": 0.0002,
+      "step": 14515
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4768631271310278e-07,
+      "loss": 0.0002,
+      "step": 14516
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.476619581100828e-07,
+      "loss": 0.0004,
+      "step": 14517
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4763760350706283e-07,
+      "loss": 0.0002,
+      "step": 14518
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4761324890404286e-07,
+      "loss": 0.0003,
+      "step": 14519
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4758889430102288e-07,
+      "loss": 0.0002,
+      "step": 14520
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4756453969800293e-07,
+      "loss": 0.0002,
+      "step": 14521
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4754018509498296e-07,
+      "loss": 0.0002,
+      "step": 14522
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.4751583049196298e-07,
+      "loss": 0.0002,
+      "step": 14523
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 1.47491475888943e-07,
+      "loss": 0.0003,
+      "step": 14524
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.4746712128592303e-07,
+      "loss": 0.0002,
+      "step": 14525
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.4744276668290306e-07,
+      "loss": 0.0001,
+      "step": 14526
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.4741841207988308e-07,
+      "loss": 0.0001,
+      "step": 14527
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.473940574768631e-07,
+      "loss": 0.0002,
+      "step": 14528
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.4736970287384316e-07,
+      "loss": 0.0001,
+      "step": 14529
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.4734534827082319e-07,
+      "loss": 0.0003,
+      "step": 14530
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.473209936678032e-07,
+      "loss": 0.0002,
+      "step": 14531
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.4729663906478324e-07,
+      "loss": 0.1079,
+      "step": 14532
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.4727228446176326e-07,
+      "loss": 0.0003,
+      "step": 14533
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 1.472479298587433e-07,
+      "loss": 0.0002,
+      "step": 14534
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.472235752557233e-07,
+      "loss": 0.0005,
+      "step": 14535
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4719922065270334e-07,
+      "loss": 0.0002,
+      "step": 14536
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.471748660496834e-07,
+      "loss": 0.0004,
+      "step": 14537
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4715051144666341e-07,
+      "loss": 0.0002,
+      "step": 14538
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4712615684364344e-07,
+      "loss": 0.0003,
+      "step": 14539
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4710180224062346e-07,
+      "loss": 0.0002,
+      "step": 14540
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.470774476376035e-07,
+      "loss": 0.0005,
+      "step": 14541
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4705309303458351e-07,
+      "loss": 0.0004,
+      "step": 14542
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4702873843156354e-07,
+      "loss": 0.0002,
+      "step": 14543
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.4700438382854356e-07,
+      "loss": 0.0003,
+      "step": 14544
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4698002922552362e-07,
+      "loss": 0.0002,
+      "step": 14545
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4695567462250364e-07,
+      "loss": 0.0002,
+      "step": 14546
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4693132001948367e-07,
+      "loss": 0.0004,
+      "step": 14547
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4690696541646372e-07,
+      "loss": 0.0003,
+      "step": 14548
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4688261081344374e-07,
+      "loss": 0.0002,
+      "step": 14549
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4685825621042377e-07,
+      "loss": 0.0004,
+      "step": 14550
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.468339016074038e-07,
+      "loss": 0.0004,
+      "step": 14551
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4680954700438382e-07,
+      "loss": 0.0003,
+      "step": 14552
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4678519240136387e-07,
+      "loss": 0.0002,
+      "step": 14553
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.467608377983439e-07,
+      "loss": 0.0003,
+      "step": 14554
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 1.4673648319532392e-07,
+      "loss": 0.0003,
+      "step": 14555
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.4671212859230395e-07,
+      "loss": 0.0003,
+      "step": 14556
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.4668777398928397e-07,
+      "loss": 0.0015,
+      "step": 14557
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.46663419386264e-07,
+      "loss": 0.0003,
+      "step": 14558
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.4663906478324402e-07,
+      "loss": 0.0002,
+      "step": 14559
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.4661471018022407e-07,
+      "loss": 0.0002,
+      "step": 14560
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.465903555772041e-07,
+      "loss": 0.0003,
+      "step": 14561
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.4656600097418412e-07,
+      "loss": 0.0004,
+      "step": 14562
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.4654164637116415e-07,
+      "loss": 0.0014,
+      "step": 14563
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.4651729176814417e-07,
+      "loss": 0.0003,
+      "step": 14564
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 1.464929371651242e-07,
+      "loss": 0.0002,
+      "step": 14565
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4646858256210422e-07,
+      "loss": 0.0003,
+      "step": 14566
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4644422795908425e-07,
+      "loss": 0.0002,
+      "step": 14567
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.464198733560643e-07,
+      "loss": 0.0005,
+      "step": 14568
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4639551875304433e-07,
+      "loss": 0.0002,
+      "step": 14569
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4637116415002435e-07,
+      "loss": 0.0003,
+      "step": 14570
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4634680954700438e-07,
+      "loss": 0.0002,
+      "step": 14571
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.463224549439844e-07,
+      "loss": 0.0004,
+      "step": 14572
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4629810034096443e-07,
+      "loss": 0.0002,
+      "step": 14573
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4627374573794445e-07,
+      "loss": 0.0004,
+      "step": 14574
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 1.4624939113492448e-07,
+      "loss": 0.0003,
+      "step": 14575
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4622503653190453e-07,
+      "loss": 0.0002,
+      "step": 14576
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4620068192888455e-07,
+      "loss": 0.0003,
+      "step": 14577
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4617632732586458e-07,
+      "loss": 0.0002,
+      "step": 14578
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.461519727228446e-07,
+      "loss": 0.0003,
+      "step": 14579
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4612761811982463e-07,
+      "loss": 0.0003,
+      "step": 14580
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4610326351680465e-07,
+      "loss": 0.0002,
+      "step": 14581
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4607890891378468e-07,
+      "loss": 0.0002,
+      "step": 14582
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.460545543107647e-07,
+      "loss": 0.0002,
+      "step": 14583
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4603019970774476e-07,
+      "loss": 0.0002,
+      "step": 14584
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4600584510472478e-07,
+      "loss": 0.0004,
+      "step": 14585
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 1.4598149050170483e-07,
+      "loss": 0.0003,
+      "step": 14586
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4595713589868486e-07,
+      "loss": 0.0002,
+      "step": 14587
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4593278129566488e-07,
+      "loss": 0.0002,
+      "step": 14588
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.459084266926449e-07,
+      "loss": 0.0002,
+      "step": 14589
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4588407208962493e-07,
+      "loss": 0.0005,
+      "step": 14590
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4585971748660496e-07,
+      "loss": 0.0002,
+      "step": 14591
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.45835362883585e-07,
+      "loss": 0.0003,
+      "step": 14592
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4581100828056504e-07,
+      "loss": 0.0002,
+      "step": 14593
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4578665367754506e-07,
+      "loss": 0.0012,
+      "step": 14594
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.4576229907452509e-07,
+      "loss": 0.0002,
+      "step": 14595
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 1.457379444715051e-07,
+      "loss": 0.0001,
+      "step": 14596
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.4571358986848514e-07,
+      "loss": 0.0001,
+      "step": 14597
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.4568923526546516e-07,
+      "loss": 0.0003,
+      "step": 14598
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.456648806624452e-07,
+      "loss": 0.0003,
+      "step": 14599
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.4564052605942524e-07,
+      "loss": 0.0002,
+      "step": 14600
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.4561617145640526e-07,
+      "loss": 0.0002,
+      "step": 14601
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.455918168533853e-07,
+      "loss": 0.0002,
+      "step": 14602
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.4556746225036531e-07,
+      "loss": 0.0003,
+      "step": 14603
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.4554310764734534e-07,
+      "loss": 0.0004,
+      "step": 14604
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.4551875304432536e-07,
+      "loss": 0.0002,
+      "step": 14605
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.454943984413054e-07,
+      "loss": 0.0003,
+      "step": 14606
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.4547004383828542e-07,
+      "loss": 0.0001,
+      "step": 14607
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.4544568923526547e-07,
+      "loss": 0.0003,
+      "step": 14608
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.454213346322455e-07,
+      "loss": 0.0002,
+      "step": 14609
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.4539698002922552e-07,
+      "loss": 0.0004,
+      "step": 14610
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.4537262542620554e-07,
+      "loss": 0.0001,
+      "step": 14611
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.4534827082318557e-07,
+      "loss": 0.0003,
+      "step": 14612
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.453239162201656e-07,
+      "loss": 0.0002,
+      "step": 14613
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.4529956161714562e-07,
+      "loss": 0.0005,
+      "step": 14614
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.4527520701412564e-07,
+      "loss": 0.0002,
+      "step": 14615
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 1.452508524111057e-07,
+      "loss": 0.0003,
+      "step": 14616
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4522649780808572e-07,
+      "loss": 0.0003,
+      "step": 14617
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4520214320506574e-07,
+      "loss": 0.0004,
+      "step": 14618
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4517778860204577e-07,
+      "loss": 0.0002,
+      "step": 14619
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.451534339990258e-07,
+      "loss": 0.0002,
+      "step": 14620
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4512907939600582e-07,
+      "loss": 0.0002,
+      "step": 14621
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4510472479298585e-07,
+      "loss": 0.0002,
+      "step": 14622
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4508037018996587e-07,
+      "loss": 0.0003,
+      "step": 14623
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4505601558694592e-07,
+      "loss": 0.0003,
+      "step": 14624
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4503166098392597e-07,
+      "loss": 0.0002,
+      "step": 14625
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.45007306380906e-07,
+      "loss": 0.0004,
+      "step": 14626
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 1.4498295177788602e-07,
+      "loss": 0.0002,
+      "step": 14627
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.4495859717486605e-07,
+      "loss": 0.0003,
+      "step": 14628
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.4493424257184607e-07,
+      "loss": 0.0003,
+      "step": 14629
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.449098879688261e-07,
+      "loss": 0.0003,
+      "step": 14630
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.4488553336580615e-07,
+      "loss": 0.0002,
+      "step": 14631
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.4486117876278618e-07,
+      "loss": 0.0003,
+      "step": 14632
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.448368241597662e-07,
+      "loss": 0.0002,
+      "step": 14633
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.4481246955674623e-07,
+      "loss": 0.0002,
+      "step": 14634
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.4478811495372625e-07,
+      "loss": 0.0003,
+      "step": 14635
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.4476376035070628e-07,
+      "loss": 0.0005,
+      "step": 14636
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 1.447394057476863e-07,
+      "loss": 0.0003,
+      "step": 14637
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4471505114466633e-07,
+      "loss": 0.0002,
+      "step": 14638
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4469069654164638e-07,
+      "loss": 0.0002,
+      "step": 14639
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.446663419386264e-07,
+      "loss": 0.0003,
+      "step": 14640
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4464198733560643e-07,
+      "loss": 0.0003,
+      "step": 14641
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4461763273258645e-07,
+      "loss": 0.0003,
+      "step": 14642
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4459327812956648e-07,
+      "loss": 0.0004,
+      "step": 14643
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.445689235265465e-07,
+      "loss": 0.0003,
+      "step": 14644
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4454456892352653e-07,
+      "loss": 0.0005,
+      "step": 14645
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.4452021432050656e-07,
+      "loss": 0.0004,
+      "step": 14646
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 1.444958597174866e-07,
+      "loss": 0.0002,
+      "step": 14647
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4447150511446663e-07,
+      "loss": 0.0002,
+      "step": 14648
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4444715051144666e-07,
+      "loss": 0.0002,
+      "step": 14649
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4442279590842668e-07,
+      "loss": 0.0003,
+      "step": 14650
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.443984413054067e-07,
+      "loss": 0.0002,
+      "step": 14651
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4437408670238673e-07,
+      "loss": 0.0002,
+      "step": 14652
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4434973209936676e-07,
+      "loss": 0.0005,
+      "step": 14653
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4432537749634678e-07,
+      "loss": 0.0003,
+      "step": 14654
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4430102289332683e-07,
+      "loss": 0.0002,
+      "step": 14655
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4427666829030686e-07,
+      "loss": 0.0002,
+      "step": 14656
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.4425231368728689e-07,
+      "loss": 0.0003,
+      "step": 14657
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.442279590842669e-07,
+      "loss": 0.0004,
+      "step": 14658
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4420360448124694e-07,
+      "loss": 0.0004,
+      "step": 14659
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4417924987822696e-07,
+      "loss": 0.0002,
+      "step": 14660
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4415489527520699e-07,
+      "loss": 0.0003,
+      "step": 14661
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.44130540672187e-07,
+      "loss": 0.0004,
+      "step": 14662
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.441061860691671e-07,
+      "loss": 0.0006,
+      "step": 14663
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4408183146614711e-07,
+      "loss": 0.0004,
+      "step": 14664
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4405747686312714e-07,
+      "loss": 0.0002,
+      "step": 14665
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4403312226010716e-07,
+      "loss": 0.0002,
+      "step": 14666
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.440087676570872e-07,
+      "loss": 0.0004,
+      "step": 14667
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 1.4398441305406721e-07,
+      "loss": 0.0002,
+      "step": 14668
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4396005845104724e-07,
+      "loss": 0.0002,
+      "step": 14669
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.439357038480273e-07,
+      "loss": 0.0005,
+      "step": 14670
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4391134924500732e-07,
+      "loss": 0.0004,
+      "step": 14671
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4388699464198734e-07,
+      "loss": 0.0002,
+      "step": 14672
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4386264003896737e-07,
+      "loss": 0.0002,
+      "step": 14673
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.438382854359474e-07,
+      "loss": 0.0004,
+      "step": 14674
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4381393083292742e-07,
+      "loss": 0.0002,
+      "step": 14675
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4378957622990744e-07,
+      "loss": 0.0003,
+      "step": 14676
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4376522162688747e-07,
+      "loss": 0.0003,
+      "step": 14677
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 1.4374086702386752e-07,
+      "loss": 0.0002,
+      "step": 14678
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.4371651242084754e-07,
+      "loss": 0.0003,
+      "step": 14679
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.4369215781782757e-07,
+      "loss": 0.0003,
+      "step": 14680
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.436678032148076e-07,
+      "loss": 0.0004,
+      "step": 14681
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.4364344861178762e-07,
+      "loss": 0.0005,
+      "step": 14682
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.4361909400876765e-07,
+      "loss": 0.0003,
+      "step": 14683
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.4359473940574767e-07,
+      "loss": 0.0003,
+      "step": 14684
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.435703848027277e-07,
+      "loss": 0.0003,
+      "step": 14685
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.4354603019970775e-07,
+      "loss": 0.0003,
+      "step": 14686
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.4352167559668777e-07,
+      "loss": 0.0007,
+      "step": 14687
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 1.434973209936678e-07,
+      "loss": 0.0005,
+      "step": 14688
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4347296639064782e-07,
+      "loss": 0.0001,
+      "step": 14689
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4344861178762785e-07,
+      "loss": 0.0003,
+      "step": 14690
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4342425718460787e-07,
+      "loss": 0.0002,
+      "step": 14691
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.433999025815879e-07,
+      "loss": 0.0002,
+      "step": 14692
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4337554797856792e-07,
+      "loss": 0.0003,
+      "step": 14693
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4335119337554798e-07,
+      "loss": 0.0004,
+      "step": 14694
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.43326838772528e-07,
+      "loss": 0.0004,
+      "step": 14695
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4330248416950803e-07,
+      "loss": 0.0002,
+      "step": 14696
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4327812956648805e-07,
+      "loss": 0.0002,
+      "step": 14697
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.4325377496346808e-07,
+      "loss": 0.0002,
+      "step": 14698
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 1.432294203604481e-07,
+      "loss": 0.0003,
+      "step": 14699
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4320506575742813e-07,
+      "loss": 0.0005,
+      "step": 14700
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4318071115440818e-07,
+      "loss": 0.0002,
+      "step": 14701
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4315635655138823e-07,
+      "loss": 0.0005,
+      "step": 14702
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4313200194836825e-07,
+      "loss": 0.0003,
+      "step": 14703
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4310764734534828e-07,
+      "loss": 0.0126,
+      "step": 14704
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.430832927423283e-07,
+      "loss": 0.0002,
+      "step": 14705
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4305893813930833e-07,
+      "loss": 0.0002,
+      "step": 14706
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4303458353628836e-07,
+      "loss": 0.0001,
+      "step": 14707
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.4301022893326838e-07,
+      "loss": 0.0003,
+      "step": 14708
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.429858743302484e-07,
+      "loss": 0.0001,
+      "step": 14709
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4296151972722846e-07,
+      "loss": 0.0002,
+      "step": 14710
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4293716512420848e-07,
+      "loss": 0.0003,
+      "step": 14711
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.429128105211885e-07,
+      "loss": 0.0004,
+      "step": 14712
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4288845591816853e-07,
+      "loss": 0.0003,
+      "step": 14713
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4286410131514856e-07,
+      "loss": 0.0002,
+      "step": 14714
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4283974671212858e-07,
+      "loss": 0.0002,
+      "step": 14715
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.428153921091086e-07,
+      "loss": 0.0003,
+      "step": 14716
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4279103750608863e-07,
+      "loss": 0.0002,
+      "step": 14717
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.4276668290306868e-07,
+      "loss": 0.0002,
+      "step": 14718
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 1.427423283000487e-07,
+      "loss": 0.0002,
+      "step": 14719
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.4271797369702874e-07,
+      "loss": 0.0004,
+      "step": 14720
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.4269361909400876e-07,
+      "loss": 0.0004,
+      "step": 14721
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.4266926449098879e-07,
+      "loss": 0.0003,
+      "step": 14722
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.426449098879688e-07,
+      "loss": 0.0004,
+      "step": 14723
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.4262055528494884e-07,
+      "loss": 0.0002,
+      "step": 14724
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.425962006819289e-07,
+      "loss": 0.0002,
+      "step": 14725
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.425718460789089e-07,
+      "loss": 0.0003,
+      "step": 14726
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.4254749147588894e-07,
+      "loss": 0.0003,
+      "step": 14727
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.4252313687286896e-07,
+      "loss": 0.0003,
+      "step": 14728
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.42498782269849e-07,
+      "loss": 0.0002,
+      "step": 14729
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 1.4247442766682901e-07,
+      "loss": 0.0002,
+      "step": 14730
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4245007306380904e-07,
+      "loss": 0.0003,
+      "step": 14731
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4242571846078906e-07,
+      "loss": 0.0002,
+      "step": 14732
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4240136385776912e-07,
+      "loss": 0.0003,
+      "step": 14733
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4237700925474914e-07,
+      "loss": 0.0002,
+      "step": 14734
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4235265465172917e-07,
+      "loss": 0.0002,
+      "step": 14735
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.423283000487092e-07,
+      "loss": 0.0002,
+      "step": 14736
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4230394544568922e-07,
+      "loss": 0.0004,
+      "step": 14737
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4227959084266924e-07,
+      "loss": 0.0003,
+      "step": 14738
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.422552362396493e-07,
+      "loss": 0.0003,
+      "step": 14739
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 1.4223088163662932e-07,
+      "loss": 0.0003,
+      "step": 14740
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.4220652703360937e-07,
+      "loss": 0.0002,
+      "step": 14741
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.421821724305894e-07,
+      "loss": 0.0002,
+      "step": 14742
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.4215781782756942e-07,
+      "loss": 0.0003,
+      "step": 14743
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.4213346322454945e-07,
+      "loss": 0.0002,
+      "step": 14744
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.4210910862152947e-07,
+      "loss": 0.0004,
+      "step": 14745
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.420847540185095e-07,
+      "loss": 0.0004,
+      "step": 14746
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.4206039941548952e-07,
+      "loss": 0.0006,
+      "step": 14747
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.4203604481246955e-07,
+      "loss": 0.0002,
+      "step": 14748
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.420116902094496e-07,
+      "loss": 0.0003,
+      "step": 14749
+    },
+    {
+      "epoch": 14.33,
+      "learning_rate": 1.4198733560642962e-07,
+      "loss": 0.0002,
+      "step": 14750
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4196298100340965e-07,
+      "loss": 0.0003,
+      "step": 14751
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4193862640038967e-07,
+      "loss": 0.0003,
+      "step": 14752
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.419142717973697e-07,
+      "loss": 0.0003,
+      "step": 14753
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4188991719434972e-07,
+      "loss": 0.0002,
+      "step": 14754
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4186556259132975e-07,
+      "loss": 0.0003,
+      "step": 14755
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4184120798830977e-07,
+      "loss": 0.0002,
+      "step": 14756
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4181685338528983e-07,
+      "loss": 0.0002,
+      "step": 14757
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4179249878226985e-07,
+      "loss": 0.0004,
+      "step": 14758
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4176814417924988e-07,
+      "loss": 0.0002,
+      "step": 14759
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.417437895762299e-07,
+      "loss": 0.0002,
+      "step": 14760
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 1.4171943497320993e-07,
+      "loss": 0.0005,
+      "step": 14761
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4169508037018995e-07,
+      "loss": 0.0002,
+      "step": 14762
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4167072576716998e-07,
+      "loss": 0.0002,
+      "step": 14763
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4164637116415e-07,
+      "loss": 0.0001,
+      "step": 14764
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4162201656113005e-07,
+      "loss": 0.0002,
+      "step": 14765
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4159766195811008e-07,
+      "loss": 0.4975,
+      "step": 14766
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.415733073550901e-07,
+      "loss": 0.0003,
+      "step": 14767
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4154895275207013e-07,
+      "loss": 0.0002,
+      "step": 14768
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4152459814905015e-07,
+      "loss": 0.0002,
+      "step": 14769
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.4150024354603018e-07,
+      "loss": 0.0002,
+      "step": 14770
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.414758889430102e-07,
+      "loss": 0.0002,
+      "step": 14771
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4145153433999023e-07,
+      "loss": 0.0002,
+      "step": 14772
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4142717973697028e-07,
+      "loss": 0.0002,
+      "step": 14773
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.414028251339503e-07,
+      "loss": 0.0003,
+      "step": 14774
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4137847053093033e-07,
+      "loss": 0.0002,
+      "step": 14775
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4135411592791036e-07,
+      "loss": 0.0003,
+      "step": 14776
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4132976132489038e-07,
+      "loss": 0.0004,
+      "step": 14777
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4130540672187043e-07,
+      "loss": 0.0003,
+      "step": 14778
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4128105211885046e-07,
+      "loss": 0.0004,
+      "step": 14779
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.412566975158305e-07,
+      "loss": 0.0002,
+      "step": 14780
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 1.4123234291281054e-07,
+      "loss": 0.0002,
+      "step": 14781
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.4120798830979056e-07,
+      "loss": 0.0004,
+      "step": 14782
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.4118363370677059e-07,
+      "loss": 0.0003,
+      "step": 14783
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.411592791037506e-07,
+      "loss": 0.0003,
+      "step": 14784
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.4113492450073064e-07,
+      "loss": 0.0005,
+      "step": 14785
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.4111056989771066e-07,
+      "loss": 0.0002,
+      "step": 14786
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.4108621529469069e-07,
+      "loss": 0.0002,
+      "step": 14787
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.4106186069167074e-07,
+      "loss": 0.0002,
+      "step": 14788
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.4103750608865076e-07,
+      "loss": 0.0002,
+      "step": 14789
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.410131514856308e-07,
+      "loss": 0.0002,
+      "step": 14790
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 1.409887968826108e-07,
+      "loss": 0.0003,
+      "step": 14791
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.4096444227959084e-07,
+      "loss": 0.0002,
+      "step": 14792
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.4094008767657086e-07,
+      "loss": 0.0002,
+      "step": 14793
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.409157330735509e-07,
+      "loss": 0.0004,
+      "step": 14794
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.4089137847053091e-07,
+      "loss": 0.0004,
+      "step": 14795
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.4086702386751097e-07,
+      "loss": 0.0004,
+      "step": 14796
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.40842669264491e-07,
+      "loss": 0.0004,
+      "step": 14797
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.4081831466147102e-07,
+      "loss": 0.0002,
+      "step": 14798
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.4079396005845104e-07,
+      "loss": 0.0004,
+      "step": 14799
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.4076960545543107e-07,
+      "loss": 0.0002,
+      "step": 14800
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.407452508524111e-07,
+      "loss": 0.0002,
+      "step": 14801
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 1.4072089624939112e-07,
+      "loss": 0.0002,
+      "step": 14802
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4069654164637114e-07,
+      "loss": 0.0003,
+      "step": 14803
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.406721870433512e-07,
+      "loss": 0.0002,
+      "step": 14804
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4064783244033122e-07,
+      "loss": 1.571,
+      "step": 14805
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4062347783731124e-07,
+      "loss": 0.0003,
+      "step": 14806
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4059912323429127e-07,
+      "loss": 0.0003,
+      "step": 14807
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.405747686312713e-07,
+      "loss": 0.0002,
+      "step": 14808
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4055041402825132e-07,
+      "loss": 0.0002,
+      "step": 14809
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4052605942523134e-07,
+      "loss": 0.0002,
+      "step": 14810
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4050170482221137e-07,
+      "loss": 0.0004,
+      "step": 14811
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 1.4047735021919142e-07,
+      "loss": 0.0003,
+      "step": 14812
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.4045299561617145e-07,
+      "loss": 0.0002,
+      "step": 14813
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.4042864101315147e-07,
+      "loss": 0.0002,
+      "step": 14814
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.404042864101315e-07,
+      "loss": 0.0002,
+      "step": 14815
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.4037993180711155e-07,
+      "loss": 0.0002,
+      "step": 14816
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.4035557720409157e-07,
+      "loss": 0.0003,
+      "step": 14817
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.403312226010716e-07,
+      "loss": 0.0002,
+      "step": 14818
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.4030686799805162e-07,
+      "loss": 0.0003,
+      "step": 14819
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.4028251339503168e-07,
+      "loss": 0.0003,
+      "step": 14820
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.402581587920117e-07,
+      "loss": 0.0002,
+      "step": 14821
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.4023380418899173e-07,
+      "loss": 0.0002,
+      "step": 14822
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.4020944958597175e-07,
+      "loss": 0.0002,
+      "step": 14823
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.4018509498295178e-07,
+      "loss": 0.0003,
+      "step": 14824
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.401607403799318e-07,
+      "loss": 0.0002,
+      "step": 14825
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.4013638577691183e-07,
+      "loss": 0.0002,
+      "step": 14826
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.4011203117389188e-07,
+      "loss": 0.0003,
+      "step": 14827
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.400876765708719e-07,
+      "loss": 0.0003,
+      "step": 14828
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.4006332196785193e-07,
+      "loss": 0.0003,
+      "step": 14829
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.4003896736483195e-07,
+      "loss": 0.0002,
+      "step": 14830
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.4001461276181198e-07,
+      "loss": 0.0003,
+      "step": 14831
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.39990258158792e-07,
+      "loss": 0.0004,
+      "step": 14832
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 1.3996590355577203e-07,
+      "loss": 0.0003,
+      "step": 14833
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3994154895275205e-07,
+      "loss": 0.0003,
+      "step": 14834
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.399171943497321e-07,
+      "loss": 0.0002,
+      "step": 14835
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3989283974671213e-07,
+      "loss": 0.0004,
+      "step": 14836
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3986848514369216e-07,
+      "loss": 0.0002,
+      "step": 14837
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3984413054067218e-07,
+      "loss": 0.0003,
+      "step": 14838
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.398197759376522e-07,
+      "loss": 0.0004,
+      "step": 14839
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3979542133463223e-07,
+      "loss": 1.9467,
+      "step": 14840
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3977106673161226e-07,
+      "loss": 0.0003,
+      "step": 14841
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3974671212859228e-07,
+      "loss": 0.0002,
+      "step": 14842
+    },
+    {
+      "epoch": 14.42,
+      "learning_rate": 1.3972235752557233e-07,
+      "loss": 0.0002,
+      "step": 14843
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.3969800292255236e-07,
+      "loss": 0.0001,
+      "step": 14844
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.3967364831953238e-07,
+      "loss": 0.0003,
+      "step": 14845
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.396492937165124e-07,
+      "loss": 0.0003,
+      "step": 14846
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.3962493911349243e-07,
+      "loss": 0.0002,
+      "step": 14847
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.3960058451047246e-07,
+      "loss": 0.0003,
+      "step": 14848
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.3957622990745248e-07,
+      "loss": 0.0003,
+      "step": 14849
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.395518753044325e-07,
+      "loss": 0.0002,
+      "step": 14850
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.3952752070141256e-07,
+      "loss": 0.0002,
+      "step": 14851
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.3950316609839259e-07,
+      "loss": 0.0003,
+      "step": 14852
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 1.394788114953726e-07,
+      "loss": 0.0002,
+      "step": 14853
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3945445689235266e-07,
+      "loss": 0.0003,
+      "step": 14854
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.394301022893327e-07,
+      "loss": 0.0005,
+      "step": 14855
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3940574768631271e-07,
+      "loss": 0.0004,
+      "step": 14856
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3938139308329274e-07,
+      "loss": 0.0002,
+      "step": 14857
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3935703848027276e-07,
+      "loss": 0.0003,
+      "step": 14858
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3933268387725282e-07,
+      "loss": 0.0005,
+      "step": 14859
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3930832927423284e-07,
+      "loss": 0.0002,
+      "step": 14860
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3928397467121287e-07,
+      "loss": 0.0004,
+      "step": 14861
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.392596200681929e-07,
+      "loss": 0.0003,
+      "step": 14862
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 1.3923526546517292e-07,
+      "loss": 0.0003,
+      "step": 14863
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3921091086215294e-07,
+      "loss": 0.0003,
+      "step": 14864
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3918655625913297e-07,
+      "loss": 0.0003,
+      "step": 14865
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.39162201656113e-07,
+      "loss": 0.0003,
+      "step": 14866
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3913784705309304e-07,
+      "loss": 0.0003,
+      "step": 14867
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3911349245007307e-07,
+      "loss": 0.0003,
+      "step": 14868
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.390891378470531e-07,
+      "loss": 0.0001,
+      "step": 14869
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3906478324403312e-07,
+      "loss": 0.0001,
+      "step": 14870
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3904042864101314e-07,
+      "loss": 0.0002,
+      "step": 14871
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3901607403799317e-07,
+      "loss": 0.0002,
+      "step": 14872
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.389917194349732e-07,
+      "loss": 0.0004,
+      "step": 14873
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.3896736483195322e-07,
+      "loss": 0.0004,
+      "step": 14874
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.3894301022893327e-07,
+      "loss": 0.0001,
+      "step": 14875
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.389186556259133e-07,
+      "loss": 0.0001,
+      "step": 14876
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.3889430102289332e-07,
+      "loss": 0.0002,
+      "step": 14877
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.3886994641987335e-07,
+      "loss": 0.0003,
+      "step": 14878
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.3884559181685337e-07,
+      "loss": 0.0003,
+      "step": 14879
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.388212372138334e-07,
+      "loss": 0.0003,
+      "step": 14880
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.3879688261081342e-07,
+      "loss": 0.0002,
+      "step": 14881
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.3877252800779345e-07,
+      "loss": 0.0003,
+      "step": 14882
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.387481734047735e-07,
+      "loss": 0.0003,
+      "step": 14883
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 1.3872381880175352e-07,
+      "loss": 0.0002,
+      "step": 14884
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3869946419873355e-07,
+      "loss": 0.0002,
+      "step": 14885
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3867510959571357e-07,
+      "loss": 0.0003,
+      "step": 14886
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.386507549926936e-07,
+      "loss": 0.0002,
+      "step": 14887
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3862640038967362e-07,
+      "loss": 0.0003,
+      "step": 14888
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3860204578665365e-07,
+      "loss": 0.0003,
+      "step": 14889
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3857769118363368e-07,
+      "loss": 0.0003,
+      "step": 14890
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3855333658061373e-07,
+      "loss": 0.0004,
+      "step": 14891
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3852898197759375e-07,
+      "loss": 0.0002,
+      "step": 14892
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.385046273745738e-07,
+      "loss": 0.0002,
+      "step": 14893
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 1.3848027277155383e-07,
+      "loss": 0.0003,
+      "step": 14894
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.3845591816853385e-07,
+      "loss": 0.0003,
+      "step": 14895
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.3843156356551388e-07,
+      "loss": 0.0002,
+      "step": 14896
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.384072089624939e-07,
+      "loss": 0.0005,
+      "step": 14897
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.3838285435947396e-07,
+      "loss": 0.0003,
+      "step": 14898
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.3835849975645398e-07,
+      "loss": 0.0002,
+      "step": 14899
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.38334145153434e-07,
+      "loss": 0.0002,
+      "step": 14900
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.3830979055041403e-07,
+      "loss": 0.0002,
+      "step": 14901
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.3828543594739406e-07,
+      "loss": 0.0002,
+      "step": 14902
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.3826108134437408e-07,
+      "loss": 0.0002,
+      "step": 14903
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.382367267413541e-07,
+      "loss": 0.0003,
+      "step": 14904
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 1.3821237213833413e-07,
+      "loss": 0.0003,
+      "step": 14905
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.3818801753531418e-07,
+      "loss": 0.0002,
+      "step": 14906
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.381636629322942e-07,
+      "loss": 0.0003,
+      "step": 14907
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.3813930832927423e-07,
+      "loss": 0.0001,
+      "step": 14908
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.3811495372625426e-07,
+      "loss": 0.0005,
+      "step": 14909
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.3809059912323428e-07,
+      "loss": 0.0002,
+      "step": 14910
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.380662445202143e-07,
+      "loss": 0.0002,
+      "step": 14911
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.3804188991719433e-07,
+      "loss": 0.0003,
+      "step": 14912
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.3801753531417436e-07,
+      "loss": 0.0002,
+      "step": 14913
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.379931807111544e-07,
+      "loss": 0.0002,
+      "step": 14914
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 1.3796882610813444e-07,
+      "loss": 0.2571,
+      "step": 14915
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.3794447150511446e-07,
+      "loss": 0.0002,
+      "step": 14916
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.379201169020945e-07,
+      "loss": 0.0003,
+      "step": 14917
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.378957622990745e-07,
+      "loss": 0.0004,
+      "step": 14918
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.3787140769605454e-07,
+      "loss": 0.0002,
+      "step": 14919
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.3784705309303456e-07,
+      "loss": 0.0002,
+      "step": 14920
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.378226984900146e-07,
+      "loss": 0.0002,
+      "step": 14921
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.3779834388699464e-07,
+      "loss": 0.0004,
+      "step": 14922
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.3777398928397466e-07,
+      "loss": 0.0002,
+      "step": 14923
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.377496346809547e-07,
+      "loss": 0.0003,
+      "step": 14924
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 1.3772528007793471e-07,
+      "loss": 0.0003,
+      "step": 14925
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.3770092547491474e-07,
+      "loss": 0.0181,
+      "step": 14926
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.3767657087189477e-07,
+      "loss": 0.0002,
+      "step": 14927
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.376522162688748e-07,
+      "loss": 0.0003,
+      "step": 14928
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.3762786166585482e-07,
+      "loss": 0.0002,
+      "step": 14929
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.3760350706283487e-07,
+      "loss": 0.0002,
+      "step": 14930
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.3757915245981492e-07,
+      "loss": 0.0003,
+      "step": 14931
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.3755479785679494e-07,
+      "loss": 0.0002,
+      "step": 14932
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.3753044325377497e-07,
+      "loss": 0.0002,
+      "step": 14933
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.37506088650755e-07,
+      "loss": 0.0002,
+      "step": 14934
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 1.3748173404773502e-07,
+      "loss": 0.0002,
+      "step": 14935
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3745737944471504e-07,
+      "loss": 0.0003,
+      "step": 14936
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.374330248416951e-07,
+      "loss": 0.0004,
+      "step": 14937
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3740867023867512e-07,
+      "loss": 0.0003,
+      "step": 14938
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3738431563565515e-07,
+      "loss": 0.0003,
+      "step": 14939
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3735996103263517e-07,
+      "loss": 0.0003,
+      "step": 14940
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.373356064296152e-07,
+      "loss": 0.0002,
+      "step": 14941
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3731125182659522e-07,
+      "loss": 0.0003,
+      "step": 14942
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3728689722357525e-07,
+      "loss": 0.0002,
+      "step": 14943
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3726254262055527e-07,
+      "loss": 0.0002,
+      "step": 14944
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3723818801753532e-07,
+      "loss": 0.0004,
+      "step": 14945
+    },
+    {
+      "epoch": 14.52,
+      "learning_rate": 1.3721383341451535e-07,
+      "loss": 0.0002,
+      "step": 14946
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.3718947881149537e-07,
+      "loss": 0.0004,
+      "step": 14947
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.371651242084754e-07,
+      "loss": 0.0002,
+      "step": 14948
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.3714076960545542e-07,
+      "loss": 0.0002,
+      "step": 14949
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.3711641500243545e-07,
+      "loss": 0.0003,
+      "step": 14950
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.3709206039941547e-07,
+      "loss": 0.0002,
+      "step": 14951
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.370677057963955e-07,
+      "loss": 0.0003,
+      "step": 14952
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.3704335119337555e-07,
+      "loss": 0.0002,
+      "step": 14953
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.3701899659035558e-07,
+      "loss": 0.0002,
+      "step": 14954
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.369946419873356e-07,
+      "loss": 0.0002,
+      "step": 14955
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 1.3697028738431563e-07,
+      "loss": 0.0002,
+      "step": 14956
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.3694593278129565e-07,
+      "loss": 0.0003,
+      "step": 14957
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.3692157817827568e-07,
+      "loss": 0.0004,
+      "step": 14958
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.368972235752557e-07,
+      "loss": 0.0003,
+      "step": 14959
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.3687286897223573e-07,
+      "loss": 0.0003,
+      "step": 14960
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.3684851436921578e-07,
+      "loss": 0.0003,
+      "step": 14961
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.368241597661958e-07,
+      "loss": 0.0003,
+      "step": 14962
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.3679980516317583e-07,
+      "loss": 0.0002,
+      "step": 14963
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.3677545056015585e-07,
+      "loss": 0.0002,
+      "step": 14964
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.3675109595713588e-07,
+      "loss": 0.0002,
+      "step": 14965
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 1.367267413541159e-07,
+      "loss": 0.0002,
+      "step": 14966
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3670238675109593e-07,
+      "loss": 0.0004,
+      "step": 14967
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3667803214807596e-07,
+      "loss": 0.0003,
+      "step": 14968
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3665367754505603e-07,
+      "loss": 0.0002,
+      "step": 14969
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3662932294203606e-07,
+      "loss": 0.0002,
+      "step": 14970
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3660496833901608e-07,
+      "loss": 0.0002,
+      "step": 14971
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.365806137359961e-07,
+      "loss": 0.0002,
+      "step": 14972
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3655625913297613e-07,
+      "loss": 0.0002,
+      "step": 14973
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3653190452995616e-07,
+      "loss": 0.0004,
+      "step": 14974
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3650754992693618e-07,
+      "loss": 0.0002,
+      "step": 14975
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.364831953239162e-07,
+      "loss": 0.0003,
+      "step": 14976
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 1.3645884072089626e-07,
+      "loss": 0.0002,
+      "step": 14977
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.3643448611787629e-07,
+      "loss": 0.0001,
+      "step": 14978
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.364101315148563e-07,
+      "loss": 0.0002,
+      "step": 14979
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.3638577691183634e-07,
+      "loss": 0.0001,
+      "step": 14980
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.3636142230881636e-07,
+      "loss": 0.0002,
+      "step": 14981
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.363370677057964e-07,
+      "loss": 0.0002,
+      "step": 14982
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.363127131027764e-07,
+      "loss": 0.0002,
+      "step": 14983
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.3628835849975644e-07,
+      "loss": 0.0004,
+      "step": 14984
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.362640038967365e-07,
+      "loss": 0.0003,
+      "step": 14985
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.3623964929371651e-07,
+      "loss": 0.0003,
+      "step": 14986
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 1.3621529469069654e-07,
+      "loss": 0.0002,
+      "step": 14987
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3619094008767656e-07,
+      "loss": 0.0002,
+      "step": 14988
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.361665854846566e-07,
+      "loss": 0.0002,
+      "step": 14989
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3614223088163662e-07,
+      "loss": 0.0002,
+      "step": 14990
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3611787627861664e-07,
+      "loss": 0.0003,
+      "step": 14991
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3609352167559667e-07,
+      "loss": 0.0007,
+      "step": 14992
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3606916707257672e-07,
+      "loss": 0.0002,
+      "step": 14993
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3604481246955674e-07,
+      "loss": 0.0002,
+      "step": 14994
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3602045786653677e-07,
+      "loss": 0.0003,
+      "step": 14995
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.359961032635168e-07,
+      "loss": 0.0003,
+      "step": 14996
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 1.3597174866049682e-07,
+      "loss": 0.0003,
+      "step": 14997
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.3594739405747684e-07,
+      "loss": 0.0002,
+      "step": 14998
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.3592303945445687e-07,
+      "loss": 0.0003,
+      "step": 14999
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.3589868485143692e-07,
+      "loss": 0.0002,
+      "step": 15000
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.3587433024841694e-07,
+      "loss": 0.0003,
+      "step": 15001
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.3584997564539697e-07,
+      "loss": 0.0002,
+      "step": 15002
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.35825621042377e-07,
+      "loss": 0.0002,
+      "step": 15003
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.3580126643935702e-07,
+      "loss": 0.0002,
+      "step": 15004
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.3577691183633705e-07,
+      "loss": 0.0003,
+      "step": 15005
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.3575255723331707e-07,
+      "loss": 0.0003,
+      "step": 15006
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 1.357282026302971e-07,
+      "loss": 0.0002,
+      "step": 15007
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.3570384802727717e-07,
+      "loss": 0.0004,
+      "step": 15008
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.356794934242572e-07,
+      "loss": 0.0003,
+      "step": 15009
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.3565513882123722e-07,
+      "loss": 0.0004,
+      "step": 15010
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.3563078421821725e-07,
+      "loss": 0.0002,
+      "step": 15011
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.3560642961519727e-07,
+      "loss": 0.0003,
+      "step": 15012
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.355820750121773e-07,
+      "loss": 0.0002,
+      "step": 15013
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.3555772040915732e-07,
+      "loss": 0.0008,
+      "step": 15014
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.3553336580613735e-07,
+      "loss": 0.0001,
+      "step": 15015
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.355090112031174e-07,
+      "loss": 0.0003,
+      "step": 15016
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.3548465660009743e-07,
+      "loss": 0.0002,
+      "step": 15017
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 1.3546030199707745e-07,
+      "loss": 0.0003,
+      "step": 15018
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.3543594739405748e-07,
+      "loss": 0.0002,
+      "step": 15019
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.354115927910375e-07,
+      "loss": 0.0002,
+      "step": 15020
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.3538723818801753e-07,
+      "loss": 0.0004,
+      "step": 15021
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.3536288358499755e-07,
+      "loss": 0.0002,
+      "step": 15022
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.3533852898197758e-07,
+      "loss": 0.0002,
+      "step": 15023
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.3531417437895763e-07,
+      "loss": 0.0004,
+      "step": 15024
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.3528981977593765e-07,
+      "loss": 0.0002,
+      "step": 15025
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.3526546517291768e-07,
+      "loss": 0.0003,
+      "step": 15026
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.352411105698977e-07,
+      "loss": 0.0003,
+      "step": 15027
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 1.3521675596687773e-07,
+      "loss": 0.0003,
+      "step": 15028
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.3519240136385776e-07,
+      "loss": 0.0002,
+      "step": 15029
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.3516804676083778e-07,
+      "loss": 0.0003,
+      "step": 15030
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.351436921578178e-07,
+      "loss": 0.0002,
+      "step": 15031
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.3511933755479786e-07,
+      "loss": 0.0002,
+      "step": 15032
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.3509498295177788e-07,
+      "loss": 0.0002,
+      "step": 15033
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.350706283487579e-07,
+      "loss": 0.0005,
+      "step": 15034
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.3504627374573793e-07,
+      "loss": 0.0002,
+      "step": 15035
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.3502191914271796e-07,
+      "loss": 0.0004,
+      "step": 15036
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.3499756453969798e-07,
+      "loss": 0.0002,
+      "step": 15037
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 1.34973209936678e-07,
+      "loss": 0.0002,
+      "step": 15038
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3494885533365803e-07,
+      "loss": 0.0004,
+      "step": 15039
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3492450073063809e-07,
+      "loss": 0.0002,
+      "step": 15040
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.349001461276181e-07,
+      "loss": 0.0001,
+      "step": 15041
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3487579152459814e-07,
+      "loss": 0.0002,
+      "step": 15042
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3485143692157816e-07,
+      "loss": 0.0002,
+      "step": 15043
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3482708231855819e-07,
+      "loss": 0.0002,
+      "step": 15044
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.348027277155382e-07,
+      "loss": 0.0003,
+      "step": 15045
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3477837311251826e-07,
+      "loss": 0.0003,
+      "step": 15046
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3475401850949831e-07,
+      "loss": 0.0004,
+      "step": 15047
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3472966390647834e-07,
+      "loss": 0.0002,
+      "step": 15048
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 1.3470530930345836e-07,
+      "loss": 0.0003,
+      "step": 15049
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.346809547004384e-07,
+      "loss": 0.0003,
+      "step": 15050
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.3465660009741841e-07,
+      "loss": 0.0003,
+      "step": 15051
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.3463224549439844e-07,
+      "loss": 0.0002,
+      "step": 15052
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.3460789089137847e-07,
+      "loss": 0.0002,
+      "step": 15053
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.345835362883585e-07,
+      "loss": 0.0027,
+      "step": 15054
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.3455918168533854e-07,
+      "loss": 0.0004,
+      "step": 15055
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.3453482708231857e-07,
+      "loss": 0.0002,
+      "step": 15056
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.345104724792986e-07,
+      "loss": 0.0003,
+      "step": 15057
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.3448611787627862e-07,
+      "loss": 0.0002,
+      "step": 15058
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 1.3446176327325864e-07,
+      "loss": 0.0002,
+      "step": 15059
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.3443740867023867e-07,
+      "loss": 0.0003,
+      "step": 15060
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.344130540672187e-07,
+      "loss": 0.0005,
+      "step": 15061
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.3438869946419872e-07,
+      "loss": 1.6396,
+      "step": 15062
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.3436434486117877e-07,
+      "loss": 0.0002,
+      "step": 15063
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.343399902581588e-07,
+      "loss": 0.0001,
+      "step": 15064
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.3431563565513882e-07,
+      "loss": 0.0003,
+      "step": 15065
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.3429128105211885e-07,
+      "loss": 0.0002,
+      "step": 15066
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.3426692644909887e-07,
+      "loss": 0.0002,
+      "step": 15067
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.342425718460789e-07,
+      "loss": 0.0002,
+      "step": 15068
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 1.3421821724305892e-07,
+      "loss": 0.0002,
+      "step": 15069
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.3419386264003895e-07,
+      "loss": 0.0003,
+      "step": 15070
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.34169508037019e-07,
+      "loss": 0.0002,
+      "step": 15071
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.3414515343399902e-07,
+      "loss": 0.0002,
+      "step": 15072
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.3412079883097905e-07,
+      "loss": 0.0003,
+      "step": 15073
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.3409644422795907e-07,
+      "loss": 0.0004,
+      "step": 15074
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.340720896249391e-07,
+      "loss": 0.0002,
+      "step": 15075
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.3404773502191912e-07,
+      "loss": 0.0002,
+      "step": 15076
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.3402338041889915e-07,
+      "loss": 0.0001,
+      "step": 15077
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.3399902581587917e-07,
+      "loss": 0.0004,
+      "step": 15078
+    },
+    {
+      "epoch": 14.65,
+      "learning_rate": 1.3397467121285923e-07,
+      "loss": 0.0002,
+      "step": 15079
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.3395031660983925e-07,
+      "loss": 0.0002,
+      "step": 15080
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.3392596200681928e-07,
+      "loss": 0.0002,
+      "step": 15081
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.339016074037993e-07,
+      "loss": 0.0003,
+      "step": 15082
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.3387725280077933e-07,
+      "loss": 0.0001,
+      "step": 15083
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.3385289819775938e-07,
+      "loss": 0.0003,
+      "step": 15084
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.338285435947394e-07,
+      "loss": 0.0002,
+      "step": 15085
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.3380418899171943e-07,
+      "loss": 0.0004,
+      "step": 15086
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.3377983438869948e-07,
+      "loss": 0.0002,
+      "step": 15087
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.337554797856795e-07,
+      "loss": 0.0002,
+      "step": 15088
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.3373112518265953e-07,
+      "loss": 0.0002,
+      "step": 15089
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 1.3370677057963956e-07,
+      "loss": 0.0003,
+      "step": 15090
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.3368241597661958e-07,
+      "loss": 0.0003,
+      "step": 15091
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.336580613735996e-07,
+      "loss": 0.0003,
+      "step": 15092
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.3363370677057963e-07,
+      "loss": 0.0004,
+      "step": 15093
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.3360935216755968e-07,
+      "loss": 0.0002,
+      "step": 15094
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.335849975645397e-07,
+      "loss": 0.0003,
+      "step": 15095
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.3356064296151973e-07,
+      "loss": 0.0006,
+      "step": 15096
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.3353628835849976e-07,
+      "loss": 0.0004,
+      "step": 15097
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.3351193375547978e-07,
+      "loss": 0.0002,
+      "step": 15098
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.334875791524598e-07,
+      "loss": 0.0002,
+      "step": 15099
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 1.3346322454943983e-07,
+      "loss": 0.0003,
+      "step": 15100
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3343886994641986e-07,
+      "loss": 0.0002,
+      "step": 15101
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.334145153433999e-07,
+      "loss": 0.0002,
+      "step": 15102
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3339016074037994e-07,
+      "loss": 0.0002,
+      "step": 15103
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3336580613735996e-07,
+      "loss": 1.6257,
+      "step": 15104
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3334145153433999e-07,
+      "loss": 0.0003,
+      "step": 15105
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3331709693132e-07,
+      "loss": 0.0002,
+      "step": 15106
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3329274232830004e-07,
+      "loss": 0.0002,
+      "step": 15107
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3326838772528006e-07,
+      "loss": 0.0003,
+      "step": 15108
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3324403312226009e-07,
+      "loss": 0.0004,
+      "step": 15109
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 1.3321967851924014e-07,
+      "loss": 0.0003,
+      "step": 15110
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.3319532391622016e-07,
+      "loss": 0.0004,
+      "step": 15111
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.331709693132002e-07,
+      "loss": 0.0003,
+      "step": 15112
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.3314661471018021e-07,
+      "loss": 0.0003,
+      "step": 15113
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.3312226010716024e-07,
+      "loss": 0.0002,
+      "step": 15114
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.3309790550414026e-07,
+      "loss": 0.0003,
+      "step": 15115
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.330735509011203e-07,
+      "loss": 0.0002,
+      "step": 15116
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.3304919629810031e-07,
+      "loss": 0.0002,
+      "step": 15117
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.3302484169508037e-07,
+      "loss": 1.4769,
+      "step": 15118
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.330004870920604e-07,
+      "loss": 0.0004,
+      "step": 15119
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.3297613248904042e-07,
+      "loss": 0.0002,
+      "step": 15120
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 1.3295177788602044e-07,
+      "loss": 0.0002,
+      "step": 15121
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.329274232830005e-07,
+      "loss": 0.0003,
+      "step": 15122
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3290306867998052e-07,
+      "loss": 0.0003,
+      "step": 15123
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3287871407696054e-07,
+      "loss": 0.0002,
+      "step": 15124
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3285435947394057e-07,
+      "loss": 0.0002,
+      "step": 15125
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3283000487092062e-07,
+      "loss": 0.0004,
+      "step": 15126
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3280565026790065e-07,
+      "loss": 0.0002,
+      "step": 15127
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3278129566488067e-07,
+      "loss": 0.0004,
+      "step": 15128
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.327569410618607e-07,
+      "loss": 0.0003,
+      "step": 15129
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3273258645884072e-07,
+      "loss": 0.0004,
+      "step": 15130
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 1.3270823185582075e-07,
+      "loss": 0.0003,
+      "step": 15131
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.3268387725280077e-07,
+      "loss": 0.0003,
+      "step": 15132
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.326595226497808e-07,
+      "loss": 0.0002,
+      "step": 15133
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.3263516804676085e-07,
+      "loss": 0.0003,
+      "step": 15134
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.3261081344374087e-07,
+      "loss": 0.0001,
+      "step": 15135
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.325864588407209e-07,
+      "loss": 0.0002,
+      "step": 15136
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.3256210423770092e-07,
+      "loss": 0.0005,
+      "step": 15137
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.3253774963468095e-07,
+      "loss": 0.0002,
+      "step": 15138
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.3251339503166097e-07,
+      "loss": 0.0003,
+      "step": 15139
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.32489040428641e-07,
+      "loss": 0.0001,
+      "step": 15140
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 1.3246468582562102e-07,
+      "loss": 0.0003,
+      "step": 15141
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.3244033122260108e-07,
+      "loss": 0.0003,
+      "step": 15142
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.324159766195811e-07,
+      "loss": 0.0002,
+      "step": 15143
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.3239162201656113e-07,
+      "loss": 0.0003,
+      "step": 15144
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.3236726741354115e-07,
+      "loss": 0.0002,
+      "step": 15145
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.3234291281052118e-07,
+      "loss": 0.0005,
+      "step": 15146
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.323185582075012e-07,
+      "loss": 0.0003,
+      "step": 15147
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.3229420360448123e-07,
+      "loss": 0.0003,
+      "step": 15148
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.3226984900146125e-07,
+      "loss": 0.0004,
+      "step": 15149
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.322454943984413e-07,
+      "loss": 0.0002,
+      "step": 15150
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.3222113979542133e-07,
+      "loss": 0.0002,
+      "step": 15151
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 1.3219678519240135e-07,
+      "loss": 0.0002,
+      "step": 15152
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3217243058938138e-07,
+      "loss": 0.0002,
+      "step": 15153
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.321480759863614e-07,
+      "loss": 0.0002,
+      "step": 15154
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3212372138334143e-07,
+      "loss": 0.0003,
+      "step": 15155
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3209936678032145e-07,
+      "loss": 0.0003,
+      "step": 15156
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3207501217730148e-07,
+      "loss": 0.0004,
+      "step": 15157
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3205065757428153e-07,
+      "loss": 0.0003,
+      "step": 15158
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3202630297126156e-07,
+      "loss": 0.0003,
+      "step": 15159
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3200194836824158e-07,
+      "loss": 0.0004,
+      "step": 15160
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3197759376522163e-07,
+      "loss": 0.0002,
+      "step": 15161
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 1.3195323916220166e-07,
+      "loss": 0.0002,
+      "step": 15162
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.3192888455918168e-07,
+      "loss": 0.0003,
+      "step": 15163
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.319045299561617e-07,
+      "loss": 0.0002,
+      "step": 15164
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.3188017535314176e-07,
+      "loss": 0.0002,
+      "step": 15165
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.3185582075012179e-07,
+      "loss": 0.0003,
+      "step": 15166
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.318314661471018e-07,
+      "loss": 0.0002,
+      "step": 15167
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.3180711154408184e-07,
+      "loss": 0.0004,
+      "step": 15168
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.3178275694106186e-07,
+      "loss": 0.0002,
+      "step": 15169
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.3175840233804189e-07,
+      "loss": 0.0002,
+      "step": 15170
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.317340477350219e-07,
+      "loss": 0.0005,
+      "step": 15171
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 1.3170969313200194e-07,
+      "loss": 0.0004,
+      "step": 15172
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.31685338528982e-07,
+      "loss": 0.0002,
+      "step": 15173
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.31660983925962e-07,
+      "loss": 0.0002,
+      "step": 15174
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.3163662932294204e-07,
+      "loss": 0.0004,
+      "step": 15175
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.3161227471992206e-07,
+      "loss": 0.0001,
+      "step": 15176
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.315879201169021e-07,
+      "loss": 0.0002,
+      "step": 15177
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.3156356551388211e-07,
+      "loss": 0.0002,
+      "step": 15178
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.3153921091086214e-07,
+      "loss": 0.0005,
+      "step": 15179
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.3151485630784216e-07,
+      "loss": 0.0003,
+      "step": 15180
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.3149050170482222e-07,
+      "loss": 0.0003,
+      "step": 15181
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 1.3146614710180224e-07,
+      "loss": 0.0001,
+      "step": 15182
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.3144179249878227e-07,
+      "loss": 0.0004,
+      "step": 15183
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.314174378957623e-07,
+      "loss": 0.0002,
+      "step": 15184
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.3139308329274232e-07,
+      "loss": 0.0004,
+      "step": 15185
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.3136872868972234e-07,
+      "loss": 0.0003,
+      "step": 15186
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.3134437408670237e-07,
+      "loss": 0.0002,
+      "step": 15187
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.313200194836824e-07,
+      "loss": 0.0005,
+      "step": 15188
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.3129566488066244e-07,
+      "loss": 0.0003,
+      "step": 15189
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.3127131027764247e-07,
+      "loss": 0.0002,
+      "step": 15190
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.312469556746225e-07,
+      "loss": 0.0002,
+      "step": 15191
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.3122260107160252e-07,
+      "loss": 0.0002,
+      "step": 15192
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 1.3119824646858254e-07,
+      "loss": 0.0002,
+      "step": 15193
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.3117389186556257e-07,
+      "loss": 0.0002,
+      "step": 15194
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.311495372625426e-07,
+      "loss": 0.0003,
+      "step": 15195
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.3112518265952262e-07,
+      "loss": 0.0002,
+      "step": 15196
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.3110082805650267e-07,
+      "loss": 0.0002,
+      "step": 15197
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.310764734534827e-07,
+      "loss": 0.0003,
+      "step": 15198
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.3105211885046275e-07,
+      "loss": 0.0002,
+      "step": 15199
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.3102776424744277e-07,
+      "loss": 0.0002,
+      "step": 15200
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.310034096444228e-07,
+      "loss": 0.0003,
+      "step": 15201
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.3097905504140282e-07,
+      "loss": 0.0003,
+      "step": 15202
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 1.3095470043838285e-07,
+      "loss": 0.0003,
+      "step": 15203
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.309303458353629e-07,
+      "loss": 0.0002,
+      "step": 15204
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3090599123234293e-07,
+      "loss": 0.0004,
+      "step": 15205
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3088163662932295e-07,
+      "loss": 0.0002,
+      "step": 15206
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3085728202630298e-07,
+      "loss": 0.0004,
+      "step": 15207
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.30832927423283e-07,
+      "loss": 0.0002,
+      "step": 15208
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3080857282026303e-07,
+      "loss": 0.0002,
+      "step": 15209
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3078421821724305e-07,
+      "loss": 0.0003,
+      "step": 15210
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3075986361422308e-07,
+      "loss": 0.0002,
+      "step": 15211
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3073550901120313e-07,
+      "loss": 0.0004,
+      "step": 15212
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 1.3071115440818315e-07,
+      "loss": 0.0002,
+      "step": 15213
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.3068679980516318e-07,
+      "loss": 0.0003,
+      "step": 15214
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.306624452021432e-07,
+      "loss": 0.0006,
+      "step": 15215
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.3063809059912323e-07,
+      "loss": 0.0003,
+      "step": 15216
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.3061373599610325e-07,
+      "loss": 0.0002,
+      "step": 15217
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.3058938139308328e-07,
+      "loss": 0.0004,
+      "step": 15218
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.305650267900633e-07,
+      "loss": 0.0004,
+      "step": 15219
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.3054067218704336e-07,
+      "loss": 0.0002,
+      "step": 15220
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.3051631758402338e-07,
+      "loss": 0.0002,
+      "step": 15221
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.304919629810034e-07,
+      "loss": 0.0003,
+      "step": 15222
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.3046760837798343e-07,
+      "loss": 0.0003,
+      "step": 15223
+    },
+    {
+      "epoch": 14.79,
+      "learning_rate": 1.3044325377496346e-07,
+      "loss": 0.0002,
+      "step": 15224
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.3041889917194348e-07,
+      "loss": 0.0002,
+      "step": 15225
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.303945445689235e-07,
+      "loss": 0.0004,
+      "step": 15226
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.3037018996590353e-07,
+      "loss": 0.0004,
+      "step": 15227
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.3034583536288358e-07,
+      "loss": 0.0004,
+      "step": 15228
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.303214807598636e-07,
+      "loss": 0.0002,
+      "step": 15229
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.3029712615684363e-07,
+      "loss": 0.0002,
+      "step": 15230
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.3027277155382366e-07,
+      "loss": 0.0002,
+      "step": 15231
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.3024841695080368e-07,
+      "loss": 0.0006,
+      "step": 15232
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.302240623477837e-07,
+      "loss": 0.0003,
+      "step": 15233
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 1.3019970774476373e-07,
+      "loss": 0.0002,
+      "step": 15234
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.3017535314174376e-07,
+      "loss": 0.0002,
+      "step": 15235
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.301509985387238e-07,
+      "loss": 0.0004,
+      "step": 15236
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.3012664393570386e-07,
+      "loss": 0.0005,
+      "step": 15237
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.301022893326839e-07,
+      "loss": 0.0002,
+      "step": 15238
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.3007793472966391e-07,
+      "loss": 0.0007,
+      "step": 15239
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.3005358012664394e-07,
+      "loss": 0.0002,
+      "step": 15240
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.3002922552362396e-07,
+      "loss": 0.0002,
+      "step": 15241
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.30004870920604e-07,
+      "loss": 0.0003,
+      "step": 15242
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.2998051631758401e-07,
+      "loss": 0.0002,
+      "step": 15243
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 1.2995616171456407e-07,
+      "loss": 0.0002,
+      "step": 15244
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.299318071115441e-07,
+      "loss": 0.0004,
+      "step": 15245
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.2990745250852412e-07,
+      "loss": 0.0002,
+      "step": 15246
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.2988309790550414e-07,
+      "loss": 0.0002,
+      "step": 15247
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.2985874330248417e-07,
+      "loss": 0.0001,
+      "step": 15248
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.298343886994642e-07,
+      "loss": 0.0002,
+      "step": 15249
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.2981003409644422e-07,
+      "loss": 0.0002,
+      "step": 15250
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.2978567949342424e-07,
+      "loss": 0.0002,
+      "step": 15251
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.297613248904043e-07,
+      "loss": 0.0002,
+      "step": 15252
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.2973697028738432e-07,
+      "loss": 0.0002,
+      "step": 15253
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 1.2971261568436434e-07,
+      "loss": 0.0002,
+      "step": 15254
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2968826108134437e-07,
+      "loss": 0.0002,
+      "step": 15255
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.296639064783244e-07,
+      "loss": 0.0004,
+      "step": 15256
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2963955187530442e-07,
+      "loss": 0.0001,
+      "step": 15257
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2961519727228444e-07,
+      "loss": 0.0006,
+      "step": 15258
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2959084266926447e-07,
+      "loss": 0.0004,
+      "step": 15259
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2956648806624452e-07,
+      "loss": 0.0003,
+      "step": 15260
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2954213346322455e-07,
+      "loss": 0.0003,
+      "step": 15261
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2951777886020457e-07,
+      "loss": 0.0003,
+      "step": 15262
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.294934242571846e-07,
+      "loss": 0.0002,
+      "step": 15263
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2946906965416462e-07,
+      "loss": 0.0001,
+      "step": 15264
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 1.2944471505114465e-07,
+      "loss": 0.0003,
+      "step": 15265
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.2942036044812467e-07,
+      "loss": 0.0003,
+      "step": 15266
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.2939600584510472e-07,
+      "loss": 0.0002,
+      "step": 15267
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.2937165124208475e-07,
+      "loss": 0.0002,
+      "step": 15268
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.2934729663906477e-07,
+      "loss": 0.0002,
+      "step": 15269
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.293229420360448e-07,
+      "loss": 0.0002,
+      "step": 15270
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.2929858743302482e-07,
+      "loss": 0.0002,
+      "step": 15271
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.2927423283000485e-07,
+      "loss": 0.0002,
+      "step": 15272
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.2924987822698488e-07,
+      "loss": 0.0003,
+      "step": 15273
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.292255236239649e-07,
+      "loss": 0.0001,
+      "step": 15274
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 1.2920116902094495e-07,
+      "loss": 0.0002,
+      "step": 15275
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.29176814417925e-07,
+      "loss": 0.0002,
+      "step": 15276
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.2915245981490503e-07,
+      "loss": 0.0005,
+      "step": 15277
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.2912810521188505e-07,
+      "loss": 0.0004,
+      "step": 15278
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.2910375060886508e-07,
+      "loss": 0.0002,
+      "step": 15279
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.290793960058451e-07,
+      "loss": 0.0003,
+      "step": 15280
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.2905504140282513e-07,
+      "loss": 0.0004,
+      "step": 15281
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.2903068679980515e-07,
+      "loss": 0.0003,
+      "step": 15282
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.290063321967852e-07,
+      "loss": 0.0003,
+      "step": 15283
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.2898197759376523e-07,
+      "loss": 0.0003,
+      "step": 15284
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 1.2895762299074526e-07,
+      "loss": 0.0002,
+      "step": 15285
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2893326838772528e-07,
+      "loss": 0.0003,
+      "step": 15286
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.289089137847053e-07,
+      "loss": 0.0003,
+      "step": 15287
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2888455918168533e-07,
+      "loss": 0.0002,
+      "step": 15288
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2886020457866536e-07,
+      "loss": 1.6757,
+      "step": 15289
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2883584997564538e-07,
+      "loss": 0.0005,
+      "step": 15290
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2881149537262543e-07,
+      "loss": 0.0002,
+      "step": 15291
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2878714076960546e-07,
+      "loss": 0.0002,
+      "step": 15292
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2876278616658548e-07,
+      "loss": 0.0003,
+      "step": 15293
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.287384315635655e-07,
+      "loss": 0.0003,
+      "step": 15294
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2871407696054553e-07,
+      "loss": 0.0002,
+      "step": 15295
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 1.2868972235752556e-07,
+      "loss": 0.0002,
+      "step": 15296
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.2866536775450558e-07,
+      "loss": 0.0003,
+      "step": 15297
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.286410131514856e-07,
+      "loss": 0.0002,
+      "step": 15298
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.2861665854846566e-07,
+      "loss": 0.0002,
+      "step": 15299
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.285923039454457e-07,
+      "loss": 0.0003,
+      "step": 15300
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.285679493424257e-07,
+      "loss": 0.0002,
+      "step": 15301
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.2854359473940574e-07,
+      "loss": 0.0002,
+      "step": 15302
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.2851924013638576e-07,
+      "loss": 0.0002,
+      "step": 15303
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.284948855333658e-07,
+      "loss": 0.0002,
+      "step": 15304
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.284705309303458e-07,
+      "loss": 0.0006,
+      "step": 15305
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 1.2844617632732584e-07,
+      "loss": 0.0002,
+      "step": 15306
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.284218217243059e-07,
+      "loss": 0.0002,
+      "step": 15307
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.2839746712128591e-07,
+      "loss": 0.0003,
+      "step": 15308
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.2837311251826594e-07,
+      "loss": 0.0003,
+      "step": 15309
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.2834875791524597e-07,
+      "loss": 0.0003,
+      "step": 15310
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.28324403312226e-07,
+      "loss": 0.0004,
+      "step": 15311
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.2830004870920602e-07,
+      "loss": 0.0002,
+      "step": 15312
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.2827569410618604e-07,
+      "loss": 0.0001,
+      "step": 15313
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.2825133950316612e-07,
+      "loss": 0.0002,
+      "step": 15314
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.2822698490014614e-07,
+      "loss": 0.0003,
+      "step": 15315
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 1.2820263029712617e-07,
+      "loss": 0.0002,
+      "step": 15316
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.281782756941062e-07,
+      "loss": 0.0004,
+      "step": 15317
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.2815392109108622e-07,
+      "loss": 0.0002,
+      "step": 15318
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.2812956648806624e-07,
+      "loss": 0.0001,
+      "step": 15319
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.2810521188504627e-07,
+      "loss": 0.0004,
+      "step": 15320
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.280808572820263e-07,
+      "loss": 0.0002,
+      "step": 15321
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.2805650267900635e-07,
+      "loss": 0.0002,
+      "step": 15322
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.2803214807598637e-07,
+      "loss": 0.0004,
+      "step": 15323
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.280077934729664e-07,
+      "loss": 0.0001,
+      "step": 15324
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.2798343886994642e-07,
+      "loss": 0.0004,
+      "step": 15325
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 1.2795908426692645e-07,
+      "loss": 0.0003,
+      "step": 15326
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.2793472966390647e-07,
+      "loss": 0.0002,
+      "step": 15327
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.279103750608865e-07,
+      "loss": 0.0003,
+      "step": 15328
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.2788602045786652e-07,
+      "loss": 0.0004,
+      "step": 15329
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.2786166585484657e-07,
+      "loss": 0.0003,
+      "step": 15330
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.278373112518266e-07,
+      "loss": 0.0002,
+      "step": 15331
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.2781295664880662e-07,
+      "loss": 0.0002,
+      "step": 15332
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.2778860204578665e-07,
+      "loss": 0.0004,
+      "step": 15333
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.2776424744276667e-07,
+      "loss": 0.0004,
+      "step": 15334
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.277398928397467e-07,
+      "loss": 0.0003,
+      "step": 15335
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.2771553823672673e-07,
+      "loss": 0.0003,
+      "step": 15336
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 1.2769118363370675e-07,
+      "loss": 0.0006,
+      "step": 15337
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.276668290306868e-07,
+      "loss": 0.0002,
+      "step": 15338
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.2764247442766683e-07,
+      "loss": 0.0003,
+      "step": 15339
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.2761811982464685e-07,
+      "loss": 0.0003,
+      "step": 15340
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.2759376522162688e-07,
+      "loss": 0.0001,
+      "step": 15341
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.275694106186069e-07,
+      "loss": 0.0002,
+      "step": 15342
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.2754505601558693e-07,
+      "loss": 0.0002,
+      "step": 15343
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.2752070141256695e-07,
+      "loss": 0.0002,
+      "step": 15344
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.2749634680954698e-07,
+      "loss": 0.0002,
+      "step": 15345
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.2747199220652703e-07,
+      "loss": 0.0002,
+      "step": 15346
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 1.2744763760350705e-07,
+      "loss": 0.0003,
+      "step": 15347
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.2742328300048708e-07,
+      "loss": 0.0002,
+      "step": 15348
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.273989283974671e-07,
+      "loss": 0.0003,
+      "step": 15349
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.2737457379444713e-07,
+      "loss": 0.0002,
+      "step": 15350
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.2735021919142716e-07,
+      "loss": 0.0002,
+      "step": 15351
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.273258645884072e-07,
+      "loss": 0.0002,
+      "step": 15352
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.2730150998538723e-07,
+      "loss": 0.0004,
+      "step": 15353
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.2727715538236728e-07,
+      "loss": 0.0004,
+      "step": 15354
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.272528007793473e-07,
+      "loss": 0.0003,
+      "step": 15355
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.2722844617632733e-07,
+      "loss": 0.0002,
+      "step": 15356
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 1.2720409157330736e-07,
+      "loss": 0.0002,
+      "step": 15357
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.2717973697028738e-07,
+      "loss": 0.0002,
+      "step": 15358
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.271553823672674e-07,
+      "loss": 0.0004,
+      "step": 15359
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.2713102776424744e-07,
+      "loss": 0.0001,
+      "step": 15360
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.2710667316122746e-07,
+      "loss": 0.0002,
+      "step": 15361
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.270823185582075e-07,
+      "loss": 0.0002,
+      "step": 15362
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.2705796395518754e-07,
+      "loss": 0.0003,
+      "step": 15363
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.2703360935216756e-07,
+      "loss": 0.0001,
+      "step": 15364
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.270092547491476e-07,
+      "loss": 0.0003,
+      "step": 15365
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.269849001461276e-07,
+      "loss": 0.0003,
+      "step": 15366
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.2696054554310764e-07,
+      "loss": 0.0002,
+      "step": 15367
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 1.2693619094008766e-07,
+      "loss": 0.0004,
+      "step": 15368
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.2691183633706771e-07,
+      "loss": 0.0003,
+      "step": 15369
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.2688748173404774e-07,
+      "loss": 0.0003,
+      "step": 15370
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.2686312713102776e-07,
+      "loss": 0.0005,
+      "step": 15371
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.268387725280078e-07,
+      "loss": 0.0002,
+      "step": 15372
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.2681441792498782e-07,
+      "loss": 0.0002,
+      "step": 15373
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.2679006332196784e-07,
+      "loss": 0.0002,
+      "step": 15374
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.2676570871894787e-07,
+      "loss": 0.0005,
+      "step": 15375
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.267413541159279e-07,
+      "loss": 0.0003,
+      "step": 15376
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.2671699951290794e-07,
+      "loss": 0.0002,
+      "step": 15377
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.2669264490988797e-07,
+      "loss": 0.0002,
+      "step": 15378
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.26668290306868e-07,
+      "loss": 0.0002,
+      "step": 15379
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.2664393570384802e-07,
+      "loss": 0.0002,
+      "step": 15380
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.2661958110082804e-07,
+      "loss": 0.0003,
+      "step": 15381
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.2659522649780807e-07,
+      "loss": 0.0005,
+      "step": 15382
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.265708718947881e-07,
+      "loss": 0.0003,
+      "step": 15383
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.2654651729176812e-07,
+      "loss": 0.0003,
+      "step": 15384
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.2652216268874817e-07,
+      "loss": 0.0004,
+      "step": 15385
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.264978080857282e-07,
+      "loss": 0.0004,
+      "step": 15386
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.2647345348270822e-07,
+      "loss": 0.0003,
+      "step": 15387
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 1.2644909887968825e-07,
+      "loss": 0.0003,
+      "step": 15388
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.2642474427666827e-07,
+      "loss": 0.0002,
+      "step": 15389
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.264003896736483e-07,
+      "loss": 0.0004,
+      "step": 15390
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.2637603507062835e-07,
+      "loss": 0.0002,
+      "step": 15391
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.2635168046760837e-07,
+      "loss": 0.0003,
+      "step": 15392
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.2632732586458842e-07,
+      "loss": 0.0002,
+      "step": 15393
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.2630297126156845e-07,
+      "loss": 0.0003,
+      "step": 15394
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.2627861665854847e-07,
+      "loss": 0.0001,
+      "step": 15395
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.262542620555285e-07,
+      "loss": 0.0003,
+      "step": 15396
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.2622990745250852e-07,
+      "loss": 0.0002,
+      "step": 15397
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 1.2620555284948855e-07,
+      "loss": 0.0002,
+      "step": 15398
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.2618119824646858e-07,
+      "loss": 0.0002,
+      "step": 15399
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.261568436434486e-07,
+      "loss": 0.0003,
+      "step": 15400
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.2613248904042865e-07,
+      "loss": 0.0004,
+      "step": 15401
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.2610813443740868e-07,
+      "loss": 0.0004,
+      "step": 15402
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.260837798343887e-07,
+      "loss": 0.0004,
+      "step": 15403
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.2605942523136873e-07,
+      "loss": 0.0003,
+      "step": 15404
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.2603507062834875e-07,
+      "loss": 0.0002,
+      "step": 15405
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.2601071602532878e-07,
+      "loss": 0.0003,
+      "step": 15406
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.259863614223088e-07,
+      "loss": 0.0004,
+      "step": 15407
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.2596200681928883e-07,
+      "loss": 0.0002,
+      "step": 15408
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 1.2593765221626888e-07,
+      "loss": 1.3821,
+      "step": 15409
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.259132976132489e-07,
+      "loss": 0.0002,
+      "step": 15410
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.2588894301022893e-07,
+      "loss": 0.0003,
+      "step": 15411
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.2586458840720896e-07,
+      "loss": 0.0005,
+      "step": 15412
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.2584023380418898e-07,
+      "loss": 0.0003,
+      "step": 15413
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.25815879201169e-07,
+      "loss": 0.0002,
+      "step": 15414
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.2579152459814903e-07,
+      "loss": 0.0002,
+      "step": 15415
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.2576716999512906e-07,
+      "loss": 0.0003,
+      "step": 15416
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.257428153921091e-07,
+      "loss": 0.0002,
+      "step": 15417
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.2571846078908913e-07,
+      "loss": 0.0004,
+      "step": 15418
+    },
+    {
+      "epoch": 14.98,
+      "learning_rate": 1.2569410618606916e-07,
+      "loss": 0.0002,
+      "step": 15419
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.2566975158304918e-07,
+      "loss": 0.0005,
+      "step": 15420
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.256453969800292e-07,
+      "loss": 0.0004,
+      "step": 15421
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.2562104237700923e-07,
+      "loss": 0.0002,
+      "step": 15422
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.2559668777398926e-07,
+      "loss": 0.0002,
+      "step": 15423
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.2557233317096928e-07,
+      "loss": 0.0002,
+      "step": 15424
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.2554797856794934e-07,
+      "loss": 0.0003,
+      "step": 15425
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.2552362396492936e-07,
+      "loss": 0.0004,
+      "step": 15426
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.2549926936190939e-07,
+      "loss": 0.0003,
+      "step": 15427
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.254749147588894e-07,
+      "loss": 0.0002,
+      "step": 15428
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 1.2545056015586946e-07,
+      "loss": 0.0003,
+      "step": 15429
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.254262055528495e-07,
+      "loss": 0.0003,
+      "step": 15430
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.254018509498295e-07,
+      "loss": 0.0002,
+      "step": 15431
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.2537749634680956e-07,
+      "loss": 0.0002,
+      "step": 15432
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.253531417437896e-07,
+      "loss": 0.0003,
+      "step": 15433
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.2532878714076961e-07,
+      "loss": 0.0003,
+      "step": 15434
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.2530443253774964e-07,
+      "loss": 0.0001,
+      "step": 15435
+    },
+    {
+      "epoch": 15.0,
+      "eval_accuracy": 0.9990277102576568,
+      "eval_loss": 0.014385281130671501,
+      "eval_runtime": 2.8034,
+      "eval_samples_per_second": 733.752,
+      "eval_steps_per_second": 11.771,
+      "step": 15435
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.2528007793472967e-07,
+      "loss": 0.0002,
+      "step": 15436
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.252557233317097e-07,
+      "loss": 0.0002,
+      "step": 15437
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.2523136872868972e-07,
+      "loss": 0.0003,
+      "step": 15438
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.2520701412566974e-07,
+      "loss": 0.0003,
+      "step": 15439
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 1.251826595226498e-07,
+      "loss": 0.0003,
+      "step": 15440
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2515830491962982e-07,
+      "loss": 0.0003,
+      "step": 15441
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2513395031660984e-07,
+      "loss": 0.0001,
+      "step": 15442
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2510959571358987e-07,
+      "loss": 0.0002,
+      "step": 15443
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.250852411105699e-07,
+      "loss": 0.0002,
+      "step": 15444
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2506088650754992e-07,
+      "loss": 0.0002,
+      "step": 15445
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2503653190452994e-07,
+      "loss": 0.0003,
+      "step": 15446
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2501217730150997e-07,
+      "loss": 0.0002,
+      "step": 15447
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2498782269849002e-07,
+      "loss": 0.0003,
+      "step": 15448
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2496346809547005e-07,
+      "loss": 0.0002,
+      "step": 15449
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 1.2493911349245007e-07,
+      "loss": 0.0003,
+      "step": 15450
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.249147588894301e-07,
+      "loss": 0.0001,
+      "step": 15451
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.2489040428641012e-07,
+      "loss": 0.0005,
+      "step": 15452
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.2486604968339015e-07,
+      "loss": 0.0002,
+      "step": 15453
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.2484169508037017e-07,
+      "loss": 0.0002,
+      "step": 15454
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.248173404773502e-07,
+      "loss": 0.0002,
+      "step": 15455
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.2479298587433025e-07,
+      "loss": 0.0003,
+      "step": 15456
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.2476863127131027e-07,
+      "loss": 0.0003,
+      "step": 15457
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.247442766682903e-07,
+      "loss": 0.0002,
+      "step": 15458
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.2471992206527032e-07,
+      "loss": 0.0003,
+      "step": 15459
+    },
+    {
+      "epoch": 15.02,
+      "learning_rate": 1.2469556746225038e-07,
+      "loss": 0.0003,
+      "step": 15460
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.246712128592304e-07,
+      "loss": 0.0005,
+      "step": 15461
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.2464685825621043e-07,
+      "loss": 0.0003,
+      "step": 15462
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.2462250365319045e-07,
+      "loss": 0.0002,
+      "step": 15463
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.2459814905017048e-07,
+      "loss": 0.0005,
+      "step": 15464
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.245737944471505e-07,
+      "loss": 0.0001,
+      "step": 15465
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.2454943984413053e-07,
+      "loss": 0.0004,
+      "step": 15466
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.2452508524111055e-07,
+      "loss": 0.0001,
+      "step": 15467
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.245007306380906e-07,
+      "loss": 0.0002,
+      "step": 15468
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.2447637603507063e-07,
+      "loss": 0.0004,
+      "step": 15469
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.2445202143205065e-07,
+      "loss": 0.0002,
+      "step": 15470
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 1.2442766682903068e-07,
+      "loss": 0.0002,
+      "step": 15471
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.244033122260107e-07,
+      "loss": 0.0003,
+      "step": 15472
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2437895762299073e-07,
+      "loss": 0.0003,
+      "step": 15473
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2435460301997075e-07,
+      "loss": 0.0001,
+      "step": 15474
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2433024841695078e-07,
+      "loss": 0.0003,
+      "step": 15475
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2430589381393083e-07,
+      "loss": 0.0003,
+      "step": 15476
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2428153921091086e-07,
+      "loss": 0.0002,
+      "step": 15477
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2425718460789088e-07,
+      "loss": 0.0002,
+      "step": 15478
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2423283000487093e-07,
+      "loss": 0.0002,
+      "step": 15479
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2420847540185096e-07,
+      "loss": 0.0003,
+      "step": 15480
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 1.2418412079883098e-07,
+      "loss": 0.0002,
+      "step": 15481
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.24159766195811e-07,
+      "loss": 0.0001,
+      "step": 15482
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.2413541159279103e-07,
+      "loss": 0.0002,
+      "step": 15483
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.2411105698977106e-07,
+      "loss": 0.0002,
+      "step": 15484
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.2408670238675108e-07,
+      "loss": 0.0002,
+      "step": 15485
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.240623477837311e-07,
+      "loss": 0.0002,
+      "step": 15486
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.2403799318071116e-07,
+      "loss": 0.0001,
+      "step": 15487
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.2401363857769119e-07,
+      "loss": 0.0007,
+      "step": 15488
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.239892839746712e-07,
+      "loss": 0.0002,
+      "step": 15489
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.2396492937165124e-07,
+      "loss": 0.0003,
+      "step": 15490
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 1.2394057476863126e-07,
+      "loss": 0.0003,
+      "step": 15491
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.2391622016561129e-07,
+      "loss": 0.0003,
+      "step": 15492
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.238918655625913e-07,
+      "loss": 0.0003,
+      "step": 15493
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.2386751095957134e-07,
+      "loss": 0.0002,
+      "step": 15494
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.238431563565514e-07,
+      "loss": 0.0003,
+      "step": 15495
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.2381880175353141e-07,
+      "loss": 0.0001,
+      "step": 15496
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.2379444715051144e-07,
+      "loss": 0.0003,
+      "step": 15497
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.2377009254749146e-07,
+      "loss": 0.0002,
+      "step": 15498
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.2374573794447152e-07,
+      "loss": 0.0003,
+      "step": 15499
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.2372138334145154e-07,
+      "loss": 0.0004,
+      "step": 15500
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 1.2369702873843157e-07,
+      "loss": 0.0002,
+      "step": 15501
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.236726741354116e-07,
+      "loss": 0.0002,
+      "step": 15502
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.2364831953239162e-07,
+      "loss": 0.0002,
+      "step": 15503
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.2362396492937164e-07,
+      "loss": 0.0003,
+      "step": 15504
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.2359961032635167e-07,
+      "loss": 0.0003,
+      "step": 15505
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.235752557233317e-07,
+      "loss": 0.0002,
+      "step": 15506
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.2355090112031174e-07,
+      "loss": 0.0003,
+      "step": 15507
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.2352654651729177e-07,
+      "loss": 0.0003,
+      "step": 15508
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.235021919142718e-07,
+      "loss": 0.0005,
+      "step": 15509
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.2347783731125182e-07,
+      "loss": 0.0004,
+      "step": 15510
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.2345348270823184e-07,
+      "loss": 0.0002,
+      "step": 15511
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 1.2342912810521187e-07,
+      "loss": 0.0003,
+      "step": 15512
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.234047735021919e-07,
+      "loss": 0.0003,
+      "step": 15513
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.2338041889917192e-07,
+      "loss": 0.0002,
+      "step": 15514
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.2335606429615197e-07,
+      "loss": 0.0003,
+      "step": 15515
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.23331709693132e-07,
+      "loss": 0.0002,
+      "step": 15516
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.2330735509011202e-07,
+      "loss": 0.0002,
+      "step": 15517
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.2328300048709205e-07,
+      "loss": 0.0005,
+      "step": 15518
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.232586458840721e-07,
+      "loss": 0.0002,
+      "step": 15519
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.2323429128105212e-07,
+      "loss": 0.0002,
+      "step": 15520
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.2320993667803215e-07,
+      "loss": 0.0002,
+      "step": 15521
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 1.2318558207501217e-07,
+      "loss": 0.0002,
+      "step": 15522
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.231612274719922e-07,
+      "loss": 0.0001,
+      "step": 15523
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.2313687286897222e-07,
+      "loss": 0.0002,
+      "step": 15524
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.2311251826595225e-07,
+      "loss": 0.0002,
+      "step": 15525
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.2308816366293227e-07,
+      "loss": 0.0003,
+      "step": 15526
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.2306380905991233e-07,
+      "loss": 0.0002,
+      "step": 15527
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.2303945445689235e-07,
+      "loss": 0.0003,
+      "step": 15528
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.2301509985387238e-07,
+      "loss": 0.0003,
+      "step": 15529
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.229907452508524e-07,
+      "loss": 0.0001,
+      "step": 15530
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.2296639064783243e-07,
+      "loss": 0.0003,
+      "step": 15531
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 1.2294203604481245e-07,
+      "loss": 0.0003,
+      "step": 15532
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2291768144179248e-07,
+      "loss": 0.0002,
+      "step": 15533
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2289332683877253e-07,
+      "loss": 0.0005,
+      "step": 15534
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2286897223575255e-07,
+      "loss": 0.0002,
+      "step": 15535
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2284461763273258e-07,
+      "loss": 0.0002,
+      "step": 15536
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.228202630297126e-07,
+      "loss": 0.0003,
+      "step": 15537
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2279590842669266e-07,
+      "loss": 0.0003,
+      "step": 15538
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2277155382367268e-07,
+      "loss": 0.0002,
+      "step": 15539
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.227471992206527e-07,
+      "loss": 0.0004,
+      "step": 15540
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2272284461763273e-07,
+      "loss": 0.0003,
+      "step": 15541
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2269849001461276e-07,
+      "loss": 0.0004,
+      "step": 15542
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 1.2267413541159278e-07,
+      "loss": 0.0002,
+      "step": 15543
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.226497808085728e-07,
+      "loss": 0.0003,
+      "step": 15544
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.2262542620555283e-07,
+      "loss": 0.0003,
+      "step": 15545
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.2260107160253288e-07,
+      "loss": 0.0002,
+      "step": 15546
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.225767169995129e-07,
+      "loss": 0.0003,
+      "step": 15547
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.2255236239649293e-07,
+      "loss": 0.0007,
+      "step": 15548
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.2252800779347296e-07,
+      "loss": 0.0003,
+      "step": 15549
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.2250365319045298e-07,
+      "loss": 0.0003,
+      "step": 15550
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.22479298587433e-07,
+      "loss": 0.0002,
+      "step": 15551
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.2245494398441303e-07,
+      "loss": 0.0004,
+      "step": 15552
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 1.2243058938139309e-07,
+      "loss": 0.0002,
+      "step": 15553
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.224062347783731e-07,
+      "loss": 0.0002,
+      "step": 15554
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.2238188017535314e-07,
+      "loss": 0.0003,
+      "step": 15555
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.2235752557233316e-07,
+      "loss": 0.0002,
+      "step": 15556
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.2233317096931319e-07,
+      "loss": 0.0002,
+      "step": 15557
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.2230881636629324e-07,
+      "loss": 0.0002,
+      "step": 15558
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.2228446176327326e-07,
+      "loss": 0.0002,
+      "step": 15559
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.222601071602533e-07,
+      "loss": 0.0003,
+      "step": 15560
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.2223575255723331e-07,
+      "loss": 0.0002,
+      "step": 15561
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.2221139795421334e-07,
+      "loss": 0.0002,
+      "step": 15562
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 1.2218704335119336e-07,
+      "loss": 0.0006,
+      "step": 15563
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.221626887481734e-07,
+      "loss": 0.0004,
+      "step": 15564
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.2213833414515341e-07,
+      "loss": 0.0003,
+      "step": 15565
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.2211397954213347e-07,
+      "loss": 0.0001,
+      "step": 15566
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.220896249391135e-07,
+      "loss": 0.0004,
+      "step": 15567
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.2206527033609352e-07,
+      "loss": 0.0003,
+      "step": 15568
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.2204091573307354e-07,
+      "loss": 0.0004,
+      "step": 15569
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.2201656113005357e-07,
+      "loss": 0.0004,
+      "step": 15570
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.219922065270336e-07,
+      "loss": 0.0002,
+      "step": 15571
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.2196785192401364e-07,
+      "loss": 0.0002,
+      "step": 15572
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 1.2194349732099367e-07,
+      "loss": 0.0002,
+      "step": 15573
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.219191427179737e-07,
+      "loss": 0.0004,
+      "step": 15574
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2189478811495372e-07,
+      "loss": 0.0005,
+      "step": 15575
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2187043351193374e-07,
+      "loss": 0.0002,
+      "step": 15576
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2184607890891377e-07,
+      "loss": 0.0002,
+      "step": 15577
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2182172430589382e-07,
+      "loss": 0.0003,
+      "step": 15578
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2179736970287385e-07,
+      "loss": 0.0001,
+      "step": 15579
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2177301509985387e-07,
+      "loss": 0.0003,
+      "step": 15580
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.217486604968339e-07,
+      "loss": 0.0002,
+      "step": 15581
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2172430589381392e-07,
+      "loss": 0.0001,
+      "step": 15582
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2169995129079395e-07,
+      "loss": 0.0003,
+      "step": 15583
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 1.2167559668777397e-07,
+      "loss": 0.0003,
+      "step": 15584
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2165124208475402e-07,
+      "loss": 0.0002,
+      "step": 15585
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2162688748173405e-07,
+      "loss": 0.0002,
+      "step": 15586
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2160253287871407e-07,
+      "loss": 0.0002,
+      "step": 15587
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.215781782756941e-07,
+      "loss": 0.0002,
+      "step": 15588
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2155382367267412e-07,
+      "loss": 0.0002,
+      "step": 15589
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2152946906965415e-07,
+      "loss": 0.0003,
+      "step": 15590
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2150511446663417e-07,
+      "loss": 0.0003,
+      "step": 15591
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2148075986361423e-07,
+      "loss": 0.0004,
+      "step": 15592
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2145640526059425e-07,
+      "loss": 0.0002,
+      "step": 15593
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 1.2143205065757428e-07,
+      "loss": 0.0002,
+      "step": 15594
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.214076960545543e-07,
+      "loss": 0.0003,
+      "step": 15595
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.2138334145153433e-07,
+      "loss": 0.0002,
+      "step": 15596
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.2135898684851438e-07,
+      "loss": 0.0003,
+      "step": 15597
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.213346322454944e-07,
+      "loss": 0.0004,
+      "step": 15598
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.2131027764247443e-07,
+      "loss": 0.0002,
+      "step": 15599
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.2128592303945445e-07,
+      "loss": 0.0003,
+      "step": 15600
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.2126156843643448e-07,
+      "loss": 0.0003,
+      "step": 15601
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.212372138334145e-07,
+      "loss": 0.0003,
+      "step": 15602
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.2121285923039453e-07,
+      "loss": 0.0001,
+      "step": 15603
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 1.2118850462737455e-07,
+      "loss": 0.0002,
+      "step": 15604
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.211641500243546e-07,
+      "loss": 0.0002,
+      "step": 15605
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.2113979542133463e-07,
+      "loss": 0.0004,
+      "step": 15606
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.2111544081831466e-07,
+      "loss": 0.0003,
+      "step": 15607
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.2109108621529468e-07,
+      "loss": 0.0003,
+      "step": 15608
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.210667316122747e-07,
+      "loss": 0.0002,
+      "step": 15609
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.2104237700925473e-07,
+      "loss": 0.0004,
+      "step": 15610
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.2101802240623478e-07,
+      "loss": 0.0004,
+      "step": 15611
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.209936678032148e-07,
+      "loss": 0.0002,
+      "step": 15612
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.2096931320019483e-07,
+      "loss": 0.0002,
+      "step": 15613
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.2094495859717486e-07,
+      "loss": 0.0003,
+      "step": 15614
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 1.2092060399415488e-07,
+      "loss": 0.0005,
+      "step": 15615
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.208962493911349e-07,
+      "loss": 0.0002,
+      "step": 15616
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.2087189478811496e-07,
+      "loss": 0.0002,
+      "step": 15617
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.2084754018509499e-07,
+      "loss": 0.0002,
+      "step": 15618
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.20823185582075e-07,
+      "loss": 0.0003,
+      "step": 15619
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.2079883097905504e-07,
+      "loss": 0.0002,
+      "step": 15620
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.2077447637603506e-07,
+      "loss": 0.0002,
+      "step": 15621
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.207501217730151e-07,
+      "loss": 0.0001,
+      "step": 15622
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.207257671699951e-07,
+      "loss": 0.0003,
+      "step": 15623
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.2070141256697514e-07,
+      "loss": 0.0003,
+      "step": 15624
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 1.206770579639552e-07,
+      "loss": 0.0002,
+      "step": 15625
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.2065270336093521e-07,
+      "loss": 0.0002,
+      "step": 15626
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.2062834875791524e-07,
+      "loss": 0.0004,
+      "step": 15627
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.2060399415489526e-07,
+      "loss": 0.0002,
+      "step": 15628
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.205796395518753e-07,
+      "loss": 0.0004,
+      "step": 15629
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.2055528494885534e-07,
+      "loss": 0.0002,
+      "step": 15630
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.2053093034583537e-07,
+      "loss": 0.0003,
+      "step": 15631
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.205065757428154e-07,
+      "loss": 0.0003,
+      "step": 15632
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.2048222113979542e-07,
+      "loss": 0.0002,
+      "step": 15633
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.2045786653677544e-07,
+      "loss": 0.0004,
+      "step": 15634
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 1.2043351193375547e-07,
+      "loss": 0.0002,
+      "step": 15635
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2040915733073552e-07,
+      "loss": 0.0003,
+      "step": 15636
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2038480272771554e-07,
+      "loss": 0.0003,
+      "step": 15637
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2036044812469557e-07,
+      "loss": 0.0001,
+      "step": 15638
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.203360935216756e-07,
+      "loss": 0.0003,
+      "step": 15639
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2031173891865562e-07,
+      "loss": 0.0003,
+      "step": 15640
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2028738431563564e-07,
+      "loss": 0.0004,
+      "step": 15641
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2026302971261567e-07,
+      "loss": 0.0001,
+      "step": 15642
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.202386751095957e-07,
+      "loss": 0.0004,
+      "step": 15643
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2021432050657575e-07,
+      "loss": 0.0006,
+      "step": 15644
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 1.2018996590355577e-07,
+      "loss": 0.0002,
+      "step": 15645
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.201656113005358e-07,
+      "loss": 0.0001,
+      "step": 15646
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.2014125669751582e-07,
+      "loss": 0.0002,
+      "step": 15647
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.2011690209449585e-07,
+      "loss": 0.0003,
+      "step": 15648
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.200925474914759e-07,
+      "loss": 0.0003,
+      "step": 15649
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.2006819288845592e-07,
+      "loss": 0.0003,
+      "step": 15650
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.2004383828543595e-07,
+      "loss": 0.0004,
+      "step": 15651
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.2001948368241597e-07,
+      "loss": 0.0003,
+      "step": 15652
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.19995129079396e-07,
+      "loss": 0.0004,
+      "step": 15653
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.1997077447637602e-07,
+      "loss": 0.0003,
+      "step": 15654
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.1994641987335605e-07,
+      "loss": 0.0002,
+      "step": 15655
+    },
+    {
+      "epoch": 15.21,
+      "learning_rate": 1.199220652703361e-07,
+      "loss": 0.0002,
+      "step": 15656
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1989771066731613e-07,
+      "loss": 0.0005,
+      "step": 15657
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1987335606429615e-07,
+      "loss": 0.0003,
+      "step": 15658
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1984900146127618e-07,
+      "loss": 0.3716,
+      "step": 15659
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.198246468582562e-07,
+      "loss": 0.0002,
+      "step": 15660
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1980029225523623e-07,
+      "loss": 0.0003,
+      "step": 15661
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1977593765221625e-07,
+      "loss": 0.0002,
+      "step": 15662
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1975158304919628e-07,
+      "loss": 0.0001,
+      "step": 15663
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1972722844617633e-07,
+      "loss": 0.0002,
+      "step": 15664
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1970287384315635e-07,
+      "loss": 0.0002,
+      "step": 15665
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 1.1967851924013638e-07,
+      "loss": 0.0002,
+      "step": 15666
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.196541646371164e-07,
+      "loss": 0.0003,
+      "step": 15667
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.1962981003409646e-07,
+      "loss": 0.0003,
+      "step": 15668
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.1960545543107648e-07,
+      "loss": 0.0002,
+      "step": 15669
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.195811008280565e-07,
+      "loss": 0.0002,
+      "step": 15670
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.1955674622503653e-07,
+      "loss": 0.0004,
+      "step": 15671
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.1953239162201656e-07,
+      "loss": 0.0004,
+      "step": 15672
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.1950803701899658e-07,
+      "loss": 0.0001,
+      "step": 15673
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.194836824159766e-07,
+      "loss": 0.0002,
+      "step": 15674
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.1945932781295663e-07,
+      "loss": 0.0002,
+      "step": 15675
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 1.1943497320993668e-07,
+      "loss": 0.0003,
+      "step": 15676
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.194106186069167e-07,
+      "loss": 0.0001,
+      "step": 15677
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.1938626400389673e-07,
+      "loss": 0.0002,
+      "step": 15678
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.1936190940087676e-07,
+      "loss": 0.0009,
+      "step": 15679
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.1933755479785678e-07,
+      "loss": 0.0002,
+      "step": 15680
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.193132001948368e-07,
+      "loss": 0.0002,
+      "step": 15681
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.1928884559181684e-07,
+      "loss": 0.0002,
+      "step": 15682
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.1926449098879686e-07,
+      "loss": 0.0003,
+      "step": 15683
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.192401363857769e-07,
+      "loss": 0.0002,
+      "step": 15684
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.1921578178275694e-07,
+      "loss": 0.0002,
+      "step": 15685
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.1919142717973696e-07,
+      "loss": 0.0002,
+      "step": 15686
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 1.19167072576717e-07,
+      "loss": 0.0001,
+      "step": 15687
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1914271797369703e-07,
+      "loss": 0.0002,
+      "step": 15688
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1911836337067706e-07,
+      "loss": 0.0003,
+      "step": 15689
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1909400876765709e-07,
+      "loss": 0.0002,
+      "step": 15690
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1906965416463711e-07,
+      "loss": 0.0002,
+      "step": 15691
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1904529956161714e-07,
+      "loss": 0.0003,
+      "step": 15692
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1902094495859718e-07,
+      "loss": 0.0003,
+      "step": 15693
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.189965903555772e-07,
+      "loss": 0.0001,
+      "step": 15694
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1897223575255723e-07,
+      "loss": 0.0002,
+      "step": 15695
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1894788114953725e-07,
+      "loss": 0.0003,
+      "step": 15696
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 1.1892352654651729e-07,
+      "loss": 0.0002,
+      "step": 15697
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1889917194349732e-07,
+      "loss": 0.0002,
+      "step": 15698
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1887481734047734e-07,
+      "loss": 0.0003,
+      "step": 15699
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1885046273745737e-07,
+      "loss": 0.0003,
+      "step": 15700
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.188261081344374e-07,
+      "loss": 0.0003,
+      "step": 15701
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1880175353141743e-07,
+      "loss": 0.0004,
+      "step": 15702
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1877739892839746e-07,
+      "loss": 0.0003,
+      "step": 15703
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1875304432537748e-07,
+      "loss": 0.0004,
+      "step": 15704
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1872868972235752e-07,
+      "loss": 0.0004,
+      "step": 15705
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1870433511933755e-07,
+      "loss": 0.0004,
+      "step": 15706
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 1.1867998051631758e-07,
+      "loss": 0.0002,
+      "step": 15707
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1865562591329761e-07,
+      "loss": 0.0002,
+      "step": 15708
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1863127131027765e-07,
+      "loss": 0.0004,
+      "step": 15709
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1860691670725767e-07,
+      "loss": 0.0003,
+      "step": 15710
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.185825621042377e-07,
+      "loss": 0.0002,
+      "step": 15711
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1855820750121772e-07,
+      "loss": 0.0002,
+      "step": 15712
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1853385289819776e-07,
+      "loss": 0.0003,
+      "step": 15713
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1850949829517779e-07,
+      "loss": 0.0001,
+      "step": 15714
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1848514369215781e-07,
+      "loss": 0.0002,
+      "step": 15715
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1846078908913784e-07,
+      "loss": 0.0004,
+      "step": 15716
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 1.1843643448611787e-07,
+      "loss": 0.0002,
+      "step": 15717
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.184120798830979e-07,
+      "loss": 0.0003,
+      "step": 15718
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1838772528007793e-07,
+      "loss": 0.0002,
+      "step": 15719
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1836337067705795e-07,
+      "loss": 0.0003,
+      "step": 15720
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1833901607403799e-07,
+      "loss": 0.0002,
+      "step": 15721
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1831466147101801e-07,
+      "loss": 0.0003,
+      "step": 15722
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1829030686799804e-07,
+      "loss": 0.0004,
+      "step": 15723
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1826595226497806e-07,
+      "loss": 0.0003,
+      "step": 15724
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.182415976619581e-07,
+      "loss": 0.0003,
+      "step": 15725
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1821724305893814e-07,
+      "loss": 0.0003,
+      "step": 15726
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1819288845591817e-07,
+      "loss": 0.0002,
+      "step": 15727
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 1.1816853385289819e-07,
+      "loss": 0.0004,
+      "step": 15728
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.1814417924987823e-07,
+      "loss": 0.0002,
+      "step": 15729
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.1811982464685825e-07,
+      "loss": 0.0002,
+      "step": 15730
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.1809547004383828e-07,
+      "loss": 0.0002,
+      "step": 15731
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.180711154408183e-07,
+      "loss": 0.0003,
+      "step": 15732
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.1804676083779834e-07,
+      "loss": 0.0002,
+      "step": 15733
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.1802240623477837e-07,
+      "loss": 0.0002,
+      "step": 15734
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.179980516317584e-07,
+      "loss": 0.0002,
+      "step": 15735
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.1797369702873842e-07,
+      "loss": 0.0003,
+      "step": 15736
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.1794934242571846e-07,
+      "loss": 0.0002,
+      "step": 15737
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 1.1792498782269848e-07,
+      "loss": 0.0001,
+      "step": 15738
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.1790063321967851e-07,
+      "loss": 0.0003,
+      "step": 15739
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.1787627861665853e-07,
+      "loss": 0.0002,
+      "step": 15740
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.1785192401363857e-07,
+      "loss": 0.0002,
+      "step": 15741
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.178275694106186e-07,
+      "loss": 0.0002,
+      "step": 15742
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.1780321480759862e-07,
+      "loss": 0.0003,
+      "step": 15743
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.1777886020457865e-07,
+      "loss": 0.0004,
+      "step": 15744
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.177545056015587e-07,
+      "loss": 0.0002,
+      "step": 15745
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.1773015099853872e-07,
+      "loss": 0.0003,
+      "step": 15746
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.1770579639551875e-07,
+      "loss": 0.0002,
+      "step": 15747
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 1.1768144179249879e-07,
+      "loss": 0.0003,
+      "step": 15748
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1765708718947881e-07,
+      "loss": 0.0004,
+      "step": 15749
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1763273258645884e-07,
+      "loss": 0.0001,
+      "step": 15750
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1760837798343886e-07,
+      "loss": 0.0002,
+      "step": 15751
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.175840233804189e-07,
+      "loss": 0.0001,
+      "step": 15752
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1755966877739893e-07,
+      "loss": 0.0002,
+      "step": 15753
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1753531417437895e-07,
+      "loss": 0.0003,
+      "step": 15754
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1751095957135898e-07,
+      "loss": 0.0003,
+      "step": 15755
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1748660496833902e-07,
+      "loss": 0.0002,
+      "step": 15756
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1746225036531904e-07,
+      "loss": 0.0002,
+      "step": 15757
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1743789576229907e-07,
+      "loss": 0.0002,
+      "step": 15758
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 1.1741354115927909e-07,
+      "loss": 0.0001,
+      "step": 15759
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.1738918655625913e-07,
+      "loss": 0.0002,
+      "step": 15760
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.1736483195323915e-07,
+      "loss": 0.0002,
+      "step": 15761
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.1734047735021918e-07,
+      "loss": 0.0004,
+      "step": 15762
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.173161227471992e-07,
+      "loss": 0.0004,
+      "step": 15763
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.1729176814417926e-07,
+      "loss": 0.0002,
+      "step": 15764
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.1726741354115928e-07,
+      "loss": 0.0002,
+      "step": 15765
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.172430589381393e-07,
+      "loss": 0.0002,
+      "step": 15766
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.1721870433511933e-07,
+      "loss": 0.0002,
+      "step": 15767
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.1719434973209937e-07,
+      "loss": 0.0004,
+      "step": 15768
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 1.171699951290794e-07,
+      "loss": 0.0002,
+      "step": 15769
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1714564052605942e-07,
+      "loss": 0.0002,
+      "step": 15770
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1712128592303945e-07,
+      "loss": 0.0004,
+      "step": 15771
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1709693132001948e-07,
+      "loss": 0.0003,
+      "step": 15772
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1707257671699951e-07,
+      "loss": 0.0004,
+      "step": 15773
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1704822211397953e-07,
+      "loss": 0.0003,
+      "step": 15774
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1702386751095956e-07,
+      "loss": 0.0002,
+      "step": 15775
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.169995129079396e-07,
+      "loss": 0.0002,
+      "step": 15776
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1697515830491962e-07,
+      "loss": 0.0002,
+      "step": 15777
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1695080370189965e-07,
+      "loss": 0.0002,
+      "step": 15778
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 1.1692644909887967e-07,
+      "loss": 0.0002,
+      "step": 15779
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1690209449585971e-07,
+      "loss": 0.0004,
+      "step": 15780
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1687773989283974e-07,
+      "loss": 0.0002,
+      "step": 15781
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1685338528981976e-07,
+      "loss": 0.0002,
+      "step": 15782
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.168290306867998e-07,
+      "loss": 0.0002,
+      "step": 15783
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1680467608377984e-07,
+      "loss": 0.0004,
+      "step": 15784
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1678032148075986e-07,
+      "loss": 0.0002,
+      "step": 15785
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1675596687773989e-07,
+      "loss": 0.0002,
+      "step": 15786
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1673161227471991e-07,
+      "loss": 0.0003,
+      "step": 15787
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1670725767169995e-07,
+      "loss": 0.0002,
+      "step": 15788
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1668290306867998e-07,
+      "loss": 0.0002,
+      "step": 15789
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 1.1665854846566e-07,
+      "loss": 0.0004,
+      "step": 15790
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.1663419386264003e-07,
+      "loss": 0.0003,
+      "step": 15791
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.1660983925962007e-07,
+      "loss": 0.0003,
+      "step": 15792
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.1658548465660009e-07,
+      "loss": 0.0002,
+      "step": 15793
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.1656113005358012e-07,
+      "loss": 0.0002,
+      "step": 15794
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.1653677545056014e-07,
+      "loss": 0.0003,
+      "step": 15795
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.1651242084754018e-07,
+      "loss": 0.0001,
+      "step": 15796
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.164880662445202e-07,
+      "loss": 0.0002,
+      "step": 15797
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.1646371164150023e-07,
+      "loss": 0.0002,
+      "step": 15798
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.1643935703848026e-07,
+      "loss": 0.0003,
+      "step": 15799
+    },
+    {
+      "epoch": 15.35,
+      "learning_rate": 1.164150024354603e-07,
+      "loss": 0.0001,
+      "step": 15800
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1639064783244032e-07,
+      "loss": 0.0004,
+      "step": 15801
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1636629322942036e-07,
+      "loss": 0.0002,
+      "step": 15802
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.163419386264004e-07,
+      "loss": 1.781,
+      "step": 15803
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1631758402338042e-07,
+      "loss": 0.0002,
+      "step": 15804
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1629322942036045e-07,
+      "loss": 0.0002,
+      "step": 15805
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1626887481734047e-07,
+      "loss": 1.9208,
+      "step": 15806
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1624452021432051e-07,
+      "loss": 0.0002,
+      "step": 15807
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1622016561130054e-07,
+      "loss": 0.0002,
+      "step": 15808
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1619581100828056e-07,
+      "loss": 0.0003,
+      "step": 15809
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 1.1617145640526059e-07,
+      "loss": 0.0004,
+      "step": 15810
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1614710180224062e-07,
+      "loss": 0.0002,
+      "step": 15811
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1612274719922065e-07,
+      "loss": 0.0003,
+      "step": 15812
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1609839259620067e-07,
+      "loss": 0.0002,
+      "step": 15813
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.160740379931807e-07,
+      "loss": 0.0003,
+      "step": 15814
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1604968339016074e-07,
+      "loss": 0.0001,
+      "step": 15815
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1602532878714076e-07,
+      "loss": 0.0004,
+      "step": 15816
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1600097418412079e-07,
+      "loss": 0.0001,
+      "step": 15817
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1597661958110081e-07,
+      "loss": 0.0003,
+      "step": 15818
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1595226497808085e-07,
+      "loss": 0.0003,
+      "step": 15819
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 1.1592791037506088e-07,
+      "loss": 0.0003,
+      "step": 15820
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1590355577204092e-07,
+      "loss": 0.0002,
+      "step": 15821
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1587920116902094e-07,
+      "loss": 0.0002,
+      "step": 15822
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1585484656600098e-07,
+      "loss": 0.0005,
+      "step": 15823
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.15830491962981e-07,
+      "loss": 0.0002,
+      "step": 15824
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1580613735996103e-07,
+      "loss": 0.0002,
+      "step": 15825
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1578178275694105e-07,
+      "loss": 0.0002,
+      "step": 15826
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1575742815392109e-07,
+      "loss": 0.0001,
+      "step": 15827
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1573307355090112e-07,
+      "loss": 0.5159,
+      "step": 15828
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1570871894788114e-07,
+      "loss": 0.0002,
+      "step": 15829
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1568436434486117e-07,
+      "loss": 0.0003,
+      "step": 15830
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 1.1566000974184121e-07,
+      "loss": 0.0002,
+      "step": 15831
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.1563565513882123e-07,
+      "loss": 0.0002,
+      "step": 15832
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.1561130053580126e-07,
+      "loss": 0.0004,
+      "step": 15833
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.1558694593278128e-07,
+      "loss": 0.0001,
+      "step": 15834
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.1556259132976132e-07,
+      "loss": 0.0002,
+      "step": 15835
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.1553823672674135e-07,
+      "loss": 0.0006,
+      "step": 15836
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.1551388212372137e-07,
+      "loss": 0.0002,
+      "step": 15837
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.154895275207014e-07,
+      "loss": 0.0002,
+      "step": 15838
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.1546517291768143e-07,
+      "loss": 0.0002,
+      "step": 15839
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.1544081831466146e-07,
+      "loss": 0.0003,
+      "step": 15840
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 1.154164637116415e-07,
+      "loss": 0.0002,
+      "step": 15841
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1539210910862152e-07,
+      "loss": 0.0002,
+      "step": 15842
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1536775450560156e-07,
+      "loss": 0.0003,
+      "step": 15843
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1534339990258159e-07,
+      "loss": 0.0003,
+      "step": 15844
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1531904529956161e-07,
+      "loss": 0.0003,
+      "step": 15845
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1529469069654164e-07,
+      "loss": 0.0003,
+      "step": 15846
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1527033609352168e-07,
+      "loss": 0.0002,
+      "step": 15847
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.152459814905017e-07,
+      "loss": 0.0004,
+      "step": 15848
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1522162688748173e-07,
+      "loss": 0.0005,
+      "step": 15849
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1519727228446175e-07,
+      "loss": 0.0003,
+      "step": 15850
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 1.1517291768144179e-07,
+      "loss": 0.0003,
+      "step": 15851
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1514856307842181e-07,
+      "loss": 0.0004,
+      "step": 15852
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1512420847540184e-07,
+      "loss": 0.0003,
+      "step": 15853
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1509985387238188e-07,
+      "loss": 0.0005,
+      "step": 15854
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.150754992693619e-07,
+      "loss": 0.0002,
+      "step": 15855
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1505114466634193e-07,
+      "loss": 0.0002,
+      "step": 15856
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1502679006332195e-07,
+      "loss": 0.0002,
+      "step": 15857
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1500243546030199e-07,
+      "loss": 0.0003,
+      "step": 15858
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1497808085728202e-07,
+      "loss": 0.0002,
+      "step": 15859
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1495372625426206e-07,
+      "loss": 0.0001,
+      "step": 15860
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1492937165124208e-07,
+      "loss": 0.0003,
+      "step": 15861
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 1.1490501704822212e-07,
+      "loss": 0.0004,
+      "step": 15862
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.1488066244520214e-07,
+      "loss": 0.0002,
+      "step": 15863
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.1485630784218217e-07,
+      "loss": 0.0002,
+      "step": 15864
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.148319532391622e-07,
+      "loss": 0.0002,
+      "step": 15865
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.1480759863614223e-07,
+      "loss": 0.0003,
+      "step": 15866
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.1478324403312226e-07,
+      "loss": 0.0003,
+      "step": 15867
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.1475888943010228e-07,
+      "loss": 0.0003,
+      "step": 15868
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.1473453482708231e-07,
+      "loss": 0.0002,
+      "step": 15869
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.1471018022406235e-07,
+      "loss": 0.0003,
+      "step": 15870
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.1468582562104237e-07,
+      "loss": 0.0002,
+      "step": 15871
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 1.146614710180224e-07,
+      "loss": 0.0001,
+      "step": 15872
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1463711641500242e-07,
+      "loss": 0.0002,
+      "step": 15873
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1461276181198246e-07,
+      "loss": 0.0002,
+      "step": 15874
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1458840720896249e-07,
+      "loss": 0.0001,
+      "step": 15875
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1456405260594251e-07,
+      "loss": 0.0003,
+      "step": 15876
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1453969800292254e-07,
+      "loss": 0.0002,
+      "step": 15877
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1451534339990257e-07,
+      "loss": 0.0002,
+      "step": 15878
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1449098879688261e-07,
+      "loss": 0.0002,
+      "step": 15879
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1446663419386264e-07,
+      "loss": 0.0001,
+      "step": 15880
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.1444227959084266e-07,
+      "loss": 0.0002,
+      "step": 15881
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 1.144179249878227e-07,
+      "loss": 0.0002,
+      "step": 15882
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1439357038480273e-07,
+      "loss": 0.0002,
+      "step": 15883
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1436921578178275e-07,
+      "loss": 0.0002,
+      "step": 15884
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1434486117876278e-07,
+      "loss": 0.0002,
+      "step": 15885
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1432050657574282e-07,
+      "loss": 0.0002,
+      "step": 15886
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1429615197272284e-07,
+      "loss": 0.0001,
+      "step": 15887
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1427179736970287e-07,
+      "loss": 0.0005,
+      "step": 15888
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1424744276668289e-07,
+      "loss": 0.0001,
+      "step": 15889
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1422308816366293e-07,
+      "loss": 0.0003,
+      "step": 15890
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1419873356064296e-07,
+      "loss": 0.0003,
+      "step": 15891
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 1.1417437895762298e-07,
+      "loss": 0.0001,
+      "step": 15892
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.14150024354603e-07,
+      "loss": 0.0002,
+      "step": 15893
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.1412566975158304e-07,
+      "loss": 0.0002,
+      "step": 15894
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.1410131514856307e-07,
+      "loss": 1.5116,
+      "step": 15895
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.140769605455431e-07,
+      "loss": 0.0003,
+      "step": 15896
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.1405260594252312e-07,
+      "loss": 0.0002,
+      "step": 15897
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.1402825133950317e-07,
+      "loss": 0.0004,
+      "step": 15898
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.140038967364832e-07,
+      "loss": 0.0003,
+      "step": 15899
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.1397954213346322e-07,
+      "loss": 0.0004,
+      "step": 15900
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.1395518753044326e-07,
+      "loss": 0.0002,
+      "step": 15901
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.1393083292742328e-07,
+      "loss": 0.0002,
+      "step": 15902
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 1.1390647832440331e-07,
+      "loss": 0.0002,
+      "step": 15903
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.1388212372138334e-07,
+      "loss": 0.0002,
+      "step": 15904
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.1385776911836337e-07,
+      "loss": 0.0002,
+      "step": 15905
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.138334145153434e-07,
+      "loss": 0.0003,
+      "step": 15906
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.1380905991232342e-07,
+      "loss": 0.0002,
+      "step": 15907
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.1378470530930345e-07,
+      "loss": 0.0002,
+      "step": 15908
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.1376035070628349e-07,
+      "loss": 0.0003,
+      "step": 15909
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.1373599610326351e-07,
+      "loss": 0.0003,
+      "step": 15910
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.1371164150024354e-07,
+      "loss": 0.0003,
+      "step": 15911
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.1368728689722356e-07,
+      "loss": 0.0006,
+      "step": 15912
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 1.136629322942036e-07,
+      "loss": 0.0001,
+      "step": 15913
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1363857769118363e-07,
+      "loss": 0.0003,
+      "step": 15914
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1361422308816365e-07,
+      "loss": 0.0002,
+      "step": 15915
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1358986848514368e-07,
+      "loss": 0.0003,
+      "step": 15916
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1356551388212373e-07,
+      "loss": 0.0003,
+      "step": 15917
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1354115927910375e-07,
+      "loss": 0.0048,
+      "step": 15918
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1351680467608378e-07,
+      "loss": 0.0002,
+      "step": 15919
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.134924500730638e-07,
+      "loss": 0.0002,
+      "step": 15920
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1346809547004384e-07,
+      "loss": 0.0004,
+      "step": 15921
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1344374086702387e-07,
+      "loss": 0.0003,
+      "step": 15922
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 1.1341938626400389e-07,
+      "loss": 0.0002,
+      "step": 15923
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1339503166098392e-07,
+      "loss": 0.0003,
+      "step": 15924
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1337067705796396e-07,
+      "loss": 0.0002,
+      "step": 15925
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1334632245494398e-07,
+      "loss": 0.0003,
+      "step": 15926
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.13321967851924e-07,
+      "loss": 0.0002,
+      "step": 15927
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1329761324890403e-07,
+      "loss": 0.0003,
+      "step": 15928
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1327325864588407e-07,
+      "loss": 0.0003,
+      "step": 15929
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.132489040428641e-07,
+      "loss": 0.0003,
+      "step": 15930
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1322454943984412e-07,
+      "loss": 0.0001,
+      "step": 15931
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1320019483682415e-07,
+      "loss": 0.0003,
+      "step": 15932
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1317584023380418e-07,
+      "loss": 0.0004,
+      "step": 15933
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 1.1315148563078421e-07,
+      "loss": 0.0001,
+      "step": 15934
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1312713102776423e-07,
+      "loss": 0.0006,
+      "step": 15935
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1310277642474427e-07,
+      "loss": 0.0005,
+      "step": 15936
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1307842182172431e-07,
+      "loss": 0.0002,
+      "step": 15937
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1305406721870434e-07,
+      "loss": 0.0003,
+      "step": 15938
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1302971261568436e-07,
+      "loss": 0.0002,
+      "step": 15939
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1300535801266439e-07,
+      "loss": 0.0002,
+      "step": 15940
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1298100340964443e-07,
+      "loss": 0.0004,
+      "step": 15941
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1295664880662445e-07,
+      "loss": 0.0003,
+      "step": 15942
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.1293229420360448e-07,
+      "loss": 0.0004,
+      "step": 15943
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 1.129079396005845e-07,
+      "loss": 0.0003,
+      "step": 15944
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1288358499756454e-07,
+      "loss": 0.0002,
+      "step": 15945
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1285923039454456e-07,
+      "loss": 1.7665,
+      "step": 15946
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1283487579152459e-07,
+      "loss": 0.0004,
+      "step": 15947
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1281052118850461e-07,
+      "loss": 0.0002,
+      "step": 15948
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1278616658548465e-07,
+      "loss": 0.0003,
+      "step": 15949
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1276181198246468e-07,
+      "loss": 0.0003,
+      "step": 15950
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.127374573794447e-07,
+      "loss": 0.0002,
+      "step": 15951
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1271310277642473e-07,
+      "loss": 0.0002,
+      "step": 15952
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1268874817340477e-07,
+      "loss": 0.0002,
+      "step": 15953
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 1.1266439357038479e-07,
+      "loss": 0.0003,
+      "step": 15954
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.1264003896736482e-07,
+      "loss": 0.0003,
+      "step": 15955
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.1261568436434487e-07,
+      "loss": 0.0003,
+      "step": 15956
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.125913297613249e-07,
+      "loss": 0.0004,
+      "step": 15957
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.1256697515830492e-07,
+      "loss": 0.0002,
+      "step": 15958
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.1254262055528494e-07,
+      "loss": 0.0003,
+      "step": 15959
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.1251826595226498e-07,
+      "loss": 0.0002,
+      "step": 15960
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.1249391134924501e-07,
+      "loss": 0.0002,
+      "step": 15961
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.1246955674622503e-07,
+      "loss": 0.0002,
+      "step": 15962
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.1244520214320506e-07,
+      "loss": 0.0004,
+      "step": 15963
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 1.124208475401851e-07,
+      "loss": 0.0003,
+      "step": 15964
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1239649293716512e-07,
+      "loss": 0.0002,
+      "step": 15965
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1237213833414515e-07,
+      "loss": 0.0005,
+      "step": 15966
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1234778373112517e-07,
+      "loss": 0.0004,
+      "step": 15967
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1232342912810521e-07,
+      "loss": 0.0003,
+      "step": 15968
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1229907452508524e-07,
+      "loss": 0.0002,
+      "step": 15969
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1227471992206526e-07,
+      "loss": 0.0002,
+      "step": 15970
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1225036531904529e-07,
+      "loss": 0.0002,
+      "step": 15971
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1222601071602532e-07,
+      "loss": 0.0003,
+      "step": 15972
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1220165611300535e-07,
+      "loss": 0.0002,
+      "step": 15973
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1217730150998537e-07,
+      "loss": 0.0003,
+      "step": 15974
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 1.1215294690696541e-07,
+      "loss": 0.0004,
+      "step": 15975
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.1212859230394545e-07,
+      "loss": 0.0002,
+      "step": 15976
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.1210423770092548e-07,
+      "loss": 0.0004,
+      "step": 15977
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.120798830979055e-07,
+      "loss": 0.0001,
+      "step": 15978
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.1205552849488553e-07,
+      "loss": 0.0002,
+      "step": 15979
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.1203117389186557e-07,
+      "loss": 0.0002,
+      "step": 15980
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.1200681928884559e-07,
+      "loss": 0.0002,
+      "step": 15981
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.1198246468582562e-07,
+      "loss": 0.0004,
+      "step": 15982
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.1195811008280564e-07,
+      "loss": 0.0004,
+      "step": 15983
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.1193375547978568e-07,
+      "loss": 0.0002,
+      "step": 15984
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 1.119094008767657e-07,
+      "loss": 0.0003,
+      "step": 15985
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1188504627374573e-07,
+      "loss": 0.0002,
+      "step": 15986
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1186069167072575e-07,
+      "loss": 0.0002,
+      "step": 15987
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1183633706770579e-07,
+      "loss": 0.0003,
+      "step": 15988
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1181198246468582e-07,
+      "loss": 0.0002,
+      "step": 15989
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1178762786166584e-07,
+      "loss": 0.0002,
+      "step": 15990
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1176327325864587e-07,
+      "loss": 0.0001,
+      "step": 15991
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1173891865562591e-07,
+      "loss": 0.0002,
+      "step": 15992
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1171456405260593e-07,
+      "loss": 0.0003,
+      "step": 15993
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.1169020944958597e-07,
+      "loss": 0.0003,
+      "step": 15994
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 1.11665854846566e-07,
+      "loss": 0.0003,
+      "step": 15995
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1164150024354603e-07,
+      "loss": 0.0011,
+      "step": 15996
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1161714564052606e-07,
+      "loss": 0.0008,
+      "step": 15997
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1159279103750608e-07,
+      "loss": 0.0002,
+      "step": 15998
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1156843643448611e-07,
+      "loss": 0.0002,
+      "step": 15999
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1154408183146615e-07,
+      "loss": 0.0002,
+      "step": 16000
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1151972722844617e-07,
+      "loss": 0.0003,
+      "step": 16001
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.114953726254262e-07,
+      "loss": 0.0004,
+      "step": 16002
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1147101802240622e-07,
+      "loss": 0.0003,
+      "step": 16003
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1144666341938626e-07,
+      "loss": 0.0001,
+      "step": 16004
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1142230881636629e-07,
+      "loss": 0.0003,
+      "step": 16005
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 1.1139795421334631e-07,
+      "loss": 0.0003,
+      "step": 16006
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1137359961032634e-07,
+      "loss": 0.0001,
+      "step": 16007
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1134924500730638e-07,
+      "loss": 0.0001,
+      "step": 16008
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.113248904042864e-07,
+      "loss": 0.0004,
+      "step": 16009
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1130053580126643e-07,
+      "loss": 0.0002,
+      "step": 16010
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1127618119824645e-07,
+      "loss": 0.0002,
+      "step": 16011
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1125182659522649e-07,
+      "loss": 0.0003,
+      "step": 16012
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1122747199220653e-07,
+      "loss": 0.0004,
+      "step": 16013
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1120311738918655e-07,
+      "loss": 0.0003,
+      "step": 16014
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1117876278616659e-07,
+      "loss": 0.0002,
+      "step": 16015
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 1.1115440818314662e-07,
+      "loss": 0.0003,
+      "step": 16016
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.1113005358012664e-07,
+      "loss": 0.0004,
+      "step": 16017
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.1110569897710667e-07,
+      "loss": 0.0002,
+      "step": 16018
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.110813443740867e-07,
+      "loss": 0.0004,
+      "step": 16019
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.1105698977106673e-07,
+      "loss": 0.0002,
+      "step": 16020
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.1103263516804676e-07,
+      "loss": 0.0002,
+      "step": 16021
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.1100828056502678e-07,
+      "loss": 0.0002,
+      "step": 16022
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.1098392596200682e-07,
+      "loss": 0.0002,
+      "step": 16023
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.1095957135898684e-07,
+      "loss": 0.0002,
+      "step": 16024
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.1093521675596687e-07,
+      "loss": 0.0003,
+      "step": 16025
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 1.109108621529469e-07,
+      "loss": 0.0002,
+      "step": 16026
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1088650754992693e-07,
+      "loss": 0.0002,
+      "step": 16027
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1086215294690696e-07,
+      "loss": 0.0002,
+      "step": 16028
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1083779834388698e-07,
+      "loss": 0.0004,
+      "step": 16029
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1081344374086701e-07,
+      "loss": 0.0002,
+      "step": 16030
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1078908913784705e-07,
+      "loss": 0.0002,
+      "step": 16031
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1076473453482709e-07,
+      "loss": 0.0003,
+      "step": 16032
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1074037993180711e-07,
+      "loss": 0.0002,
+      "step": 16033
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1071602532878714e-07,
+      "loss": 0.0002,
+      "step": 16034
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.1069167072576717e-07,
+      "loss": 0.0002,
+      "step": 16035
+    },
+    {
+      "epoch": 15.58,
+      "learning_rate": 1.106673161227472e-07,
+      "loss": 0.0003,
+      "step": 16036
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1064296151972722e-07,
+      "loss": 0.0002,
+      "step": 16037
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1061860691670725e-07,
+      "loss": 0.0003,
+      "step": 16038
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1059425231368729e-07,
+      "loss": 0.0003,
+      "step": 16039
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1056989771066731e-07,
+      "loss": 0.0003,
+      "step": 16040
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1054554310764734e-07,
+      "loss": 0.0003,
+      "step": 16041
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1052118850462736e-07,
+      "loss": 0.0005,
+      "step": 16042
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.104968339016074e-07,
+      "loss": 0.0003,
+      "step": 16043
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1047247929858743e-07,
+      "loss": 0.0003,
+      "step": 16044
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1044812469556745e-07,
+      "loss": 0.0003,
+      "step": 16045
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1042377009254748e-07,
+      "loss": 0.0003,
+      "step": 16046
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 1.1039941548952752e-07,
+      "loss": 0.0002,
+      "step": 16047
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1037506088650754e-07,
+      "loss": 0.0002,
+      "step": 16048
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1035070628348757e-07,
+      "loss": 0.0002,
+      "step": 16049
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1032635168046759e-07,
+      "loss": 0.0005,
+      "step": 16050
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1030199707744764e-07,
+      "loss": 0.0002,
+      "step": 16051
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1027764247442767e-07,
+      "loss": 0.0003,
+      "step": 16052
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.102532878714077e-07,
+      "loss": 0.0002,
+      "step": 16053
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1022893326838772e-07,
+      "loss": 0.0003,
+      "step": 16054
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1020457866536776e-07,
+      "loss": 0.0004,
+      "step": 16055
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1018022406234778e-07,
+      "loss": 0.0001,
+      "step": 16056
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 1.1015586945932781e-07,
+      "loss": 0.0003,
+      "step": 16057
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.1013151485630783e-07,
+      "loss": 0.0002,
+      "step": 16058
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.1010716025328787e-07,
+      "loss": 0.0002,
+      "step": 16059
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.100828056502679e-07,
+      "loss": 0.0003,
+      "step": 16060
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.1005845104724792e-07,
+      "loss": 0.0003,
+      "step": 16061
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.1003409644422795e-07,
+      "loss": 0.0002,
+      "step": 16062
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.1000974184120798e-07,
+      "loss": 0.0002,
+      "step": 16063
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.0998538723818801e-07,
+      "loss": 0.0002,
+      "step": 16064
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.0996103263516804e-07,
+      "loss": 0.0003,
+      "step": 16065
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.0993667803214806e-07,
+      "loss": 0.0003,
+      "step": 16066
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 1.099123234291281e-07,
+      "loss": 0.0003,
+      "step": 16067
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0988796882610812e-07,
+      "loss": 0.0002,
+      "step": 16068
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0986361422308815e-07,
+      "loss": 0.0003,
+      "step": 16069
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.098392596200682e-07,
+      "loss": 0.0003,
+      "step": 16070
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0981490501704823e-07,
+      "loss": 0.0002,
+      "step": 16071
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0979055041402825e-07,
+      "loss": 0.0001,
+      "step": 16072
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0976619581100828e-07,
+      "loss": 0.0001,
+      "step": 16073
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0974184120798831e-07,
+      "loss": 0.0004,
+      "step": 16074
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0971748660496834e-07,
+      "loss": 0.0002,
+      "step": 16075
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0969313200194837e-07,
+      "loss": 0.0002,
+      "step": 16076
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0966877739892839e-07,
+      "loss": 0.0001,
+      "step": 16077
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 1.0964442279590843e-07,
+      "loss": 0.0003,
+      "step": 16078
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0962006819288845e-07,
+      "loss": 0.0002,
+      "step": 16079
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0959571358986848e-07,
+      "loss": 0.0002,
+      "step": 16080
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.095713589868485e-07,
+      "loss": 0.0002,
+      "step": 16081
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0954700438382854e-07,
+      "loss": 0.0001,
+      "step": 16082
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0952264978080857e-07,
+      "loss": 0.0002,
+      "step": 16083
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0949829517778859e-07,
+      "loss": 0.0006,
+      "step": 16084
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0947394057476862e-07,
+      "loss": 0.0003,
+      "step": 16085
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0944958597174866e-07,
+      "loss": 0.0003,
+      "step": 16086
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0942523136872868e-07,
+      "loss": 0.0003,
+      "step": 16087
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 1.0940087676570871e-07,
+      "loss": 0.0002,
+      "step": 16088
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0937652216268873e-07,
+      "loss": 0.0002,
+      "step": 16089
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0935216755966878e-07,
+      "loss": 0.0002,
+      "step": 16090
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0932781295664881e-07,
+      "loss": 0.0002,
+      "step": 16091
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0930345835362883e-07,
+      "loss": 0.0002,
+      "step": 16092
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0927910375060886e-07,
+      "loss": 0.0018,
+      "step": 16093
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.092547491475889e-07,
+      "loss": 0.0002,
+      "step": 16094
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0923039454456892e-07,
+      "loss": 0.0003,
+      "step": 16095
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0920603994154895e-07,
+      "loss": 0.0003,
+      "step": 16096
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0918168533852897e-07,
+      "loss": 0.0001,
+      "step": 16097
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 1.0915733073550901e-07,
+      "loss": 0.0002,
+      "step": 16098
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0913297613248904e-07,
+      "loss": 0.0002,
+      "step": 16099
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0910862152946906e-07,
+      "loss": 0.0003,
+      "step": 16100
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0908426692644909e-07,
+      "loss": 0.0003,
+      "step": 16101
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0905991232342913e-07,
+      "loss": 0.0002,
+      "step": 16102
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0903555772040915e-07,
+      "loss": 0.0005,
+      "step": 16103
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0901120311738918e-07,
+      "loss": 0.0004,
+      "step": 16104
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.089868485143692e-07,
+      "loss": 0.0004,
+      "step": 16105
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0896249391134924e-07,
+      "loss": 0.0002,
+      "step": 16106
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0893813930832926e-07,
+      "loss": 0.0002,
+      "step": 16107
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 1.0891378470530929e-07,
+      "loss": 0.0003,
+      "step": 16108
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0888943010228933e-07,
+      "loss": 0.0002,
+      "step": 16109
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0886507549926937e-07,
+      "loss": 0.0003,
+      "step": 16110
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0884072089624939e-07,
+      "loss": 0.0002,
+      "step": 16111
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0881636629322942e-07,
+      "loss": 0.0002,
+      "step": 16112
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0879201169020944e-07,
+      "loss": 0.0003,
+      "step": 16113
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0876765708718948e-07,
+      "loss": 0.0002,
+      "step": 16114
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.087433024841695e-07,
+      "loss": 0.0003,
+      "step": 16115
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0871894788114953e-07,
+      "loss": 0.0003,
+      "step": 16116
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0869459327812956e-07,
+      "loss": 0.0002,
+      "step": 16117
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.086702386751096e-07,
+      "loss": 0.0002,
+      "step": 16118
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 1.0864588407208962e-07,
+      "loss": 0.0003,
+      "step": 16119
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0862152946906964e-07,
+      "loss": 0.0002,
+      "step": 16120
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0859717486604967e-07,
+      "loss": 0.0003,
+      "step": 16121
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0857282026302971e-07,
+      "loss": 0.0002,
+      "step": 16122
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0854846566000973e-07,
+      "loss": 0.0003,
+      "step": 16123
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0852411105698976e-07,
+      "loss": 0.0002,
+      "step": 16124
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.084997564539698e-07,
+      "loss": 0.0001,
+      "step": 16125
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0847540185094982e-07,
+      "loss": 0.0003,
+      "step": 16126
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0845104724792985e-07,
+      "loss": 0.0002,
+      "step": 16127
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0842669264490989e-07,
+      "loss": 0.0003,
+      "step": 16128
+    },
+    {
+      "epoch": 15.67,
+      "learning_rate": 1.0840233804188992e-07,
+      "loss": 0.0003,
+      "step": 16129
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0837798343886995e-07,
+      "loss": 0.0002,
+      "step": 16130
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0835362883584997e-07,
+      "loss": 0.0002,
+      "step": 16131
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0832927423283e-07,
+      "loss": 0.0003,
+      "step": 16132
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0830491962981004e-07,
+      "loss": 0.0003,
+      "step": 16133
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0828056502679006e-07,
+      "loss": 0.0002,
+      "step": 16134
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0825621042377009e-07,
+      "loss": 0.0003,
+      "step": 16135
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0823185582075011e-07,
+      "loss": 0.0001,
+      "step": 16136
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0820750121773015e-07,
+      "loss": 0.0001,
+      "step": 16137
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.0818314661471018e-07,
+      "loss": 0.0003,
+      "step": 16138
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 1.081587920116902e-07,
+      "loss": 0.0003,
+      "step": 16139
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0813443740867023e-07,
+      "loss": 0.0003,
+      "step": 16140
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0811008280565027e-07,
+      "loss": 0.0002,
+      "step": 16141
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0808572820263029e-07,
+      "loss": 0.0003,
+      "step": 16142
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0806137359961032e-07,
+      "loss": 0.0002,
+      "step": 16143
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0803701899659034e-07,
+      "loss": 0.0003,
+      "step": 16144
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0801266439357038e-07,
+      "loss": 0.027,
+      "step": 16145
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.079883097905504e-07,
+      "loss": 0.0004,
+      "step": 16146
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0796395518753044e-07,
+      "loss": 0.0005,
+      "step": 16147
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0793960058451047e-07,
+      "loss": 0.0002,
+      "step": 16148
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.079152459814905e-07,
+      "loss": 0.0002,
+      "step": 16149
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 1.0789089137847053e-07,
+      "loss": 0.0002,
+      "step": 16150
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0786653677545056e-07,
+      "loss": 0.0002,
+      "step": 16151
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0784218217243058e-07,
+      "loss": 0.0003,
+      "step": 16152
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0781782756941062e-07,
+      "loss": 0.0004,
+      "step": 16153
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0779347296639065e-07,
+      "loss": 0.0001,
+      "step": 16154
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0776911836337067e-07,
+      "loss": 0.0002,
+      "step": 16155
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.077447637603507e-07,
+      "loss": 0.0003,
+      "step": 16156
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0772040915733073e-07,
+      "loss": 0.0002,
+      "step": 16157
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0769605455431076e-07,
+      "loss": 0.0002,
+      "step": 16158
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0767169995129078e-07,
+      "loss": 0.0003,
+      "step": 16159
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 1.0764734534827081e-07,
+      "loss": 0.0002,
+      "step": 16160
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.0762299074525085e-07,
+      "loss": 0.0003,
+      "step": 16161
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.0759863614223087e-07,
+      "loss": 0.0002,
+      "step": 16162
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.075742815392109e-07,
+      "loss": 0.0003,
+      "step": 16163
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.0754992693619092e-07,
+      "loss": 0.0003,
+      "step": 16164
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.0752557233317096e-07,
+      "loss": 0.0002,
+      "step": 16165
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.07501217730151e-07,
+      "loss": 0.0002,
+      "step": 16166
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.0747686312713103e-07,
+      "loss": 0.0002,
+      "step": 16167
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.0745250852411105e-07,
+      "loss": 0.0002,
+      "step": 16168
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.0742815392109109e-07,
+      "loss": 0.0003,
+      "step": 16169
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 1.0740379931807111e-07,
+      "loss": 0.0002,
+      "step": 16170
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0737944471505114e-07,
+      "loss": 0.0002,
+      "step": 16171
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0735509011203118e-07,
+      "loss": 0.0004,
+      "step": 16172
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.073307355090112e-07,
+      "loss": 0.0002,
+      "step": 16173
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0730638090599123e-07,
+      "loss": 0.0003,
+      "step": 16174
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0728202630297125e-07,
+      "loss": 0.0001,
+      "step": 16175
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0725767169995129e-07,
+      "loss": 0.0002,
+      "step": 16176
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0723331709693132e-07,
+      "loss": 0.0003,
+      "step": 16177
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0720896249391134e-07,
+      "loss": 0.0002,
+      "step": 16178
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0718460789089137e-07,
+      "loss": 0.0003,
+      "step": 16179
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.071602532878714e-07,
+      "loss": 0.0003,
+      "step": 16180
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 1.0713589868485143e-07,
+      "loss": 0.0003,
+      "step": 16181
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0711154408183146e-07,
+      "loss": 0.0003,
+      "step": 16182
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0708718947881148e-07,
+      "loss": 0.0003,
+      "step": 16183
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0706283487579152e-07,
+      "loss": 0.0003,
+      "step": 16184
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0703848027277156e-07,
+      "loss": 0.0002,
+      "step": 16185
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0701412566975158e-07,
+      "loss": 0.0003,
+      "step": 16186
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0698977106673161e-07,
+      "loss": 0.0002,
+      "step": 16187
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0696541646371165e-07,
+      "loss": 0.0003,
+      "step": 16188
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0694106186069167e-07,
+      "loss": 0.0003,
+      "step": 16189
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.069167072576717e-07,
+      "loss": 0.0002,
+      "step": 16190
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 1.0689235265465172e-07,
+      "loss": 0.0002,
+      "step": 16191
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0686799805163176e-07,
+      "loss": 0.0003,
+      "step": 16192
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0684364344861179e-07,
+      "loss": 0.0003,
+      "step": 16193
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0681928884559181e-07,
+      "loss": 0.0002,
+      "step": 16194
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0679493424257184e-07,
+      "loss": 0.0026,
+      "step": 16195
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0677057963955187e-07,
+      "loss": 0.0002,
+      "step": 16196
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.067462250365319e-07,
+      "loss": 0.0001,
+      "step": 16197
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0672187043351192e-07,
+      "loss": 0.0002,
+      "step": 16198
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0669751583049195e-07,
+      "loss": 0.0001,
+      "step": 16199
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0667316122747199e-07,
+      "loss": 0.0003,
+      "step": 16200
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 1.0664880662445201e-07,
+      "loss": 0.0004,
+      "step": 16201
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0662445202143204e-07,
+      "loss": 0.0003,
+      "step": 16202
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0660009741841206e-07,
+      "loss": 0.0002,
+      "step": 16203
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0657574281539212e-07,
+      "loss": 0.0001,
+      "step": 16204
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0655138821237214e-07,
+      "loss": 0.0003,
+      "step": 16205
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0652703360935217e-07,
+      "loss": 0.0003,
+      "step": 16206
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0650267900633219e-07,
+      "loss": 0.0002,
+      "step": 16207
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0647832440331223e-07,
+      "loss": 0.0003,
+      "step": 16208
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0645396980029225e-07,
+      "loss": 0.0001,
+      "step": 16209
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.0642961519727228e-07,
+      "loss": 0.0002,
+      "step": 16210
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 1.064052605942523e-07,
+      "loss": 0.0003,
+      "step": 16211
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0638090599123234e-07,
+      "loss": 0.0001,
+      "step": 16212
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0635655138821237e-07,
+      "loss": 0.0003,
+      "step": 16213
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.063321967851924e-07,
+      "loss": 0.0002,
+      "step": 16214
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0630784218217242e-07,
+      "loss": 0.0003,
+      "step": 16215
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0628348757915246e-07,
+      "loss": 0.0003,
+      "step": 16216
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0625913297613248e-07,
+      "loss": 0.0003,
+      "step": 16217
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0623477837311251e-07,
+      "loss": 0.0002,
+      "step": 16218
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0621042377009253e-07,
+      "loss": 0.1312,
+      "step": 16219
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0618606916707257e-07,
+      "loss": 0.0001,
+      "step": 16220
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.061617145640526e-07,
+      "loss": 0.6025,
+      "step": 16221
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 1.0613735996103262e-07,
+      "loss": 0.0002,
+      "step": 16222
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.0611300535801265e-07,
+      "loss": 0.001,
+      "step": 16223
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.060886507549927e-07,
+      "loss": 0.0001,
+      "step": 16224
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.0606429615197272e-07,
+      "loss": 0.0003,
+      "step": 16225
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.0603994154895275e-07,
+      "loss": 0.0001,
+      "step": 16226
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.0601558694593279e-07,
+      "loss": 0.0002,
+      "step": 16227
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.0599123234291281e-07,
+      "loss": 0.0002,
+      "step": 16228
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.0596687773989284e-07,
+      "loss": 0.0002,
+      "step": 16229
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.0594252313687286e-07,
+      "loss": 0.0005,
+      "step": 16230
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.059181685338529e-07,
+      "loss": 0.0003,
+      "step": 16231
+    },
+    {
+      "epoch": 15.77,
+      "learning_rate": 1.0589381393083293e-07,
+      "loss": 0.0004,
+      "step": 16232
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0586945932781295e-07,
+      "loss": 0.0002,
+      "step": 16233
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0584510472479298e-07,
+      "loss": 0.0003,
+      "step": 16234
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0582075012177301e-07,
+      "loss": 0.0001,
+      "step": 16235
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0579639551875304e-07,
+      "loss": 0.0002,
+      "step": 16236
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0577204091573307e-07,
+      "loss": 0.0003,
+      "step": 16237
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0574768631271309e-07,
+      "loss": 0.0004,
+      "step": 16238
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0572333170969313e-07,
+      "loss": 0.0002,
+      "step": 16239
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0569897710667315e-07,
+      "loss": 0.0004,
+      "step": 16240
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.0567462250365318e-07,
+      "loss": 0.0002,
+      "step": 16241
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 1.056502679006332e-07,
+      "loss": 0.0002,
+      "step": 16242
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0562591329761326e-07,
+      "loss": 0.0002,
+      "step": 16243
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0560155869459328e-07,
+      "loss": 0.0002,
+      "step": 16244
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.055772040915733e-07,
+      "loss": 0.0002,
+      "step": 16245
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0555284948855333e-07,
+      "loss": 0.0038,
+      "step": 16246
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0552849488553337e-07,
+      "loss": 0.0002,
+      "step": 16247
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.055041402825134e-07,
+      "loss": 0.0002,
+      "step": 16248
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0547978567949342e-07,
+      "loss": 0.0002,
+      "step": 16249
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0545543107647345e-07,
+      "loss": 0.0001,
+      "step": 16250
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0543107647345348e-07,
+      "loss": 0.0002,
+      "step": 16251
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0540672187043351e-07,
+      "loss": 0.0002,
+      "step": 16252
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 1.0538236726741353e-07,
+      "loss": 0.0004,
+      "step": 16253
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.0535801266439356e-07,
+      "loss": 0.0002,
+      "step": 16254
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.053336580613736e-07,
+      "loss": 0.0002,
+      "step": 16255
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.0530930345835362e-07,
+      "loss": 0.0002,
+      "step": 16256
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.0528494885533365e-07,
+      "loss": 0.0002,
+      "step": 16257
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.0526059425231367e-07,
+      "loss": 0.0002,
+      "step": 16258
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.0523623964929371e-07,
+      "loss": 0.0003,
+      "step": 16259
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.0521188504627374e-07,
+      "loss": 0.0002,
+      "step": 16260
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.0518753044325376e-07,
+      "loss": 0.0004,
+      "step": 16261
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.051631758402338e-07,
+      "loss": 0.0004,
+      "step": 16262
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 1.0513882123721384e-07,
+      "loss": 0.0002,
+      "step": 16263
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0511446663419386e-07,
+      "loss": 0.0002,
+      "step": 16264
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0509011203117389e-07,
+      "loss": 0.0004,
+      "step": 16265
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0506575742815391e-07,
+      "loss": 0.0002,
+      "step": 16266
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0504140282513395e-07,
+      "loss": 0.0002,
+      "step": 16267
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0501704822211398e-07,
+      "loss": 0.0002,
+      "step": 16268
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.04992693619094e-07,
+      "loss": 0.0003,
+      "step": 16269
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0496833901607403e-07,
+      "loss": 0.0002,
+      "step": 16270
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0494398441305407e-07,
+      "loss": 0.0004,
+      "step": 16271
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0491962981003409e-07,
+      "loss": 0.0003,
+      "step": 16272
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 1.0489527520701412e-07,
+      "loss": 0.0002,
+      "step": 16273
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.0487092060399414e-07,
+      "loss": 0.0002,
+      "step": 16274
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.0484656600097418e-07,
+      "loss": 0.0002,
+      "step": 16275
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.048222113979542e-07,
+      "loss": 0.0003,
+      "step": 16276
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.0479785679493423e-07,
+      "loss": 0.0002,
+      "step": 16277
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.0477350219191426e-07,
+      "loss": 0.0002,
+      "step": 16278
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.047491475888943e-07,
+      "loss": 0.0002,
+      "step": 16279
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.0472479298587432e-07,
+      "loss": 0.0002,
+      "step": 16280
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.0470043838285436e-07,
+      "loss": 0.0002,
+      "step": 16281
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.046760837798344e-07,
+      "loss": 0.0003,
+      "step": 16282
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 1.0465172917681442e-07,
+      "loss": 0.0001,
+      "step": 16283
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0462737457379445e-07,
+      "loss": 0.0002,
+      "step": 16284
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0460301997077447e-07,
+      "loss": 0.0003,
+      "step": 16285
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0457866536775451e-07,
+      "loss": 0.0003,
+      "step": 16286
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0455431076473454e-07,
+      "loss": 0.0003,
+      "step": 16287
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0452995616171456e-07,
+      "loss": 0.0002,
+      "step": 16288
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0450560155869459e-07,
+      "loss": 0.0003,
+      "step": 16289
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0448124695567462e-07,
+      "loss": 0.0002,
+      "step": 16290
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0445689235265465e-07,
+      "loss": 0.0002,
+      "step": 16291
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0443253774963467e-07,
+      "loss": 0.0004,
+      "step": 16292
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.044081831466147e-07,
+      "loss": 0.0003,
+      "step": 16293
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 1.0438382854359474e-07,
+      "loss": 0.0002,
+      "step": 16294
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0435947394057476e-07,
+      "loss": 0.0003,
+      "step": 16295
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0433511933755479e-07,
+      "loss": 0.0002,
+      "step": 16296
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0431076473453481e-07,
+      "loss": 0.0002,
+      "step": 16297
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0428641013151485e-07,
+      "loss": 0.0002,
+      "step": 16298
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0426205552849488e-07,
+      "loss": 0.0002,
+      "step": 16299
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0423770092547492e-07,
+      "loss": 0.0002,
+      "step": 16300
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0421334632245494e-07,
+      "loss": 0.0003,
+      "step": 16301
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0418899171943498e-07,
+      "loss": 0.0002,
+      "step": 16302
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.04164637116415e-07,
+      "loss": 0.0002,
+      "step": 16303
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 1.0414028251339503e-07,
+      "loss": 0.0002,
+      "step": 16304
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0411592791037505e-07,
+      "loss": 0.0003,
+      "step": 16305
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0409157330735509e-07,
+      "loss": 0.0003,
+      "step": 16306
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0406721870433512e-07,
+      "loss": 0.0003,
+      "step": 16307
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0404286410131514e-07,
+      "loss": 0.0002,
+      "step": 16308
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0401850949829517e-07,
+      "loss": 0.0002,
+      "step": 16309
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0399415489527521e-07,
+      "loss": 0.0004,
+      "step": 16310
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0396980029225523e-07,
+      "loss": 0.0003,
+      "step": 16311
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0394544568923526e-07,
+      "loss": 0.0007,
+      "step": 16312
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0392109108621528e-07,
+      "loss": 0.0001,
+      "step": 16313
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 1.0389673648319532e-07,
+      "loss": 0.0003,
+      "step": 16314
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0387238188017535e-07,
+      "loss": 0.0003,
+      "step": 16315
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0384802727715537e-07,
+      "loss": 0.0003,
+      "step": 16316
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.038236726741354e-07,
+      "loss": 0.0003,
+      "step": 16317
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0379931807111543e-07,
+      "loss": 0.0002,
+      "step": 16318
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0377496346809547e-07,
+      "loss": 0.0002,
+      "step": 16319
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.037506088650755e-07,
+      "loss": 0.0002,
+      "step": 16320
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0372625426205552e-07,
+      "loss": 0.0002,
+      "step": 16321
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0370189965903556e-07,
+      "loss": 0.0002,
+      "step": 16322
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0367754505601559e-07,
+      "loss": 0.0003,
+      "step": 16323
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0365319045299561e-07,
+      "loss": 0.0003,
+      "step": 16324
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 1.0362883584997564e-07,
+      "loss": 0.0001,
+      "step": 16325
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.0360448124695568e-07,
+      "loss": 0.0003,
+      "step": 16326
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.035801266439357e-07,
+      "loss": 0.0003,
+      "step": 16327
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.0355577204091573e-07,
+      "loss": 0.0004,
+      "step": 16328
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.0353141743789575e-07,
+      "loss": 0.0001,
+      "step": 16329
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.0350706283487579e-07,
+      "loss": 0.0001,
+      "step": 16330
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.0348270823185581e-07,
+      "loss": 0.0003,
+      "step": 16331
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.0345835362883584e-07,
+      "loss": 0.0001,
+      "step": 16332
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.0343399902581586e-07,
+      "loss": 0.0004,
+      "step": 16333
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.034096444227959e-07,
+      "loss": 0.0002,
+      "step": 16334
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 1.0338528981977593e-07,
+      "loss": 0.0002,
+      "step": 16335
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0336093521675595e-07,
+      "loss": 2.3707,
+      "step": 16336
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0333658061373598e-07,
+      "loss": 0.0003,
+      "step": 16337
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0331222601071602e-07,
+      "loss": 0.0002,
+      "step": 16338
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0328787140769606e-07,
+      "loss": 0.0002,
+      "step": 16339
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0326351680467608e-07,
+      "loss": 0.0005,
+      "step": 16340
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0323916220165612e-07,
+      "loss": 0.0002,
+      "step": 16341
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0321480759863614e-07,
+      "loss": 0.0003,
+      "step": 16342
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0319045299561617e-07,
+      "loss": 0.0003,
+      "step": 16343
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.031660983925962e-07,
+      "loss": 0.0002,
+      "step": 16344
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 1.0314174378957623e-07,
+      "loss": 0.0002,
+      "step": 16345
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0311738918655626e-07,
+      "loss": 0.0003,
+      "step": 16346
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0309303458353628e-07,
+      "loss": 0.0004,
+      "step": 16347
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0306867998051631e-07,
+      "loss": 0.0002,
+      "step": 16348
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0304432537749635e-07,
+      "loss": 0.0003,
+      "step": 16349
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0301997077447637e-07,
+      "loss": 0.0004,
+      "step": 16350
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.029956161714564e-07,
+      "loss": 0.0002,
+      "step": 16351
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0297126156843642e-07,
+      "loss": 0.0002,
+      "step": 16352
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0294690696541646e-07,
+      "loss": 1.7842,
+      "step": 16353
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0292255236239649e-07,
+      "loss": 0.0002,
+      "step": 16354
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 1.0289819775937651e-07,
+      "loss": 0.0003,
+      "step": 16355
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0287384315635654e-07,
+      "loss": 0.0031,
+      "step": 16356
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0284948855333657e-07,
+      "loss": 0.0002,
+      "step": 16357
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0282513395031661e-07,
+      "loss": 0.0003,
+      "step": 16358
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0280077934729664e-07,
+      "loss": 0.0002,
+      "step": 16359
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0277642474427666e-07,
+      "loss": 0.0004,
+      "step": 16360
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.027520701412567e-07,
+      "loss": 0.0002,
+      "step": 16361
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0272771553823673e-07,
+      "loss": 0.0002,
+      "step": 16362
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0270336093521675e-07,
+      "loss": 0.0002,
+      "step": 16363
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0267900633219678e-07,
+      "loss": 0.0002,
+      "step": 16364
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0265465172917682e-07,
+      "loss": 0.0002,
+      "step": 16365
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 1.0263029712615684e-07,
+      "loss": 0.0004,
+      "step": 16366
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.0260594252313687e-07,
+      "loss": 0.0003,
+      "step": 16367
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.0258158792011689e-07,
+      "loss": 0.0002,
+      "step": 16368
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.0255723331709693e-07,
+      "loss": 0.0002,
+      "step": 16369
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.0253287871407695e-07,
+      "loss": 0.0001,
+      "step": 16370
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.0250852411105698e-07,
+      "loss": 0.0004,
+      "step": 16371
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.02484169508037e-07,
+      "loss": 0.0004,
+      "step": 16372
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.0245981490501704e-07,
+      "loss": 0.0003,
+      "step": 16373
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.0243546030199707e-07,
+      "loss": 0.0002,
+      "step": 16374
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.024111056989771e-07,
+      "loss": 0.0004,
+      "step": 16375
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 1.0238675109595712e-07,
+      "loss": 0.0002,
+      "step": 16376
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.0236239649293717e-07,
+      "loss": 0.0002,
+      "step": 16377
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.023380418899172e-07,
+      "loss": 0.0003,
+      "step": 16378
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.0231368728689722e-07,
+      "loss": 0.0001,
+      "step": 16379
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.0228933268387725e-07,
+      "loss": 0.0002,
+      "step": 16380
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.0226497808085728e-07,
+      "loss": 0.0002,
+      "step": 16381
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.0224062347783731e-07,
+      "loss": 0.0002,
+      "step": 16382
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.0221626887481733e-07,
+      "loss": 0.0002,
+      "step": 16383
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.0219191427179736e-07,
+      "loss": 0.0003,
+      "step": 16384
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.021675596687774e-07,
+      "loss": 0.0003,
+      "step": 16385
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 1.0214320506575742e-07,
+      "loss": 0.0003,
+      "step": 16386
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0211885046273745e-07,
+      "loss": 0.0004,
+      "step": 16387
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0209449585971747e-07,
+      "loss": 0.0004,
+      "step": 16388
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0207014125669751e-07,
+      "loss": 0.0004,
+      "step": 16389
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0204578665367754e-07,
+      "loss": 0.0003,
+      "step": 16390
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0202143205065756e-07,
+      "loss": 0.0003,
+      "step": 16391
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.019970774476376e-07,
+      "loss": 0.0003,
+      "step": 16392
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0197272284461763e-07,
+      "loss": 0.0003,
+      "step": 16393
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0194836824159765e-07,
+      "loss": 0.0002,
+      "step": 16394
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0192401363857768e-07,
+      "loss": 0.0002,
+      "step": 16395
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0189965903555773e-07,
+      "loss": 0.0002,
+      "step": 16396
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 1.0187530443253775e-07,
+      "loss": 0.0003,
+      "step": 16397
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.0185094982951778e-07,
+      "loss": 0.0002,
+      "step": 16398
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.018265952264978e-07,
+      "loss": 0.0002,
+      "step": 16399
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.0180224062347784e-07,
+      "loss": 0.0002,
+      "step": 16400
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.0177788602045787e-07,
+      "loss": 0.0002,
+      "step": 16401
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.0175353141743789e-07,
+      "loss": 0.0002,
+      "step": 16402
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.0172917681441792e-07,
+      "loss": 0.0002,
+      "step": 16403
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.0170482221139796e-07,
+      "loss": 0.0004,
+      "step": 16404
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.0168046760837798e-07,
+      "loss": 0.0002,
+      "step": 16405
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.01656113005358e-07,
+      "loss": 0.0004,
+      "step": 16406
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 1.0163175840233803e-07,
+      "loss": 0.0003,
+      "step": 16407
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0160740379931807e-07,
+      "loss": 0.0004,
+      "step": 16408
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.015830491962981e-07,
+      "loss": 0.0002,
+      "step": 16409
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0155869459327812e-07,
+      "loss": 0.0001,
+      "step": 16410
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0153433999025815e-07,
+      "loss": 0.0003,
+      "step": 16411
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0150998538723818e-07,
+      "loss": 0.0002,
+      "step": 16412
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0148563078421821e-07,
+      "loss": 0.0002,
+      "step": 16413
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0146127618119823e-07,
+      "loss": 0.0001,
+      "step": 16414
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0143692157817827e-07,
+      "loss": 0.0004,
+      "step": 16415
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0141256697515831e-07,
+      "loss": 0.0002,
+      "step": 16416
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 1.0138821237213834e-07,
+      "loss": 0.0003,
+      "step": 16417
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0136385776911836e-07,
+      "loss": 0.0001,
+      "step": 16418
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0133950316609839e-07,
+      "loss": 0.0003,
+      "step": 16419
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0131514856307842e-07,
+      "loss": 0.0002,
+      "step": 16420
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0129079396005845e-07,
+      "loss": 0.0003,
+      "step": 16421
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0126643935703848e-07,
+      "loss": 0.0003,
+      "step": 16422
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.012420847540185e-07,
+      "loss": 0.0001,
+      "step": 16423
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0121773015099854e-07,
+      "loss": 0.0002,
+      "step": 16424
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0119337554797856e-07,
+      "loss": 0.0002,
+      "step": 16425
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0116902094495859e-07,
+      "loss": 0.0002,
+      "step": 16426
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 1.0114466634193861e-07,
+      "loss": 0.0002,
+      "step": 16427
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0112031173891865e-07,
+      "loss": 0.0002,
+      "step": 16428
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0109595713589868e-07,
+      "loss": 0.0003,
+      "step": 16429
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.010716025328787e-07,
+      "loss": 0.0002,
+      "step": 16430
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0104724792985873e-07,
+      "loss": 0.0003,
+      "step": 16431
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0102289332683877e-07,
+      "loss": 0.0002,
+      "step": 16432
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0099853872381879e-07,
+      "loss": 0.0002,
+      "step": 16433
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0097418412079883e-07,
+      "loss": 0.0001,
+      "step": 16434
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0094982951777886e-07,
+      "loss": 0.0002,
+      "step": 16435
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.009254749147589e-07,
+      "loss": 0.0002,
+      "step": 16436
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0090112031173892e-07,
+      "loss": 0.0001,
+      "step": 16437
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 1.0087676570871894e-07,
+      "loss": 0.0002,
+      "step": 16438
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0085241110569897e-07,
+      "loss": 0.0002,
+      "step": 16439
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0082805650267901e-07,
+      "loss": 0.0002,
+      "step": 16440
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0080370189965903e-07,
+      "loss": 0.0002,
+      "step": 16441
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0077934729663906e-07,
+      "loss": 0.0001,
+      "step": 16442
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.007549926936191e-07,
+      "loss": 0.0002,
+      "step": 16443
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0073063809059912e-07,
+      "loss": 0.0002,
+      "step": 16444
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0070628348757915e-07,
+      "loss": 0.0002,
+      "step": 16445
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0068192888455917e-07,
+      "loss": 0.0004,
+      "step": 16446
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0065757428153921e-07,
+      "loss": 0.0002,
+      "step": 16447
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 1.0063321967851924e-07,
+      "loss": 0.0003,
+      "step": 16448
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0060886507549926e-07,
+      "loss": 0.0002,
+      "step": 16449
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0058451047247929e-07,
+      "loss": 2.0331,
+      "step": 16450
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0056015586945932e-07,
+      "loss": 0.0002,
+      "step": 16451
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0053580126643935e-07,
+      "loss": 0.0001,
+      "step": 16452
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0051144666341939e-07,
+      "loss": 0.0003,
+      "step": 16453
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0048709206039941e-07,
+      "loss": 0.0003,
+      "step": 16454
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0046273745737945e-07,
+      "loss": 0.0002,
+      "step": 16455
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0043838285435948e-07,
+      "loss": 0.0001,
+      "step": 16456
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.004140282513395e-07,
+      "loss": 0.0001,
+      "step": 16457
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 1.0038967364831953e-07,
+      "loss": 0.001,
+      "step": 16458
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0036531904529957e-07,
+      "loss": 0.0002,
+      "step": 16459
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0034096444227959e-07,
+      "loss": 0.0002,
+      "step": 16460
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0031660983925962e-07,
+      "loss": 0.0002,
+      "step": 16461
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0029225523623964e-07,
+      "loss": 0.0001,
+      "step": 16462
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0026790063321968e-07,
+      "loss": 0.0011,
+      "step": 16463
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.002435460301997e-07,
+      "loss": 0.0004,
+      "step": 16464
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.9985415653864852,
+      "eval_loss": 0.014664237387478352,
+      "eval_runtime": 2.7894,
+      "eval_samples_per_second": 737.445,
+      "eval_steps_per_second": 11.831,
+      "step": 16464
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0021919142717973e-07,
+      "loss": 0.0003,
+      "step": 16465
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0019483682415975e-07,
+      "loss": 0.0002,
+      "step": 16466
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0017048222113979e-07,
+      "loss": 0.0001,
+      "step": 16467
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0014612761811982e-07,
+      "loss": 0.0002,
+      "step": 16468
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 1.0012177301509984e-07,
+      "loss": 0.0003,
+      "step": 16469
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1.0009741841207987e-07,
+      "loss": 0.0003,
+      "step": 16470
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1.0007306380905991e-07,
+      "loss": 0.0003,
+      "step": 16471
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1.0004870920603993e-07,
+      "loss": 0.0002,
+      "step": 16472
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1.0002435460301997e-07,
+      "loss": 0.0002,
+      "step": 16473
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 1e-07,
+      "loss": 0.0002,
+      "step": 16474
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 9.997564539698003e-08,
+      "loss": 0.0003,
+      "step": 16475
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 9.995129079396006e-08,
+      "loss": 0.0004,
+      "step": 16476
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 9.992693619094008e-08,
+      "loss": 0.0002,
+      "step": 16477
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 9.990258158792011e-08,
+      "loss": 0.0002,
+      "step": 16478
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 9.987822698490015e-08,
+      "loss": 0.0002,
+      "step": 16479
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.985387238188017e-08,
+      "loss": 0.0003,
+      "step": 16480
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.98295177788602e-08,
+      "loss": 0.0002,
+      "step": 16481
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.980516317584022e-08,
+      "loss": 0.0002,
+      "step": 16482
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.978080857282026e-08,
+      "loss": 0.0002,
+      "step": 16483
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.975645396980029e-08,
+      "loss": 0.0002,
+      "step": 16484
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.973209936678031e-08,
+      "loss": 0.0001,
+      "step": 16485
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.970774476376034e-08,
+      "loss": 0.0002,
+      "step": 16486
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.968339016074038e-08,
+      "loss": 0.0002,
+      "step": 16487
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.96590355577204e-08,
+      "loss": 0.0003,
+      "step": 16488
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 9.963468095470043e-08,
+      "loss": 0.0003,
+      "step": 16489
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.961032635168045e-08,
+      "loss": 0.0002,
+      "step": 16490
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.958597174866049e-08,
+      "loss": 0.0002,
+      "step": 16491
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.956161714564053e-08,
+      "loss": 0.0002,
+      "step": 16492
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.953726254262055e-08,
+      "loss": 0.0001,
+      "step": 16493
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.951290793960059e-08,
+      "loss": 0.0002,
+      "step": 16494
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.948855333658062e-08,
+      "loss": 0.0004,
+      "step": 16495
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.946419873356064e-08,
+      "loss": 0.0003,
+      "step": 16496
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.943984413054067e-08,
+      "loss": 0.0003,
+      "step": 16497
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.94154895275207e-08,
+      "loss": 0.0003,
+      "step": 16498
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.939113492450073e-08,
+      "loss": 0.0002,
+      "step": 16499
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.936678032148076e-08,
+      "loss": 0.0001,
+      "step": 16500
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.934242571846078e-08,
+      "loss": 0.0002,
+      "step": 16501
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.931807111544082e-08,
+      "loss": 2.0709,
+      "step": 16502
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.929371651242084e-08,
+      "loss": 0.0001,
+      "step": 16503
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.926936190940087e-08,
+      "loss": 0.0004,
+      "step": 16504
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.92450073063809e-08,
+      "loss": 0.0004,
+      "step": 16505
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.922065270336093e-08,
+      "loss": 0.0002,
+      "step": 16506
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.919629810034096e-08,
+      "loss": 0.0002,
+      "step": 16507
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.917194349732098e-08,
+      "loss": 0.0003,
+      "step": 16508
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.914758889430101e-08,
+      "loss": 0.0002,
+      "step": 16509
+    },
+    {
+      "epoch": 16.04,
+      "learning_rate": 9.912323429128105e-08,
+      "loss": 0.0005,
+      "step": 16510
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.909887968826109e-08,
+      "loss": 0.0002,
+      "step": 16511
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.907452508524111e-08,
+      "loss": 0.0002,
+      "step": 16512
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.905017048222114e-08,
+      "loss": 0.0002,
+      "step": 16513
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.902581587920117e-08,
+      "loss": 0.0002,
+      "step": 16514
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.90014612761812e-08,
+      "loss": 0.0002,
+      "step": 16515
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.897710667316122e-08,
+      "loss": 0.0002,
+      "step": 16516
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.895275207014125e-08,
+      "loss": 0.0002,
+      "step": 16517
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.892839746712129e-08,
+      "loss": 0.0003,
+      "step": 16518
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.890404286410131e-08,
+      "loss": 0.0003,
+      "step": 16519
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.887968826108134e-08,
+      "loss": 0.0005,
+      "step": 16520
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.885533365806136e-08,
+      "loss": 0.0006,
+      "step": 16521
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.88309790550414e-08,
+      "loss": 0.0002,
+      "step": 16522
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.880662445202143e-08,
+      "loss": 0.0002,
+      "step": 16523
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.878226984900145e-08,
+      "loss": 0.0004,
+      "step": 16524
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.875791524598148e-08,
+      "loss": 0.0002,
+      "step": 16525
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.873356064296152e-08,
+      "loss": 0.0004,
+      "step": 16526
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.870920603994154e-08,
+      "loss": 0.0002,
+      "step": 16527
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.868485143692157e-08,
+      "loss": 0.0002,
+      "step": 16528
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.866049683390159e-08,
+      "loss": 0.0123,
+      "step": 16529
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 9.863614223088164e-08,
+      "loss": 0.0005,
+      "step": 16530
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.861178762786167e-08,
+      "loss": 0.0002,
+      "step": 16531
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.858743302484169e-08,
+      "loss": 0.0002,
+      "step": 16532
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.856307842182172e-08,
+      "loss": 0.0002,
+      "step": 16533
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.853872381880176e-08,
+      "loss": 0.0003,
+      "step": 16534
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.851436921578178e-08,
+      "loss": 0.0003,
+      "step": 16535
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.849001461276181e-08,
+      "loss": 0.0002,
+      "step": 16536
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.846566000974183e-08,
+      "loss": 0.0002,
+      "step": 16537
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.844130540672187e-08,
+      "loss": 0.0004,
+      "step": 16538
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.84169508037019e-08,
+      "loss": 0.0002,
+      "step": 16539
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.839259620068192e-08,
+      "loss": 0.0001,
+      "step": 16540
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 9.836824159766195e-08,
+      "loss": 0.0004,
+      "step": 16541
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.834388699464198e-08,
+      "loss": 0.0003,
+      "step": 16542
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.831953239162201e-08,
+      "loss": 0.0002,
+      "step": 16543
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.829517778860203e-08,
+      "loss": 0.0001,
+      "step": 16544
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.827082318558206e-08,
+      "loss": 0.0004,
+      "step": 16545
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.82464685825621e-08,
+      "loss": 0.0002,
+      "step": 16546
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.822211397954212e-08,
+      "loss": 0.0004,
+      "step": 16547
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.819775937652215e-08,
+      "loss": 0.0001,
+      "step": 16548
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.81734047735022e-08,
+      "loss": 0.0001,
+      "step": 16549
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.814905017048223e-08,
+      "loss": 0.0004,
+      "step": 16550
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.812469556746225e-08,
+      "loss": 0.0002,
+      "step": 16551
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.810034096444228e-08,
+      "loss": 0.0002,
+      "step": 16552
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.807598636142231e-08,
+      "loss": 0.0002,
+      "step": 16553
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.805163175840234e-08,
+      "loss": 0.0006,
+      "step": 16554
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.802727715538236e-08,
+      "loss": 0.0003,
+      "step": 16555
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.800292255236239e-08,
+      "loss": 0.0002,
+      "step": 16556
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.797856794934243e-08,
+      "loss": 0.0003,
+      "step": 16557
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.795421334632245e-08,
+      "loss": 0.0003,
+      "step": 16558
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.792985874330248e-08,
+      "loss": 0.0002,
+      "step": 16559
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.79055041402825e-08,
+      "loss": 0.0002,
+      "step": 16560
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 9.788114953726254e-08,
+      "loss": 0.0002,
+      "step": 16561
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.785679493424257e-08,
+      "loss": 0.0003,
+      "step": 16562
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.783244033122259e-08,
+      "loss": 0.0002,
+      "step": 16563
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.780808572820262e-08,
+      "loss": 0.0001,
+      "step": 16564
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.778373112518266e-08,
+      "loss": 0.0001,
+      "step": 16565
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.775937652216268e-08,
+      "loss": 0.0002,
+      "step": 16566
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.77350219191427e-08,
+      "loss": 0.0001,
+      "step": 16567
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.771066731612274e-08,
+      "loss": 0.0002,
+      "step": 16568
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.768631271310278e-08,
+      "loss": 0.0003,
+      "step": 16569
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.766195811008281e-08,
+      "loss": 0.0003,
+      "step": 16570
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.763760350706283e-08,
+      "loss": 0.0003,
+      "step": 16571
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 9.761324890404286e-08,
+      "loss": 0.0003,
+      "step": 16572
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.75888943010229e-08,
+      "loss": 0.0003,
+      "step": 16573
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.756453969800292e-08,
+      "loss": 0.0001,
+      "step": 16574
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.754018509498295e-08,
+      "loss": 0.0002,
+      "step": 16575
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.751583049196297e-08,
+      "loss": 0.0003,
+      "step": 16576
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.749147588894301e-08,
+      "loss": 0.0003,
+      "step": 16577
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.746712128592304e-08,
+      "loss": 0.0004,
+      "step": 16578
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.744276668290306e-08,
+      "loss": 0.0199,
+      "step": 16579
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.741841207988309e-08,
+      "loss": 0.0002,
+      "step": 16580
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.739405747686312e-08,
+      "loss": 0.0003,
+      "step": 16581
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.736970287384315e-08,
+      "loss": 0.0003,
+      "step": 16582
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.734534827082318e-08,
+      "loss": 0.0008,
+      "step": 16583
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.73209936678032e-08,
+      "loss": 0.0002,
+      "step": 16584
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.729663906478324e-08,
+      "loss": 0.0002,
+      "step": 16585
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.727228446176326e-08,
+      "loss": 0.0004,
+      "step": 16586
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.72479298587433e-08,
+      "loss": 0.0002,
+      "step": 16587
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.722357525572333e-08,
+      "loss": 0.0003,
+      "step": 16588
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.719922065270337e-08,
+      "loss": 0.0002,
+      "step": 16589
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.717486604968339e-08,
+      "loss": 0.0001,
+      "step": 16590
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.715051144666342e-08,
+      "loss": 0.0002,
+      "step": 16591
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 9.712615684364344e-08,
+      "loss": 0.0004,
+      "step": 16592
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.710180224062348e-08,
+      "loss": 0.0002,
+      "step": 16593
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.70774476376035e-08,
+      "loss": 0.0002,
+      "step": 16594
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.705309303458353e-08,
+      "loss": 0.0002,
+      "step": 16595
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.702873843156356e-08,
+      "loss": 0.0001,
+      "step": 16596
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.70043838285436e-08,
+      "loss": 0.0004,
+      "step": 16597
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.698002922552362e-08,
+      "loss": 0.0002,
+      "step": 16598
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.695567462250364e-08,
+      "loss": 0.0002,
+      "step": 16599
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.693132001948367e-08,
+      "loss": 0.0001,
+      "step": 16600
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.690696541646371e-08,
+      "loss": 0.0013,
+      "step": 16601
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.688261081344373e-08,
+      "loss": 0.0004,
+      "step": 16602
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.685825621042376e-08,
+      "loss": 0.0003,
+      "step": 16603
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.683390160740378e-08,
+      "loss": 0.0002,
+      "step": 16604
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.680954700438382e-08,
+      "loss": 0.0001,
+      "step": 16605
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.678519240136385e-08,
+      "loss": 0.0001,
+      "step": 16606
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.676083779834389e-08,
+      "loss": 0.0003,
+      "step": 16607
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.673648319532392e-08,
+      "loss": 0.0002,
+      "step": 16608
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.671212859230395e-08,
+      "loss": 0.0002,
+      "step": 16609
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.668777398928397e-08,
+      "loss": 0.0003,
+      "step": 16610
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.6663419386264e-08,
+      "loss": 0.0002,
+      "step": 16611
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.663906478324404e-08,
+      "loss": 0.0003,
+      "step": 16612
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 9.661471018022406e-08,
+      "loss": 0.0003,
+      "step": 16613
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.659035557720409e-08,
+      "loss": 0.0002,
+      "step": 16614
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.656600097418411e-08,
+      "loss": 0.0002,
+      "step": 16615
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.654164637116415e-08,
+      "loss": 0.0002,
+      "step": 16616
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.651729176814418e-08,
+      "loss": 0.0003,
+      "step": 16617
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.64929371651242e-08,
+      "loss": 0.0001,
+      "step": 16618
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.646858256210423e-08,
+      "loss": 0.0002,
+      "step": 16619
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.644422795908427e-08,
+      "loss": 0.0002,
+      "step": 16620
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.641987335606429e-08,
+      "loss": 0.0002,
+      "step": 16621
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.639551875304432e-08,
+      "loss": 0.0001,
+      "step": 16622
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 9.637116415002434e-08,
+      "loss": 0.0004,
+      "step": 16623
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.634680954700438e-08,
+      "loss": 0.0002,
+      "step": 16624
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.63224549439844e-08,
+      "loss": 0.0004,
+      "step": 16625
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.629810034096444e-08,
+      "loss": 0.0004,
+      "step": 16626
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.627374573794447e-08,
+      "loss": 0.0002,
+      "step": 16627
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.62493911349245e-08,
+      "loss": 0.0004,
+      "step": 16628
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.622503653190453e-08,
+      "loss": 0.0001,
+      "step": 16629
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.620068192888456e-08,
+      "loss": 0.0003,
+      "step": 16630
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.617632732586458e-08,
+      "loss": 0.0002,
+      "step": 16631
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.615197272284462e-08,
+      "loss": 0.0003,
+      "step": 16632
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.612761811982465e-08,
+      "loss": 0.0002,
+      "step": 16633
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.610326351680467e-08,
+      "loss": 0.0003,
+      "step": 16634
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.60789089137847e-08,
+      "loss": 0.0001,
+      "step": 16635
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.605455431076473e-08,
+      "loss": 0.0002,
+      "step": 16636
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.603019970774476e-08,
+      "loss": 0.0002,
+      "step": 16637
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.600584510472478e-08,
+      "loss": 0.0002,
+      "step": 16638
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.598149050170481e-08,
+      "loss": 0.0005,
+      "step": 16639
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.595713589868485e-08,
+      "loss": 0.0003,
+      "step": 16640
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.593278129566487e-08,
+      "loss": 0.0003,
+      "step": 16641
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.59084266926449e-08,
+      "loss": 0.0003,
+      "step": 16642
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.588407208962492e-08,
+      "loss": 0.0002,
+      "step": 16643
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 9.585971748660496e-08,
+      "loss": 0.0002,
+      "step": 16644
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.5835362883585e-08,
+      "loss": 0.0002,
+      "step": 16645
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.581100828056503e-08,
+      "loss": 0.0004,
+      "step": 16646
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.578665367754505e-08,
+      "loss": 0.0003,
+      "step": 16647
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.576229907452509e-08,
+      "loss": 0.0003,
+      "step": 16648
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.573794447150511e-08,
+      "loss": 0.0006,
+      "step": 16649
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.571358986848514e-08,
+      "loss": 0.0003,
+      "step": 16650
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.568923526546516e-08,
+      "loss": 0.0003,
+      "step": 16651
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.56648806624452e-08,
+      "loss": 0.0002,
+      "step": 16652
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.564052605942523e-08,
+      "loss": 0.0002,
+      "step": 16653
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 9.561617145640525e-08,
+      "loss": 0.0003,
+      "step": 16654
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.559181685338528e-08,
+      "loss": 0.0002,
+      "step": 16655
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.556746225036532e-08,
+      "loss": 0.0005,
+      "step": 16656
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.554310764734534e-08,
+      "loss": 0.0002,
+      "step": 16657
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.551875304432537e-08,
+      "loss": 0.0002,
+      "step": 16658
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.549439844130539e-08,
+      "loss": 0.0002,
+      "step": 16659
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.547004383828543e-08,
+      "loss": 0.0002,
+      "step": 16660
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.544568923526546e-08,
+      "loss": 0.0002,
+      "step": 16661
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.542133463224548e-08,
+      "loss": 0.0001,
+      "step": 16662
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.539698002922552e-08,
+      "loss": 0.0001,
+      "step": 16663
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.537262542620556e-08,
+      "loss": 0.0002,
+      "step": 16664
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.534827082318558e-08,
+      "loss": 0.0002,
+      "step": 16665
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.532391622016561e-08,
+      "loss": 0.0002,
+      "step": 16666
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.529956161714565e-08,
+      "loss": 0.0006,
+      "step": 16667
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.527520701412567e-08,
+      "loss": 0.0003,
+      "step": 16668
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.52508524111057e-08,
+      "loss": 0.0003,
+      "step": 16669
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.522649780808572e-08,
+      "loss": 0.0001,
+      "step": 16670
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.520214320506576e-08,
+      "loss": 0.0002,
+      "step": 16671
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.517778860204579e-08,
+      "loss": 0.0002,
+      "step": 16672
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.515343399902581e-08,
+      "loss": 0.0002,
+      "step": 16673
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 9.512907939600584e-08,
+      "loss": 0.0001,
+      "step": 16674
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.510472479298587e-08,
+      "loss": 0.0004,
+      "step": 16675
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.50803701899659e-08,
+      "loss": 0.0003,
+      "step": 16676
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.505601558694592e-08,
+      "loss": 0.0005,
+      "step": 16677
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.503166098392595e-08,
+      "loss": 0.0002,
+      "step": 16678
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.500730638090599e-08,
+      "loss": 0.0001,
+      "step": 16679
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.498295177788601e-08,
+      "loss": 0.0002,
+      "step": 16680
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.495859717486604e-08,
+      "loss": 0.0002,
+      "step": 16681
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.493424257184606e-08,
+      "loss": 0.0003,
+      "step": 16682
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.490988796882612e-08,
+      "loss": 0.0002,
+      "step": 16683
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.488553336580614e-08,
+      "loss": 0.0003,
+      "step": 16684
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.486117876278617e-08,
+      "loss": 0.0002,
+      "step": 16685
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.483682415976619e-08,
+      "loss": 0.0004,
+      "step": 16686
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.481246955674623e-08,
+      "loss": 0.0002,
+      "step": 16687
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.478811495372625e-08,
+      "loss": 0.0001,
+      "step": 16688
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.476376035070628e-08,
+      "loss": 0.0002,
+      "step": 16689
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.47394057476863e-08,
+      "loss": 0.0002,
+      "step": 16690
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.471505114466634e-08,
+      "loss": 0.0002,
+      "step": 16691
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.469069654164637e-08,
+      "loss": 0.0002,
+      "step": 16692
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.46663419386264e-08,
+      "loss": 0.0002,
+      "step": 16693
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.464198733560642e-08,
+      "loss": 0.0003,
+      "step": 16694
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 9.461763273258646e-08,
+      "loss": 0.0003,
+      "step": 16695
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.459327812956648e-08,
+      "loss": 0.0003,
+      "step": 16696
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.456892352654651e-08,
+      "loss": 0.0002,
+      "step": 16697
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.454456892352653e-08,
+      "loss": 0.0002,
+      "step": 16698
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.452021432050657e-08,
+      "loss": 0.0001,
+      "step": 16699
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.44958597174866e-08,
+      "loss": 0.0002,
+      "step": 16700
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.447150511446662e-08,
+      "loss": 0.0003,
+      "step": 16701
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.444715051144666e-08,
+      "loss": 0.0004,
+      "step": 16702
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.44227959084267e-08,
+      "loss": 0.0003,
+      "step": 16703
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.439844130540672e-08,
+      "loss": 0.0002,
+      "step": 16704
+    },
+    {
+      "epoch": 16.23,
+      "learning_rate": 9.437408670238675e-08,
+      "loss": 0.0002,
+      "step": 16705
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.434973209936677e-08,
+      "loss": 0.0002,
+      "step": 16706
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.432537749634681e-08,
+      "loss": 0.0002,
+      "step": 16707
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.430102289332684e-08,
+      "loss": 0.0003,
+      "step": 16708
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.427666829030686e-08,
+      "loss": 0.0002,
+      "step": 16709
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.42523136872869e-08,
+      "loss": 0.0002,
+      "step": 16710
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.422795908426693e-08,
+      "loss": 0.0003,
+      "step": 16711
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.420360448124695e-08,
+      "loss": 0.0003,
+      "step": 16712
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.417924987822698e-08,
+      "loss": 0.0001,
+      "step": 16713
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.415489527520701e-08,
+      "loss": 0.0003,
+      "step": 16714
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.413054067218704e-08,
+      "loss": 0.0002,
+      "step": 16715
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.410618606916706e-08,
+      "loss": 0.0002,
+      "step": 16716
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.408183146614709e-08,
+      "loss": 0.0004,
+      "step": 16717
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.405747686312713e-08,
+      "loss": 0.0002,
+      "step": 16718
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.403312226010715e-08,
+      "loss": 0.0002,
+      "step": 16719
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.400876765708718e-08,
+      "loss": 0.0002,
+      "step": 16720
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.39844130540672e-08,
+      "loss": 0.0003,
+      "step": 16721
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.396005845104726e-08,
+      "loss": 0.0002,
+      "step": 16722
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.393570384802728e-08,
+      "loss": 0.0003,
+      "step": 16723
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.39113492450073e-08,
+      "loss": 0.0004,
+      "step": 16724
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.388699464198733e-08,
+      "loss": 0.0002,
+      "step": 16725
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 9.386264003896737e-08,
+      "loss": 0.0003,
+      "step": 16726
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.38382854359474e-08,
+      "loss": 0.0003,
+      "step": 16727
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.381393083292742e-08,
+      "loss": 0.0003,
+      "step": 16728
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.378957622990744e-08,
+      "loss": 0.0002,
+      "step": 16729
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.376522162688748e-08,
+      "loss": 0.0003,
+      "step": 16730
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.374086702386751e-08,
+      "loss": 0.0002,
+      "step": 16731
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.371651242084753e-08,
+      "loss": 0.0003,
+      "step": 16732
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.369215781782756e-08,
+      "loss": 0.0004,
+      "step": 16733
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.36678032148076e-08,
+      "loss": 0.0002,
+      "step": 16734
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.364344861178762e-08,
+      "loss": 0.0001,
+      "step": 16735
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 9.361909400876765e-08,
+      "loss": 0.0004,
+      "step": 16736
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.359473940574767e-08,
+      "loss": 0.0003,
+      "step": 16737
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.357038480272771e-08,
+      "loss": 0.0003,
+      "step": 16738
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.354603019970774e-08,
+      "loss": 0.0002,
+      "step": 16739
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.352167559668776e-08,
+      "loss": 0.0001,
+      "step": 16740
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.34973209936678e-08,
+      "loss": 0.0004,
+      "step": 16741
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.347296639064784e-08,
+      "loss": 0.0004,
+      "step": 16742
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.344861178762786e-08,
+      "loss": 0.0002,
+      "step": 16743
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.342425718460789e-08,
+      "loss": 0.0002,
+      "step": 16744
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.339990258158791e-08,
+      "loss": 0.0001,
+      "step": 16745
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.337554797856795e-08,
+      "loss": 0.0003,
+      "step": 16746
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.335119337554798e-08,
+      "loss": 0.0002,
+      "step": 16747
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.3326838772528e-08,
+      "loss": 0.0002,
+      "step": 16748
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.330248416950803e-08,
+      "loss": 0.0025,
+      "step": 16749
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.327812956648807e-08,
+      "loss": 0.0004,
+      "step": 16750
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.325377496346809e-08,
+      "loss": 0.0002,
+      "step": 16751
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.322942036044812e-08,
+      "loss": 0.0002,
+      "step": 16752
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.320506575742814e-08,
+      "loss": 0.0003,
+      "step": 16753
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.318071115440818e-08,
+      "loss": 0.0003,
+      "step": 16754
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.31563565513882e-08,
+      "loss": 0.0003,
+      "step": 16755
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.313200194836823e-08,
+      "loss": 0.0002,
+      "step": 16756
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 9.310764734534826e-08,
+      "loss": 0.0002,
+      "step": 16757
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.30832927423283e-08,
+      "loss": 0.0006,
+      "step": 16758
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.305893813930832e-08,
+      "loss": 0.0001,
+      "step": 16759
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.303458353628836e-08,
+      "loss": 0.0002,
+      "step": 16760
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.30102289332684e-08,
+      "loss": 0.0002,
+      "step": 16761
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.298587433024842e-08,
+      "loss": 0.0003,
+      "step": 16762
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.296151972722845e-08,
+      "loss": 0.0002,
+      "step": 16763
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.293716512420847e-08,
+      "loss": 0.1054,
+      "step": 16764
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.291281052118851e-08,
+      "loss": 0.0001,
+      "step": 16765
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.288845591816853e-08,
+      "loss": 0.0003,
+      "step": 16766
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.286410131514856e-08,
+      "loss": 0.0002,
+      "step": 16767
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.283974671212859e-08,
+      "loss": 0.0002,
+      "step": 16768
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.281539210910862e-08,
+      "loss": 0.0002,
+      "step": 16769
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.279103750608865e-08,
+      "loss": 0.0002,
+      "step": 16770
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.276668290306867e-08,
+      "loss": 0.0003,
+      "step": 16771
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.27423283000487e-08,
+      "loss": 0.0002,
+      "step": 16772
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.271797369702874e-08,
+      "loss": 0.0003,
+      "step": 16773
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.269361909400876e-08,
+      "loss": 0.0003,
+      "step": 16774
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.266926449098879e-08,
+      "loss": 0.0002,
+      "step": 16775
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.264490988796881e-08,
+      "loss": 0.0002,
+      "step": 16776
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 9.262055528494885e-08,
+      "loss": 0.0003,
+      "step": 16777
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.259620068192888e-08,
+      "loss": 0.0001,
+      "step": 16778
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.257184607890891e-08,
+      "loss": 0.0003,
+      "step": 16779
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.254749147588894e-08,
+      "loss": 0.0002,
+      "step": 16780
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.252313687286898e-08,
+      "loss": 0.0003,
+      "step": 16781
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.2498782269849e-08,
+      "loss": 0.0002,
+      "step": 16782
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.247442766682903e-08,
+      "loss": 0.0001,
+      "step": 16783
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.245007306380905e-08,
+      "loss": 0.0002,
+      "step": 16784
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.242571846078909e-08,
+      "loss": 0.0002,
+      "step": 16785
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.240136385776912e-08,
+      "loss": 0.0002,
+      "step": 16786
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.237700925474914e-08,
+      "loss": 0.0001,
+      "step": 16787
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 9.235265465172917e-08,
+      "loss": 0.0002,
+      "step": 16788
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.23283000487092e-08,
+      "loss": 0.0002,
+      "step": 16789
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.230394544568923e-08,
+      "loss": 0.0002,
+      "step": 16790
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.227959084266926e-08,
+      "loss": 0.0002,
+      "step": 16791
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.225523623964928e-08,
+      "loss": 0.0002,
+      "step": 16792
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.223088163662932e-08,
+      "loss": 0.0004,
+      "step": 16793
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.220652703360935e-08,
+      "loss": 0.0002,
+      "step": 16794
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.218217243058937e-08,
+      "loss": 0.0003,
+      "step": 16795
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.21578178275694e-08,
+      "loss": 0.0002,
+      "step": 16796
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.213346322454943e-08,
+      "loss": 0.0003,
+      "step": 16797
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.210910862152947e-08,
+      "loss": 0.0001,
+      "step": 16798
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.20847540185095e-08,
+      "loss": 0.0001,
+      "step": 16799
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.206039941548952e-08,
+      "loss": 0.0004,
+      "step": 16800
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.203604481246956e-08,
+      "loss": 0.0003,
+      "step": 16801
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.201169020944959e-08,
+      "loss": 0.0002,
+      "step": 16802
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.198733560642961e-08,
+      "loss": 0.0004,
+      "step": 16803
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.196298100340964e-08,
+      "loss": 0.0002,
+      "step": 16804
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.193862640038968e-08,
+      "loss": 0.0004,
+      "step": 16805
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.19142717973697e-08,
+      "loss": 0.0003,
+      "step": 16806
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.188991719434973e-08,
+      "loss": 0.0002,
+      "step": 16807
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 9.186556259132975e-08,
+      "loss": 0.0002,
+      "step": 16808
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.184120798830979e-08,
+      "loss": 0.0003,
+      "step": 16809
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.181685338528981e-08,
+      "loss": 0.0003,
+      "step": 16810
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.179249878226984e-08,
+      "loss": 0.0003,
+      "step": 16811
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.176814417924986e-08,
+      "loss": 0.0003,
+      "step": 16812
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.17437895762299e-08,
+      "loss": 0.0002,
+      "step": 16813
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.171943497320993e-08,
+      "loss": 0.0002,
+      "step": 16814
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.169508037018995e-08,
+      "loss": 0.0002,
+      "step": 16815
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.167072576716998e-08,
+      "loss": 0.0001,
+      "step": 16816
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.164637116415003e-08,
+      "loss": 0.0002,
+      "step": 16817
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.162201656113006e-08,
+      "loss": 0.0004,
+      "step": 16818
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 9.159766195811008e-08,
+      "loss": 0.0002,
+      "step": 16819
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.157330735509012e-08,
+      "loss": 0.0004,
+      "step": 16820
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.154895275207014e-08,
+      "loss": 0.0005,
+      "step": 16821
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.152459814905017e-08,
+      "loss": 0.0002,
+      "step": 16822
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.15002435460302e-08,
+      "loss": 0.0002,
+      "step": 16823
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.147588894301023e-08,
+      "loss": 0.0002,
+      "step": 16824
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.145153433999026e-08,
+      "loss": 1.8796,
+      "step": 16825
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.142717973697028e-08,
+      "loss": 0.0003,
+      "step": 16826
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.140282513395031e-08,
+      "loss": 0.0002,
+      "step": 16827
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.137847053093035e-08,
+      "loss": 0.0003,
+      "step": 16828
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.135411592791037e-08,
+      "loss": 0.0002,
+      "step": 16829
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.13297613248904e-08,
+      "loss": 0.0002,
+      "step": 16830
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.130540672187042e-08,
+      "loss": 0.0001,
+      "step": 16831
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.128105211885046e-08,
+      "loss": 0.0001,
+      "step": 16832
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.125669751583049e-08,
+      "loss": 0.0002,
+      "step": 16833
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.123234291281051e-08,
+      "loss": 0.0002,
+      "step": 16834
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.120798830979054e-08,
+      "loss": 0.0003,
+      "step": 16835
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.118363370677059e-08,
+      "loss": 0.0002,
+      "step": 16836
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.115927910375061e-08,
+      "loss": 0.0004,
+      "step": 16837
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.113492450073064e-08,
+      "loss": 0.0004,
+      "step": 16838
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 9.111056989771066e-08,
+      "loss": 0.0001,
+      "step": 16839
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.10862152946907e-08,
+      "loss": 2.2771,
+      "step": 16840
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.106186069167073e-08,
+      "loss": 0.0004,
+      "step": 16841
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.103750608865075e-08,
+      "loss": 0.0002,
+      "step": 16842
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.101315148563078e-08,
+      "loss": 0.0003,
+      "step": 16843
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.098879688261082e-08,
+      "loss": 0.0001,
+      "step": 16844
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.096444227959084e-08,
+      "loss": 0.0004,
+      "step": 16845
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.094008767657087e-08,
+      "loss": 0.0002,
+      "step": 16846
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.091573307355089e-08,
+      "loss": 0.0001,
+      "step": 16847
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.089137847053093e-08,
+      "loss": 0.0003,
+      "step": 16848
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.086702386751095e-08,
+      "loss": 0.0006,
+      "step": 16849
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.084266926449098e-08,
+      "loss": 0.0003,
+      "step": 16850
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.0818314661471e-08,
+      "loss": 0.0003,
+      "step": 16851
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.079396005845104e-08,
+      "loss": 0.0003,
+      "step": 16852
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.076960545543107e-08,
+      "loss": 0.0003,
+      "step": 16853
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.07452508524111e-08,
+      "loss": 0.0001,
+      "step": 16854
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.072089624939112e-08,
+      "loss": 0.0002,
+      "step": 16855
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.069654164637117e-08,
+      "loss": 0.0004,
+      "step": 16856
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.06721870433512e-08,
+      "loss": 0.0002,
+      "step": 16857
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.064783244033122e-08,
+      "loss": 0.0002,
+      "step": 16858
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.062347783731125e-08,
+      "loss": 0.0002,
+      "step": 16859
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 9.059912323429128e-08,
+      "loss": 0.0002,
+      "step": 16860
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.057476863127131e-08,
+      "loss": 0.0002,
+      "step": 16861
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.055041402825133e-08,
+      "loss": 0.0002,
+      "step": 16862
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.052605942523136e-08,
+      "loss": 0.0002,
+      "step": 16863
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.05017048222114e-08,
+      "loss": 0.0004,
+      "step": 16864
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.047735021919142e-08,
+      "loss": 0.0003,
+      "step": 16865
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.045299561617145e-08,
+      "loss": 0.0003,
+      "step": 16866
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.042864101315147e-08,
+      "loss": 0.0003,
+      "step": 16867
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.040428641013151e-08,
+      "loss": 0.0003,
+      "step": 16868
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.037993180711154e-08,
+      "loss": 0.0003,
+      "step": 16869
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 9.035557720409156e-08,
+      "loss": 0.0003,
+      "step": 16870
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.033122260107159e-08,
+      "loss": 0.0002,
+      "step": 16871
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.030686799805163e-08,
+      "loss": 0.0004,
+      "step": 16872
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.028251339503165e-08,
+      "loss": 0.0006,
+      "step": 16873
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.025815879201168e-08,
+      "loss": 0.0002,
+      "step": 16874
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.023380418899173e-08,
+      "loss": 0.0003,
+      "step": 16875
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.020944958597175e-08,
+      "loss": 0.0002,
+      "step": 16876
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.018509498295178e-08,
+      "loss": 0.0003,
+      "step": 16877
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.01607403799318e-08,
+      "loss": 0.0003,
+      "step": 16878
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.013638577691184e-08,
+      "loss": 0.0001,
+      "step": 16879
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.011203117389187e-08,
+      "loss": 0.0002,
+      "step": 16880
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 9.008767657087189e-08,
+      "loss": 0.0002,
+      "step": 16881
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 9.006332196785192e-08,
+      "loss": 0.0001,
+      "step": 16882
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 9.003896736483196e-08,
+      "loss": 0.0001,
+      "step": 16883
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 9.001461276181198e-08,
+      "loss": 0.0002,
+      "step": 16884
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 8.9990258158792e-08,
+      "loss": 0.0002,
+      "step": 16885
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 8.996590355577203e-08,
+      "loss": 0.0003,
+      "step": 16886
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 8.994154895275207e-08,
+      "loss": 0.0003,
+      "step": 16887
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 8.99171943497321e-08,
+      "loss": 0.0005,
+      "step": 16888
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 8.989283974671212e-08,
+      "loss": 0.0003,
+      "step": 16889
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 8.986848514369215e-08,
+      "loss": 0.0002,
+      "step": 16890
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 8.984413054067218e-08,
+      "loss": 0.0002,
+      "step": 16891
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.981977593765221e-08,
+      "loss": 0.0003,
+      "step": 16892
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.979542133463223e-08,
+      "loss": 0.0003,
+      "step": 16893
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.977106673161227e-08,
+      "loss": 0.0007,
+      "step": 16894
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.974671212859231e-08,
+      "loss": 0.0002,
+      "step": 16895
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.972235752557234e-08,
+      "loss": 0.0002,
+      "step": 16896
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.969800292255236e-08,
+      "loss": 0.0002,
+      "step": 16897
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.967364831953239e-08,
+      "loss": 0.0002,
+      "step": 16898
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.964929371651242e-08,
+      "loss": 0.0003,
+      "step": 16899
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.962493911349245e-08,
+      "loss": 0.0002,
+      "step": 16900
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 8.960058451047247e-08,
+      "loss": 0.0003,
+      "step": 16901
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.95762299074525e-08,
+      "loss": 0.0002,
+      "step": 16902
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.955187530443254e-08,
+      "loss": 0.0003,
+      "step": 16903
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.952752070141256e-08,
+      "loss": 0.0003,
+      "step": 16904
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.950316609839259e-08,
+      "loss": 0.0004,
+      "step": 16905
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.947881149537261e-08,
+      "loss": 0.0003,
+      "step": 16906
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.945445689235265e-08,
+      "loss": 0.0001,
+      "step": 16907
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.943010228933268e-08,
+      "loss": 0.0002,
+      "step": 16908
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.94057476863127e-08,
+      "loss": 0.0003,
+      "step": 16909
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.938139308329273e-08,
+      "loss": 0.0001,
+      "step": 16910
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 8.935703848027277e-08,
+      "loss": 0.0003,
+      "step": 16911
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.933268387725279e-08,
+      "loss": 0.0004,
+      "step": 16912
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.930832927423283e-08,
+      "loss": 0.0002,
+      "step": 16913
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.928397467121285e-08,
+      "loss": 0.0002,
+      "step": 16914
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.925962006819289e-08,
+      "loss": 0.0005,
+      "step": 16915
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.923526546517292e-08,
+      "loss": 0.0006,
+      "step": 16916
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.921091086215294e-08,
+      "loss": 0.0002,
+      "step": 16917
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.918655625913297e-08,
+      "loss": 0.0002,
+      "step": 16918
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.916220165611301e-08,
+      "loss": 0.0004,
+      "step": 16919
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.913784705309303e-08,
+      "loss": 0.0002,
+      "step": 16920
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 8.911349245007306e-08,
+      "loss": 0.0002,
+      "step": 16921
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.908913784705308e-08,
+      "loss": 0.0004,
+      "step": 16922
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.906478324403312e-08,
+      "loss": 0.0003,
+      "step": 16923
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.904042864101315e-08,
+      "loss": 0.0002,
+      "step": 16924
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.901607403799317e-08,
+      "loss": 0.0002,
+      "step": 16925
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.89917194349732e-08,
+      "loss": 0.0002,
+      "step": 16926
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.896736483195323e-08,
+      "loss": 0.0004,
+      "step": 16927
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.894301022893326e-08,
+      "loss": 0.001,
+      "step": 16928
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.891865562591329e-08,
+      "loss": 0.0003,
+      "step": 16929
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.889430102289332e-08,
+      "loss": 0.0003,
+      "step": 16930
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.886994641987335e-08,
+      "loss": 0.0002,
+      "step": 16931
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 8.884559181685339e-08,
+      "loss": 0.0002,
+      "step": 16932
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.882123721383341e-08,
+      "loss": 0.0002,
+      "step": 16933
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.879688261081345e-08,
+      "loss": 0.0001,
+      "step": 16934
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.877252800779348e-08,
+      "loss": 0.0002,
+      "step": 16935
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.87481734047735e-08,
+      "loss": 0.0002,
+      "step": 16936
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.872381880175353e-08,
+      "loss": 0.0002,
+      "step": 16937
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.869946419873356e-08,
+      "loss": 0.0003,
+      "step": 16938
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.867510959571359e-08,
+      "loss": 0.0003,
+      "step": 16939
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.865075499269362e-08,
+      "loss": 0.0003,
+      "step": 16940
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.862640038967364e-08,
+      "loss": 0.0001,
+      "step": 16941
+    },
+    {
+      "epoch": 16.46,
+      "learning_rate": 8.860204578665368e-08,
+      "loss": 0.0003,
+      "step": 16942
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.85776911836337e-08,
+      "loss": 0.0003,
+      "step": 16943
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.855333658061373e-08,
+      "loss": 0.0002,
+      "step": 16944
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.852898197759375e-08,
+      "loss": 0.0002,
+      "step": 16945
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.850462737457379e-08,
+      "loss": 0.0002,
+      "step": 16946
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.848027277155382e-08,
+      "loss": 0.0003,
+      "step": 16947
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.845591816853384e-08,
+      "loss": 0.0002,
+      "step": 16948
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.843156356551387e-08,
+      "loss": 0.0002,
+      "step": 16949
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.84072089624939e-08,
+      "loss": 0.0002,
+      "step": 16950
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.838285435947394e-08,
+      "loss": 0.0001,
+      "step": 16951
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 8.835849975645397e-08,
+      "loss": 0.0002,
+      "step": 16952
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.8334145153434e-08,
+      "loss": 0.0003,
+      "step": 16953
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.830979055041403e-08,
+      "loss": 0.0004,
+      "step": 16954
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.828543594739406e-08,
+      "loss": 0.0002,
+      "step": 16955
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.826108134437408e-08,
+      "loss": 0.0002,
+      "step": 16956
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.823672674135411e-08,
+      "loss": 0.0003,
+      "step": 16957
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.821237213833415e-08,
+      "loss": 0.0002,
+      "step": 16958
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.818801753531417e-08,
+      "loss": 0.0002,
+      "step": 16959
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.81636629322942e-08,
+      "loss": 0.0003,
+      "step": 16960
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.813930832927422e-08,
+      "loss": 0.0002,
+      "step": 16961
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.811495372625426e-08,
+      "loss": 0.0006,
+      "step": 16962
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 8.809059912323429e-08,
+      "loss": 0.0003,
+      "step": 16963
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.806624452021431e-08,
+      "loss": 0.0003,
+      "step": 16964
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.804188991719434e-08,
+      "loss": 0.0003,
+      "step": 16965
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.801753531417438e-08,
+      "loss": 0.0003,
+      "step": 16966
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.79931807111544e-08,
+      "loss": 0.0002,
+      "step": 16967
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.796882610813443e-08,
+      "loss": 0.0002,
+      "step": 16968
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.794447150511445e-08,
+      "loss": 0.0002,
+      "step": 16969
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.792011690209449e-08,
+      "loss": 0.3922,
+      "step": 16970
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.789576229907453e-08,
+      "loss": 0.0002,
+      "step": 16971
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.787140769605455e-08,
+      "loss": 0.0002,
+      "step": 16972
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 8.784705309303458e-08,
+      "loss": 0.0003,
+      "step": 16973
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.782269849001462e-08,
+      "loss": 0.0003,
+      "step": 16974
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.779834388699464e-08,
+      "loss": 0.0002,
+      "step": 16975
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.777398928397467e-08,
+      "loss": 0.0003,
+      "step": 16976
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.774963468095469e-08,
+      "loss": 0.0002,
+      "step": 16977
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.772528007793473e-08,
+      "loss": 0.0003,
+      "step": 16978
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.770092547491476e-08,
+      "loss": 0.0003,
+      "step": 16979
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.767657087189478e-08,
+      "loss": 0.0006,
+      "step": 16980
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.765221626887482e-08,
+      "loss": 0.0001,
+      "step": 16981
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.762786166585484e-08,
+      "loss": 0.0002,
+      "step": 16982
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 8.760350706283487e-08,
+      "loss": 0.0002,
+      "step": 16983
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.75791524598149e-08,
+      "loss": 0.0002,
+      "step": 16984
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.755479785679493e-08,
+      "loss": 0.0003,
+      "step": 16985
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.753044325377496e-08,
+      "loss": 0.0002,
+      "step": 16986
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.750608865075498e-08,
+      "loss": 0.0002,
+      "step": 16987
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.748173404773501e-08,
+      "loss": 0.0001,
+      "step": 16988
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.745737944471505e-08,
+      "loss": 0.0002,
+      "step": 16989
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.743302484169509e-08,
+      "loss": 0.0004,
+      "step": 16990
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.740867023867511e-08,
+      "loss": 0.0003,
+      "step": 16991
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.738431563565514e-08,
+      "loss": 0.0001,
+      "step": 16992
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 8.735996103263517e-08,
+      "loss": 0.0002,
+      "step": 16993
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.73356064296152e-08,
+      "loss": 0.0002,
+      "step": 16994
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.731125182659522e-08,
+      "loss": 0.0002,
+      "step": 16995
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.728689722357525e-08,
+      "loss": 0.0003,
+      "step": 16996
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.726254262055529e-08,
+      "loss": 0.0002,
+      "step": 16997
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.723818801753531e-08,
+      "loss": 0.0002,
+      "step": 16998
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.721383341451534e-08,
+      "loss": 0.0003,
+      "step": 16999
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.718947881149536e-08,
+      "loss": 0.0003,
+      "step": 17000
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.71651242084754e-08,
+      "loss": 0.0004,
+      "step": 17001
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.714076960545543e-08,
+      "loss": 0.0004,
+      "step": 17002
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.711641500243545e-08,
+      "loss": 0.0003,
+      "step": 17003
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 8.709206039941548e-08,
+      "loss": 0.0004,
+      "step": 17004
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.706770579639552e-08,
+      "loss": 0.0003,
+      "step": 17005
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.704335119337554e-08,
+      "loss": 0.0002,
+      "step": 17006
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.701899659035557e-08,
+      "loss": 0.0004,
+      "step": 17007
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.699464198733559e-08,
+      "loss": 0.0003,
+      "step": 17008
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.697028738431564e-08,
+      "loss": 0.0001,
+      "step": 17009
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.694593278129567e-08,
+      "loss": 0.0002,
+      "step": 17010
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.692157817827569e-08,
+      "loss": 0.0003,
+      "step": 17011
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.689722357525572e-08,
+      "loss": 0.0002,
+      "step": 17012
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.687286897223576e-08,
+      "loss": 0.0003,
+      "step": 17013
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 8.684851436921578e-08,
+      "loss": 0.0001,
+      "step": 17014
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.682415976619581e-08,
+      "loss": 0.0002,
+      "step": 17015
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.679980516317583e-08,
+      "loss": 0.0003,
+      "step": 17016
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.677545056015587e-08,
+      "loss": 0.0106,
+      "step": 17017
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.67510959571359e-08,
+      "loss": 0.0003,
+      "step": 17018
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.672674135411592e-08,
+      "loss": 0.0001,
+      "step": 17019
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.670238675109595e-08,
+      "loss": 0.0001,
+      "step": 17020
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.667803214807598e-08,
+      "loss": 0.0002,
+      "step": 17021
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.665367754505601e-08,
+      "loss": 0.0002,
+      "step": 17022
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.662932294203603e-08,
+      "loss": 0.0002,
+      "step": 17023
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 8.660496833901606e-08,
+      "loss": 0.0003,
+      "step": 17024
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.65806137359961e-08,
+      "loss": 0.0002,
+      "step": 17025
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.655625913297612e-08,
+      "loss": 0.0002,
+      "step": 17026
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.653190452995615e-08,
+      "loss": 0.0002,
+      "step": 17027
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.650754992693619e-08,
+      "loss": 0.0002,
+      "step": 17028
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.648319532391623e-08,
+      "loss": 0.0002,
+      "step": 17029
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.645884072089625e-08,
+      "loss": 0.0002,
+      "step": 17030
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.643448611787628e-08,
+      "loss": 0.0002,
+      "step": 17031
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.641013151485631e-08,
+      "loss": 0.0002,
+      "step": 17032
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.638577691183634e-08,
+      "loss": 0.0003,
+      "step": 17033
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.636142230881636e-08,
+      "loss": 0.0008,
+      "step": 17034
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 8.633706770579639e-08,
+      "loss": 0.0003,
+      "step": 17035
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.631271310277643e-08,
+      "loss": 0.0003,
+      "step": 17036
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.628835849975645e-08,
+      "loss": 0.0003,
+      "step": 17037
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.626400389673648e-08,
+      "loss": 0.0002,
+      "step": 17038
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.62396492937165e-08,
+      "loss": 0.0002,
+      "step": 17039
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.621529469069654e-08,
+      "loss": 0.0002,
+      "step": 17040
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.619094008767657e-08,
+      "loss": 0.0002,
+      "step": 17041
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.616658548465659e-08,
+      "loss": 0.0002,
+      "step": 17042
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.614223088163662e-08,
+      "loss": 0.0002,
+      "step": 17043
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.611787627861666e-08,
+      "loss": 0.0001,
+      "step": 17044
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 8.609352167559668e-08,
+      "loss": 0.0004,
+      "step": 17045
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.60691670725767e-08,
+      "loss": 0.0001,
+      "step": 17046
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.604481246955674e-08,
+      "loss": 0.0002,
+      "step": 17047
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.602045786653678e-08,
+      "loss": 0.0003,
+      "step": 17048
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.599610326351681e-08,
+      "loss": 0.0002,
+      "step": 17049
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.597174866049683e-08,
+      "loss": 0.0002,
+      "step": 17050
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.594739405747686e-08,
+      "loss": 0.0003,
+      "step": 17051
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.59230394544569e-08,
+      "loss": 0.0003,
+      "step": 17052
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.589868485143692e-08,
+      "loss": 0.0003,
+      "step": 17053
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.587433024841695e-08,
+      "loss": 0.0002,
+      "step": 17054
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 8.584997564539697e-08,
+      "loss": 0.0002,
+      "step": 17055
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.582562104237701e-08,
+      "loss": 0.0003,
+      "step": 17056
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.580126643935704e-08,
+      "loss": 0.0003,
+      "step": 17057
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.577691183633706e-08,
+      "loss": 0.0004,
+      "step": 17058
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.575255723331709e-08,
+      "loss": 0.0003,
+      "step": 17059
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.572820263029712e-08,
+      "loss": 0.0002,
+      "step": 17060
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.570384802727715e-08,
+      "loss": 0.0001,
+      "step": 17061
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.567949342425717e-08,
+      "loss": 0.0001,
+      "step": 17062
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.56551388212372e-08,
+      "loss": 0.0002,
+      "step": 17063
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.563078421821724e-08,
+      "loss": 0.0001,
+      "step": 17064
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 8.560642961519726e-08,
+      "loss": 0.0001,
+      "step": 17065
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.55820750121773e-08,
+      "loss": 0.0002,
+      "step": 17066
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.555772040915733e-08,
+      "loss": 0.0004,
+      "step": 17067
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.553336580613737e-08,
+      "loss": 0.0003,
+      "step": 17068
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.550901120311739e-08,
+      "loss": 0.0003,
+      "step": 17069
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.548465660009742e-08,
+      "loss": 2.5643,
+      "step": 17070
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.546030199707744e-08,
+      "loss": 0.0002,
+      "step": 17071
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.543594739405748e-08,
+      "loss": 0.0002,
+      "step": 17072
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.54115927910375e-08,
+      "loss": 0.0003,
+      "step": 17073
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.538723818801753e-08,
+      "loss": 0.0003,
+      "step": 17074
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.536288358499756e-08,
+      "loss": 0.0003,
+      "step": 17075
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 8.53385289819776e-08,
+      "loss": 0.0002,
+      "step": 17076
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.531417437895762e-08,
+      "loss": 0.0002,
+      "step": 17077
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.528981977593764e-08,
+      "loss": 0.0004,
+      "step": 17078
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.526546517291767e-08,
+      "loss": 0.0003,
+      "step": 17079
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.524111056989771e-08,
+      "loss": 0.0001,
+      "step": 17080
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.521675596687773e-08,
+      "loss": 0.0003,
+      "step": 17081
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.519240136385776e-08,
+      "loss": 0.0002,
+      "step": 17082
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.516804676083778e-08,
+      "loss": 0.0008,
+      "step": 17083
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.514369215781782e-08,
+      "loss": 0.0003,
+      "step": 17084
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.511933755479786e-08,
+      "loss": 0.0002,
+      "step": 17085
+    },
+    {
+      "epoch": 16.6,
+      "learning_rate": 8.509498295177788e-08,
+      "loss": 0.0003,
+      "step": 17086
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.507062834875792e-08,
+      "loss": 0.0004,
+      "step": 17087
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.504627374573795e-08,
+      "loss": 0.0003,
+      "step": 17088
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.502191914271797e-08,
+      "loss": 0.0002,
+      "step": 17089
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.4997564539698e-08,
+      "loss": 0.0003,
+      "step": 17090
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.497320993667804e-08,
+      "loss": 0.0001,
+      "step": 17091
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.494885533365806e-08,
+      "loss": 0.0003,
+      "step": 17092
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.492450073063809e-08,
+      "loss": 0.0002,
+      "step": 17093
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.490014612761811e-08,
+      "loss": 0.0005,
+      "step": 17094
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.487579152459815e-08,
+      "loss": 0.0001,
+      "step": 17095
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 8.485143692157818e-08,
+      "loss": 0.001,
+      "step": 17096
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.48270823185582e-08,
+      "loss": 0.0002,
+      "step": 17097
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.480272771553823e-08,
+      "loss": 0.0003,
+      "step": 17098
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.477837311251826e-08,
+      "loss": 0.0004,
+      "step": 17099
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.475401850949829e-08,
+      "loss": 0.0002,
+      "step": 17100
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.472966390647832e-08,
+      "loss": 0.0002,
+      "step": 17101
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.470530930345834e-08,
+      "loss": 0.0003,
+      "step": 17102
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.468095470043838e-08,
+      "loss": 0.0003,
+      "step": 17103
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.46566000974184e-08,
+      "loss": 0.0002,
+      "step": 17104
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.463224549439844e-08,
+      "loss": 0.0002,
+      "step": 17105
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.460789089137847e-08,
+      "loss": 0.0001,
+      "step": 17106
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 8.45835362883585e-08,
+      "loss": 0.0005,
+      "step": 17107
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.455918168533853e-08,
+      "loss": 0.0002,
+      "step": 17108
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.453482708231856e-08,
+      "loss": 0.0002,
+      "step": 17109
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.451047247929858e-08,
+      "loss": 0.0003,
+      "step": 17110
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.448611787627862e-08,
+      "loss": 0.0003,
+      "step": 17111
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.446176327325864e-08,
+      "loss": 0.0001,
+      "step": 17112
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.443740867023867e-08,
+      "loss": 0.0003,
+      "step": 17113
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.44130540672187e-08,
+      "loss": 0.0004,
+      "step": 17114
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.438869946419873e-08,
+      "loss": 0.0003,
+      "step": 17115
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.436434486117876e-08,
+      "loss": 0.0002,
+      "step": 17116
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 8.433999025815878e-08,
+      "loss": 0.0002,
+      "step": 17117
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.431563565513881e-08,
+      "loss": 0.0002,
+      "step": 17118
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.429128105211885e-08,
+      "loss": 0.0002,
+      "step": 17119
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.426692644909887e-08,
+      "loss": 0.0002,
+      "step": 17120
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.42425718460789e-08,
+      "loss": 0.0003,
+      "step": 17121
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.421821724305892e-08,
+      "loss": 0.0003,
+      "step": 17122
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.419386264003896e-08,
+      "loss": 0.0003,
+      "step": 17123
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.4169508037019e-08,
+      "loss": 0.0003,
+      "step": 17124
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.414515343399903e-08,
+      "loss": 0.0001,
+      "step": 17125
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.412079883097905e-08,
+      "loss": 0.0003,
+      "step": 17126
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 8.409644422795909e-08,
+      "loss": 0.0002,
+      "step": 17127
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.407208962493911e-08,
+      "loss": 0.0002,
+      "step": 17128
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.404773502191914e-08,
+      "loss": 0.0002,
+      "step": 17129
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.402338041889916e-08,
+      "loss": 0.0003,
+      "step": 17130
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.39990258158792e-08,
+      "loss": 0.0003,
+      "step": 17131
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.397467121285923e-08,
+      "loss": 0.0002,
+      "step": 17132
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.395031660983925e-08,
+      "loss": 0.0003,
+      "step": 17133
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.392596200681928e-08,
+      "loss": 0.0032,
+      "step": 17134
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.390160740379932e-08,
+      "loss": 0.0002,
+      "step": 17135
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.387725280077934e-08,
+      "loss": 0.0003,
+      "step": 17136
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 8.385289819775937e-08,
+      "loss": 0.0002,
+      "step": 17137
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.382854359473939e-08,
+      "loss": 1.783,
+      "step": 17138
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.380418899171943e-08,
+      "loss": 0.0003,
+      "step": 17139
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.377983438869946e-08,
+      "loss": 0.0002,
+      "step": 17140
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.375547978567948e-08,
+      "loss": 0.0002,
+      "step": 17141
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.37311251826595e-08,
+      "loss": 0.0003,
+      "step": 17142
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.370677057963956e-08,
+      "loss": 0.0004,
+      "step": 17143
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.368241597661958e-08,
+      "loss": 0.0003,
+      "step": 17144
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.365806137359961e-08,
+      "loss": 0.0003,
+      "step": 17145
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.363370677057965e-08,
+      "loss": 0.0001,
+      "step": 17146
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.360935216755967e-08,
+      "loss": 0.0003,
+      "step": 17147
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 8.35849975645397e-08,
+      "loss": 0.0003,
+      "step": 17148
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.356064296151972e-08,
+      "loss": 0.0002,
+      "step": 17149
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.353628835849976e-08,
+      "loss": 0.0003,
+      "step": 17150
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.351193375547979e-08,
+      "loss": 0.0002,
+      "step": 17151
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.348757915245981e-08,
+      "loss": 0.0001,
+      "step": 17152
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.346322454943984e-08,
+      "loss": 0.0002,
+      "step": 17153
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.343886994641987e-08,
+      "loss": 0.0003,
+      "step": 17154
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.34145153433999e-08,
+      "loss": 0.0004,
+      "step": 17155
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.339016074037992e-08,
+      "loss": 0.0002,
+      "step": 17156
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.336580613735995e-08,
+      "loss": 0.0003,
+      "step": 17157
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 8.334145153433999e-08,
+      "loss": 0.0002,
+      "step": 17158
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.331709693132001e-08,
+      "loss": 0.0004,
+      "step": 17159
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.329274232830004e-08,
+      "loss": 0.0002,
+      "step": 17160
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.326838772528006e-08,
+      "loss": 0.0003,
+      "step": 17161
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.324403312226011e-08,
+      "loss": 0.0002,
+      "step": 17162
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.321967851924014e-08,
+      "loss": 0.0003,
+      "step": 17163
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.319532391622017e-08,
+      "loss": 0.0002,
+      "step": 17164
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.317096931320019e-08,
+      "loss": 0.0001,
+      "step": 17165
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.314661471018023e-08,
+      "loss": 0.0003,
+      "step": 17166
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.312226010716025e-08,
+      "loss": 0.0003,
+      "step": 17167
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 8.309790550414028e-08,
+      "loss": 0.0002,
+      "step": 17168
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.30735509011203e-08,
+      "loss": 0.0003,
+      "step": 17169
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.304919629810034e-08,
+      "loss": 0.0004,
+      "step": 17170
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.302484169508037e-08,
+      "loss": 0.0002,
+      "step": 17171
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.300048709206039e-08,
+      "loss": 0.0002,
+      "step": 17172
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.297613248904042e-08,
+      "loss": 0.0002,
+      "step": 17173
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.295177788602046e-08,
+      "loss": 0.0001,
+      "step": 17174
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.292742328300048e-08,
+      "loss": 0.0003,
+      "step": 17175
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.290306867998051e-08,
+      "loss": 0.0003,
+      "step": 17176
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.287871407696053e-08,
+      "loss": 0.0003,
+      "step": 17177
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.285435947394057e-08,
+      "loss": 0.0003,
+      "step": 17178
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 8.28300048709206e-08,
+      "loss": 0.0003,
+      "step": 17179
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.280565026790062e-08,
+      "loss": 0.0001,
+      "step": 17180
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.278129566488066e-08,
+      "loss": 0.0002,
+      "step": 17181
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.27569410618607e-08,
+      "loss": 0.0001,
+      "step": 17182
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.273258645884072e-08,
+      "loss": 0.0001,
+      "step": 17183
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.270823185582075e-08,
+      "loss": 0.0002,
+      "step": 17184
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.268387725280077e-08,
+      "loss": 0.0002,
+      "step": 17185
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.265952264978081e-08,
+      "loss": 0.0002,
+      "step": 17186
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.263516804676084e-08,
+      "loss": 0.0001,
+      "step": 17187
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.261081344374086e-08,
+      "loss": 0.0004,
+      "step": 17188
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 8.258645884072089e-08,
+      "loss": 0.0001,
+      "step": 17189
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.256210423770093e-08,
+      "loss": 0.0002,
+      "step": 17190
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.253774963468095e-08,
+      "loss": 0.0002,
+      "step": 17191
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.251339503166098e-08,
+      "loss": 0.0002,
+      "step": 17192
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.2489040428641e-08,
+      "loss": 0.0002,
+      "step": 17193
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.246468582562104e-08,
+      "loss": 0.0002,
+      "step": 17194
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.244033122260106e-08,
+      "loss": 0.0002,
+      "step": 17195
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.241597661958109e-08,
+      "loss": 0.0002,
+      "step": 17196
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.239162201656111e-08,
+      "loss": 0.0002,
+      "step": 17197
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.236726741354115e-08,
+      "loss": 0.0002,
+      "step": 17198
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 8.234291281052118e-08,
+      "loss": 0.0002,
+      "step": 17199
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.231855820750122e-08,
+      "loss": 0.0005,
+      "step": 17200
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.229420360448126e-08,
+      "loss": 0.0002,
+      "step": 17201
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.226984900146128e-08,
+      "loss": 0.0002,
+      "step": 17202
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.22454943984413e-08,
+      "loss": 0.0003,
+      "step": 17203
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.222113979542133e-08,
+      "loss": 0.0003,
+      "step": 17204
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.219678519240137e-08,
+      "loss": 0.0002,
+      "step": 17205
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.21724305893814e-08,
+      "loss": 0.0004,
+      "step": 17206
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.214807598636142e-08,
+      "loss": 0.0002,
+      "step": 17207
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.212372138334144e-08,
+      "loss": 0.0003,
+      "step": 17208
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.209936678032148e-08,
+      "loss": 0.0002,
+      "step": 17209
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 8.207501217730151e-08,
+      "loss": 0.0002,
+      "step": 17210
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.205065757428153e-08,
+      "loss": 0.0002,
+      "step": 17211
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.202630297126156e-08,
+      "loss": 0.0001,
+      "step": 17212
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.20019483682416e-08,
+      "loss": 0.0002,
+      "step": 17213
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.197759376522162e-08,
+      "loss": 0.0002,
+      "step": 17214
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.195323916220165e-08,
+      "loss": 0.0003,
+      "step": 17215
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.192888455918167e-08,
+      "loss": 0.0003,
+      "step": 17216
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.190452995616171e-08,
+      "loss": 0.0002,
+      "step": 17217
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.188017535314174e-08,
+      "loss": 0.0003,
+      "step": 17218
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.185582075012177e-08,
+      "loss": 0.0002,
+      "step": 17219
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 8.18314661471018e-08,
+      "loss": 0.0004,
+      "step": 17220
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.180711154408184e-08,
+      "loss": 0.0004,
+      "step": 17221
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.178275694106186e-08,
+      "loss": 0.0002,
+      "step": 17222
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.175840233804189e-08,
+      "loss": 0.0002,
+      "step": 17223
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.173404773502191e-08,
+      "loss": 0.0003,
+      "step": 17224
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.170969313200195e-08,
+      "loss": 0.0002,
+      "step": 17225
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.168533852898198e-08,
+      "loss": 0.0003,
+      "step": 17226
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.1660983925962e-08,
+      "loss": 0.0002,
+      "step": 17227
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.163662932294203e-08,
+      "loss": 0.0002,
+      "step": 17228
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.161227471992207e-08,
+      "loss": 0.0003,
+      "step": 17229
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 8.158792011690209e-08,
+      "loss": 0.0002,
+      "step": 17230
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.156356551388212e-08,
+      "loss": 0.0003,
+      "step": 17231
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.153921091086214e-08,
+      "loss": 0.0003,
+      "step": 17232
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.151485630784218e-08,
+      "loss": 0.0003,
+      "step": 17233
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.14905017048222e-08,
+      "loss": 0.0005,
+      "step": 17234
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.146614710180223e-08,
+      "loss": 0.0002,
+      "step": 17235
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.144179249878226e-08,
+      "loss": 0.0004,
+      "step": 17236
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.14174378957623e-08,
+      "loss": 0.0002,
+      "step": 17237
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.139308329274232e-08,
+      "loss": 0.0002,
+      "step": 17238
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.136872868972236e-08,
+      "loss": 0.0001,
+      "step": 17239
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 8.134437408670238e-08,
+      "loss": 0.0003,
+      "step": 17240
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.132001948368242e-08,
+      "loss": 0.0002,
+      "step": 17241
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.129566488066245e-08,
+      "loss": 0.0002,
+      "step": 17242
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.127131027764247e-08,
+      "loss": 0.0004,
+      "step": 17243
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.12469556746225e-08,
+      "loss": 0.0003,
+      "step": 17244
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.122260107160253e-08,
+      "loss": 0.0002,
+      "step": 17245
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.119824646858256e-08,
+      "loss": 0.0002,
+      "step": 17246
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.117389186556258e-08,
+      "loss": 0.0004,
+      "step": 17247
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.114953726254261e-08,
+      "loss": 0.0002,
+      "step": 17248
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.112518265952265e-08,
+      "loss": 0.0001,
+      "step": 17249
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.110082805650267e-08,
+      "loss": 0.0002,
+      "step": 17250
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 8.10764734534827e-08,
+      "loss": 0.0003,
+      "step": 17251
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.105211885046274e-08,
+      "loss": 0.0002,
+      "step": 17252
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.102776424744276e-08,
+      "loss": 0.0002,
+      "step": 17253
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.100340964442279e-08,
+      "loss": 0.0001,
+      "step": 17254
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.097905504140281e-08,
+      "loss": 0.0003,
+      "step": 17255
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.095470043838285e-08,
+      "loss": 0.0003,
+      "step": 17256
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.093034583536288e-08,
+      "loss": 0.0002,
+      "step": 17257
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.090599123234291e-08,
+      "loss": 0.0001,
+      "step": 17258
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.088163662932294e-08,
+      "loss": 0.0003,
+      "step": 17259
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.085728202630298e-08,
+      "loss": 0.0004,
+      "step": 17260
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 8.0832927423283e-08,
+      "loss": 0.0002,
+      "step": 17261
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.080857282026303e-08,
+      "loss": 0.0001,
+      "step": 17262
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.078421821724305e-08,
+      "loss": 0.0002,
+      "step": 17263
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.075986361422309e-08,
+      "loss": 0.0001,
+      "step": 17264
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.073550901120312e-08,
+      "loss": 0.0002,
+      "step": 17265
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.071115440818314e-08,
+      "loss": 0.0003,
+      "step": 17266
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.068679980516317e-08,
+      "loss": 0.0002,
+      "step": 17267
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.06624452021432e-08,
+      "loss": 0.0002,
+      "step": 17268
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.063809059912323e-08,
+      "loss": 0.0003,
+      "step": 17269
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.061373599610326e-08,
+      "loss": 0.0002,
+      "step": 17270
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 8.058938139308328e-08,
+      "loss": 0.0002,
+      "step": 17271
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.056502679006332e-08,
+      "loss": 0.0002,
+      "step": 17272
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.054067218704335e-08,
+      "loss": 0.0002,
+      "step": 17273
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.051631758402337e-08,
+      "loss": 0.0002,
+      "step": 17274
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.04919629810034e-08,
+      "loss": 0.0003,
+      "step": 17275
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.046760837798343e-08,
+      "loss": 0.0004,
+      "step": 17276
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.044325377496347e-08,
+      "loss": 0.0002,
+      "step": 17277
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.04188991719435e-08,
+      "loss": 0.0002,
+      "step": 17278
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.039454456892352e-08,
+      "loss": 0.0001,
+      "step": 17279
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.037018996590356e-08,
+      "loss": 0.0001,
+      "step": 17280
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.034583536288359e-08,
+      "loss": 0.0002,
+      "step": 17281
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 8.032148075986361e-08,
+      "loss": 0.0002,
+      "step": 17282
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.029712615684364e-08,
+      "loss": 0.0004,
+      "step": 17283
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.027277155382367e-08,
+      "loss": 0.0003,
+      "step": 17284
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.02484169508037e-08,
+      "loss": 0.0002,
+      "step": 17285
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.022406234778373e-08,
+      "loss": 0.0002,
+      "step": 17286
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.019970774476375e-08,
+      "loss": 0.0001,
+      "step": 17287
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.017535314174379e-08,
+      "loss": 0.0002,
+      "step": 17288
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.015099853872381e-08,
+      "loss": 0.0004,
+      "step": 17289
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.012664393570384e-08,
+      "loss": 0.0003,
+      "step": 17290
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.010228933268386e-08,
+      "loss": 0.0002,
+      "step": 17291
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 8.00779347296639e-08,
+      "loss": 0.0003,
+      "step": 17292
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 8.005358012664393e-08,
+      "loss": 0.0002,
+      "step": 17293
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 8.002922552362395e-08,
+      "loss": 0.0004,
+      "step": 17294
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 8.000487092060398e-08,
+      "loss": 0.0002,
+      "step": 17295
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 7.998051631758403e-08,
+      "loss": 0.0003,
+      "step": 17296
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 7.995616171456405e-08,
+      "loss": 0.0003,
+      "step": 17297
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 7.993180711154408e-08,
+      "loss": 1.2905,
+      "step": 17298
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 7.990745250852412e-08,
+      "loss": 0.0002,
+      "step": 17299
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 7.988309790550414e-08,
+      "loss": 0.0002,
+      "step": 17300
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 7.985874330248417e-08,
+      "loss": 0.0003,
+      "step": 17301
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 7.98343886994642e-08,
+      "loss": 0.0002,
+      "step": 17302
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.981003409644423e-08,
+      "loss": 0.0002,
+      "step": 17303
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.978567949342426e-08,
+      "loss": 0.0002,
+      "step": 17304
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.976132489040428e-08,
+      "loss": 0.0002,
+      "step": 17305
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.973697028738431e-08,
+      "loss": 0.0002,
+      "step": 17306
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.971261568436435e-08,
+      "loss": 0.0002,
+      "step": 17307
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.968826108134437e-08,
+      "loss": 0.0001,
+      "step": 17308
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.96639064783244e-08,
+      "loss": 0.0002,
+      "step": 17309
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.963955187530442e-08,
+      "loss": 0.0003,
+      "step": 17310
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.961519727228446e-08,
+      "loss": 0.0003,
+      "step": 17311
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 7.959084266926449e-08,
+      "loss": 0.0003,
+      "step": 17312
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.956648806624451e-08,
+      "loss": 0.0002,
+      "step": 17313
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.954213346322454e-08,
+      "loss": 0.0002,
+      "step": 17314
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.951777886020459e-08,
+      "loss": 0.0004,
+      "step": 17315
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.949342425718461e-08,
+      "loss": 1.1745,
+      "step": 17316
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.946906965416464e-08,
+      "loss": 0.0003,
+      "step": 17317
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.944471505114466e-08,
+      "loss": 0.0023,
+      "step": 17318
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.94203604481247e-08,
+      "loss": 0.0001,
+      "step": 17319
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.939600584510473e-08,
+      "loss": 0.0001,
+      "step": 17320
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.937165124208475e-08,
+      "loss": 0.0004,
+      "step": 17321
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.934729663906478e-08,
+      "loss": 0.0001,
+      "step": 17322
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 7.932294203604482e-08,
+      "loss": 0.0001,
+      "step": 17323
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.929858743302484e-08,
+      "loss": 0.0003,
+      "step": 17324
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.927423283000487e-08,
+      "loss": 0.0004,
+      "step": 17325
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.924987822698489e-08,
+      "loss": 0.0001,
+      "step": 17326
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.922552362396493e-08,
+      "loss": 0.0003,
+      "step": 17327
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.920116902094495e-08,
+      "loss": 0.0002,
+      "step": 17328
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.917681441792498e-08,
+      "loss": 0.0003,
+      "step": 17329
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.9152459814905e-08,
+      "loss": 0.0003,
+      "step": 17330
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.912810521188504e-08,
+      "loss": 0.0002,
+      "step": 17331
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.910375060886507e-08,
+      "loss": 0.0003,
+      "step": 17332
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 7.907939600584509e-08,
+      "loss": 0.0003,
+      "step": 17333
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.905504140282513e-08,
+      "loss": 0.0004,
+      "step": 17334
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.903068679980517e-08,
+      "loss": 0.0002,
+      "step": 17335
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.90063321967852e-08,
+      "loss": 0.0001,
+      "step": 17336
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.898197759376522e-08,
+      "loss": 0.0003,
+      "step": 17337
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.895762299074525e-08,
+      "loss": 0.0017,
+      "step": 17338
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.893326838772528e-08,
+      "loss": 0.0001,
+      "step": 17339
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.890891378470531e-08,
+      "loss": 0.0002,
+      "step": 17340
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.888455918168533e-08,
+      "loss": 0.0003,
+      "step": 17341
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.886020457866536e-08,
+      "loss": 0.0002,
+      "step": 17342
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 7.88358499756454e-08,
+      "loss": 0.0002,
+      "step": 17343
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.881149537262542e-08,
+      "loss": 0.0002,
+      "step": 17344
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.878714076960545e-08,
+      "loss": 0.0002,
+      "step": 17345
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.876278616658547e-08,
+      "loss": 0.0002,
+      "step": 17346
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.873843156356551e-08,
+      "loss": 0.0002,
+      "step": 17347
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.871407696054554e-08,
+      "loss": 0.0003,
+      "step": 17348
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.868972235752556e-08,
+      "loss": 0.0004,
+      "step": 17349
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.866536775450559e-08,
+      "loss": 0.0003,
+      "step": 17350
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.864101315148563e-08,
+      "loss": 0.3574,
+      "step": 17351
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.861665854846565e-08,
+      "loss": 0.0002,
+      "step": 17352
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.859230394544568e-08,
+      "loss": 0.0004,
+      "step": 17353
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 7.856794934242573e-08,
+      "loss": 0.0001,
+      "step": 17354
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.854359473940575e-08,
+      "loss": 0.0002,
+      "step": 17355
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.851924013638578e-08,
+      "loss": 0.0004,
+      "step": 17356
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.84948855333658e-08,
+      "loss": 0.0003,
+      "step": 17357
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.847053093034584e-08,
+      "loss": 0.0003,
+      "step": 17358
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.844617632732587e-08,
+      "loss": 0.0002,
+      "step": 17359
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.842182172430589e-08,
+      "loss": 0.0003,
+      "step": 17360
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.839746712128592e-08,
+      "loss": 0.0003,
+      "step": 17361
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.837311251826596e-08,
+      "loss": 0.0002,
+      "step": 17362
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.834875791524598e-08,
+      "loss": 0.0003,
+      "step": 17363
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 7.8324403312226e-08,
+      "loss": 0.0002,
+      "step": 17364
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.830004870920603e-08,
+      "loss": 0.0001,
+      "step": 17365
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.827569410618607e-08,
+      "loss": 0.0002,
+      "step": 17366
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.82513395031661e-08,
+      "loss": 0.0002,
+      "step": 17367
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.822698490014612e-08,
+      "loss": 0.0001,
+      "step": 17368
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.820263029712614e-08,
+      "loss": 0.0004,
+      "step": 17369
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.817827569410618e-08,
+      "loss": 0.0002,
+      "step": 17370
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.815392109108621e-08,
+      "loss": 0.0005,
+      "step": 17371
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.812956648806623e-08,
+      "loss": 0.0002,
+      "step": 17372
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.810521188504627e-08,
+      "loss": 0.0002,
+      "step": 17373
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 7.808085728202631e-08,
+      "loss": 0.0002,
+      "step": 17374
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.805650267900634e-08,
+      "loss": 0.0003,
+      "step": 17375
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.803214807598636e-08,
+      "loss": 0.0002,
+      "step": 17376
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.800779347296639e-08,
+      "loss": 0.0001,
+      "step": 17377
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.798343886994642e-08,
+      "loss": 0.0002,
+      "step": 17378
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.795908426692645e-08,
+      "loss": 0.0003,
+      "step": 17379
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.793472966390647e-08,
+      "loss": 0.0001,
+      "step": 17380
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.79103750608865e-08,
+      "loss": 0.0002,
+      "step": 17381
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.788602045786654e-08,
+      "loss": 0.0003,
+      "step": 17382
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.786166585484656e-08,
+      "loss": 0.0003,
+      "step": 17383
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 7.783731125182659e-08,
+      "loss": 0.0003,
+      "step": 17384
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.781295664880661e-08,
+      "loss": 0.0003,
+      "step": 17385
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.778860204578665e-08,
+      "loss": 0.0001,
+      "step": 17386
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.776424744276668e-08,
+      "loss": 0.0002,
+      "step": 17387
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.77398928397467e-08,
+      "loss": 0.0002,
+      "step": 17388
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.771553823672673e-08,
+      "loss": 0.0003,
+      "step": 17389
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.769118363370677e-08,
+      "loss": 0.0005,
+      "step": 17390
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.766682903068679e-08,
+      "loss": 0.0004,
+      "step": 17391
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.764247442766683e-08,
+      "loss": 0.0003,
+      "step": 17392
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.761811982464685e-08,
+      "loss": 0.0002,
+      "step": 17393
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.759376522162689e-08,
+      "loss": 0.0003,
+      "step": 17394
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 7.756941061860692e-08,
+      "loss": 0.0004,
+      "step": 17395
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.754505601558694e-08,
+      "loss": 0.0003,
+      "step": 17396
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.752070141256697e-08,
+      "loss": 0.0003,
+      "step": 17397
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.749634680954701e-08,
+      "loss": 0.0002,
+      "step": 17398
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.747199220652703e-08,
+      "loss": 0.0002,
+      "step": 17399
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.744763760350706e-08,
+      "loss": 0.0002,
+      "step": 17400
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.742328300048708e-08,
+      "loss": 0.0004,
+      "step": 17401
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.739892839746712e-08,
+      "loss": 0.0002,
+      "step": 17402
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.737457379444715e-08,
+      "loss": 0.0002,
+      "step": 17403
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.735021919142717e-08,
+      "loss": 0.0006,
+      "step": 17404
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 7.73258645884072e-08,
+      "loss": 0.0002,
+      "step": 17405
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.730150998538723e-08,
+      "loss": 0.0004,
+      "step": 17406
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.727715538236726e-08,
+      "loss": 0.0002,
+      "step": 17407
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.725280077934729e-08,
+      "loss": 0.0001,
+      "step": 17408
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.722844617632731e-08,
+      "loss": 0.0002,
+      "step": 17409
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.720409157330735e-08,
+      "loss": 0.0002,
+      "step": 17410
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.717973697028739e-08,
+      "loss": 0.0002,
+      "step": 17411
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.715538236726741e-08,
+      "loss": 0.0002,
+      "step": 17412
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.713102776424745e-08,
+      "loss": 0.0002,
+      "step": 17413
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.710667316122748e-08,
+      "loss": 0.0002,
+      "step": 17414
+    },
+    {
+      "epoch": 16.92,
+      "learning_rate": 7.70823185582075e-08,
+      "loss": 0.0003,
+      "step": 17415
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.705796395518753e-08,
+      "loss": 0.0003,
+      "step": 17416
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.703360935216756e-08,
+      "loss": 0.0002,
+      "step": 17417
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.700925474914759e-08,
+      "loss": 0.0003,
+      "step": 17418
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.698490014612761e-08,
+      "loss": 0.0002,
+      "step": 17419
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.696054554310764e-08,
+      "loss": 0.0002,
+      "step": 17420
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.693619094008768e-08,
+      "loss": 0.0001,
+      "step": 17421
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.69118363370677e-08,
+      "loss": 0.0002,
+      "step": 17422
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.688748173404773e-08,
+      "loss": 0.0002,
+      "step": 17423
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.686312713102775e-08,
+      "loss": 0.0001,
+      "step": 17424
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.683877252800779e-08,
+      "loss": 0.0003,
+      "step": 17425
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 7.681441792498782e-08,
+      "loss": 0.0002,
+      "step": 17426
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.679006332196784e-08,
+      "loss": 0.0003,
+      "step": 17427
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.676570871894787e-08,
+      "loss": 0.0003,
+      "step": 17428
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.67413541159279e-08,
+      "loss": 0.0001,
+      "step": 17429
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.671699951290794e-08,
+      "loss": 0.0002,
+      "step": 17430
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.669264490988797e-08,
+      "loss": 0.0002,
+      "step": 17431
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.6668290306868e-08,
+      "loss": 0.0003,
+      "step": 17432
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.664393570384803e-08,
+      "loss": 0.0001,
+      "step": 17433
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.661958110082806e-08,
+      "loss": 0.0002,
+      "step": 17434
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.659522649780808e-08,
+      "loss": 0.0004,
+      "step": 17435
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 7.657087189478811e-08,
+      "loss": 0.0003,
+      "step": 17436
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.654651729176815e-08,
+      "loss": 0.0002,
+      "step": 17437
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.652216268874817e-08,
+      "loss": 0.0003,
+      "step": 17438
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.64978080857282e-08,
+      "loss": 0.0003,
+      "step": 17439
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.647345348270822e-08,
+      "loss": 0.0001,
+      "step": 17440
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.644909887968826e-08,
+      "loss": 0.0002,
+      "step": 17441
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.642474427666829e-08,
+      "loss": 0.0001,
+      "step": 17442
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.640038967364831e-08,
+      "loss": 0.0003,
+      "step": 17443
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.637603507062834e-08,
+      "loss": 0.0003,
+      "step": 17444
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.635168046760837e-08,
+      "loss": 0.0003,
+      "step": 17445
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 7.63273258645884e-08,
+      "loss": 0.0003,
+      "step": 17446
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.630297126156843e-08,
+      "loss": 0.0003,
+      "step": 17447
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.627861665854845e-08,
+      "loss": 0.0003,
+      "step": 17448
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.62542620555285e-08,
+      "loss": 0.0002,
+      "step": 17449
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.622990745250853e-08,
+      "loss": 0.0003,
+      "step": 17450
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.620555284948855e-08,
+      "loss": 0.0002,
+      "step": 17451
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.618119824646858e-08,
+      "loss": 0.0003,
+      "step": 17452
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.615684364344862e-08,
+      "loss": 0.0001,
+      "step": 17453
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.613248904042864e-08,
+      "loss": 0.0002,
+      "step": 17454
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.610813443740867e-08,
+      "loss": 0.0001,
+      "step": 17455
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 7.608377983438869e-08,
+      "loss": 0.0003,
+      "step": 17456
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.605942523136873e-08,
+      "loss": 0.0002,
+      "step": 17457
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.603507062834876e-08,
+      "loss": 0.0003,
+      "step": 17458
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.601071602532878e-08,
+      "loss": 0.0003,
+      "step": 17459
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.59863614223088e-08,
+      "loss": 0.0017,
+      "step": 17460
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.596200681928884e-08,
+      "loss": 0.0002,
+      "step": 17461
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.593765221626887e-08,
+      "loss": 0.0002,
+      "step": 17462
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.59132976132489e-08,
+      "loss": 0.0004,
+      "step": 17463
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.588894301022892e-08,
+      "loss": 0.0002,
+      "step": 17464
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.586458840720896e-08,
+      "loss": 0.0003,
+      "step": 17465
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.584023380418898e-08,
+      "loss": 0.0004,
+      "step": 17466
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 7.581587920116901e-08,
+      "loss": 0.0001,
+      "step": 17467
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.579152459814906e-08,
+      "loss": 0.0003,
+      "step": 17468
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.576716999512908e-08,
+      "loss": 0.0003,
+      "step": 17469
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.574281539210911e-08,
+      "loss": 0.0005,
+      "step": 17470
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.571846078908914e-08,
+      "loss": 0.0003,
+      "step": 17471
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.569410618606917e-08,
+      "loss": 0.0004,
+      "step": 17472
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.56697515830492e-08,
+      "loss": 0.0003,
+      "step": 17473
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.564539698002922e-08,
+      "loss": 0.0002,
+      "step": 17474
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.562104237700925e-08,
+      "loss": 0.0002,
+      "step": 17475
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.559668777398929e-08,
+      "loss": 0.0002,
+      "step": 17476
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 7.557233317096931e-08,
+      "loss": 0.0002,
+      "step": 17477
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.554797856794934e-08,
+      "loss": 0.0002,
+      "step": 17478
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.552362396492936e-08,
+      "loss": 0.0002,
+      "step": 17479
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.54992693619094e-08,
+      "loss": 0.0004,
+      "step": 17480
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.547491475888943e-08,
+      "loss": 0.0004,
+      "step": 17481
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.545056015586945e-08,
+      "loss": 0.0002,
+      "step": 17482
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.542620555284948e-08,
+      "loss": 0.0003,
+      "step": 17483
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.540185094982952e-08,
+      "loss": 0.0001,
+      "step": 17484
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.537749634680954e-08,
+      "loss": 0.0002,
+      "step": 17485
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.535314174378957e-08,
+      "loss": 0.0003,
+      "step": 17486
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 7.532878714076959e-08,
+      "loss": 0.0002,
+      "step": 17487
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.530443253774964e-08,
+      "loss": 0.0002,
+      "step": 17488
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.528007793472967e-08,
+      "loss": 0.0003,
+      "step": 17489
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.525572333170969e-08,
+      "loss": 0.0002,
+      "step": 17490
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.523136872868972e-08,
+      "loss": 0.0002,
+      "step": 17491
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.520701412566976e-08,
+      "loss": 0.0004,
+      "step": 17492
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.518265952264978e-08,
+      "loss": 0.0003,
+      "step": 17493
+    },
+    {
+      "epoch": 17.0,
+      "eval_accuracy": 0.9990277102576568,
+      "eval_loss": 0.014137467369437218,
+      "eval_runtime": 2.7889,
+      "eval_samples_per_second": 737.567,
+      "eval_steps_per_second": 11.833,
+      "step": 17493
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.51583049196298e-08,
+      "loss": 0.0002,
+      "step": 17494
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.513395031660983e-08,
+      "loss": 0.0003,
+      "step": 17495
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.510959571358987e-08,
+      "loss": 0.0002,
+      "step": 17496
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.50852411105699e-08,
+      "loss": 0.0004,
+      "step": 17497
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 7.506088650754992e-08,
+      "loss": 0.0002,
+      "step": 17498
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.503653190452995e-08,
+      "loss": 0.0006,
+      "step": 17499
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.501217730150998e-08,
+      "loss": 0.0002,
+      "step": 17500
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.498782269849001e-08,
+      "loss": 0.0001,
+      "step": 17501
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.496346809547003e-08,
+      "loss": 0.0003,
+      "step": 17502
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.493911349245006e-08,
+      "loss": 0.0002,
+      "step": 17503
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.49147588894301e-08,
+      "loss": 0.0001,
+      "step": 17504
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.489040428641012e-08,
+      "loss": 0.0003,
+      "step": 17505
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.486604968339015e-08,
+      "loss": 0.0001,
+      "step": 17506
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.484169508037019e-08,
+      "loss": 0.0007,
+      "step": 17507
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 7.481734047735023e-08,
+      "loss": 0.0003,
+      "step": 17508
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.479298587433025e-08,
+      "loss": 0.0003,
+      "step": 17509
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.476863127131028e-08,
+      "loss": 0.0003,
+      "step": 17510
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.47442766682903e-08,
+      "loss": 0.0002,
+      "step": 17511
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.471992206527034e-08,
+      "loss": 0.0001,
+      "step": 17512
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.469556746225036e-08,
+      "loss": 0.0002,
+      "step": 17513
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.467121285923039e-08,
+      "loss": 0.0002,
+      "step": 17514
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.464685825621041e-08,
+      "loss": 0.0002,
+      "step": 17515
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.462250365319045e-08,
+      "loss": 0.0003,
+      "step": 17516
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.459814905017048e-08,
+      "loss": 0.0002,
+      "step": 17517
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 7.45737944471505e-08,
+      "loss": 0.0003,
+      "step": 17518
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.454943984413054e-08,
+      "loss": 0.0002,
+      "step": 17519
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.452508524111057e-08,
+      "loss": 0.0001,
+      "step": 17520
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.450073063809059e-08,
+      "loss": 0.0002,
+      "step": 17521
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.447637603507062e-08,
+      "loss": 0.0003,
+      "step": 17522
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.445202143205066e-08,
+      "loss": 0.0001,
+      "step": 17523
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.442766682903068e-08,
+      "loss": 0.0004,
+      "step": 17524
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.44033122260107e-08,
+      "loss": 0.0001,
+      "step": 17525
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.437895762299074e-08,
+      "loss": 0.0004,
+      "step": 17526
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.435460301997078e-08,
+      "loss": 0.0003,
+      "step": 17527
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.433024841695081e-08,
+      "loss": 0.0003,
+      "step": 17528
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 7.430589381393083e-08,
+      "loss": 0.0002,
+      "step": 17529
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.428153921091086e-08,
+      "loss": 0.0001,
+      "step": 17530
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.42571846078909e-08,
+      "loss": 0.0002,
+      "step": 17531
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.423283000487092e-08,
+      "loss": 0.0004,
+      "step": 17532
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.420847540185095e-08,
+      "loss": 0.0004,
+      "step": 17533
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.418412079883097e-08,
+      "loss": 0.0003,
+      "step": 17534
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.415976619581101e-08,
+      "loss": 0.0002,
+      "step": 17535
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.413541159279104e-08,
+      "loss": 0.0003,
+      "step": 17536
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.411105698977106e-08,
+      "loss": 0.0002,
+      "step": 17537
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.408670238675109e-08,
+      "loss": 0.0003,
+      "step": 17538
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 7.406234778373112e-08,
+      "loss": 0.0003,
+      "step": 17539
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.403799318071115e-08,
+      "loss": 0.0002,
+      "step": 17540
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.401363857769117e-08,
+      "loss": 0.0002,
+      "step": 17541
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.39892839746712e-08,
+      "loss": 0.0002,
+      "step": 17542
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.396492937165124e-08,
+      "loss": 0.0001,
+      "step": 17543
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.394057476863126e-08,
+      "loss": 0.0002,
+      "step": 17544
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.39162201656113e-08,
+      "loss": 0.0001,
+      "step": 17545
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.389186556259133e-08,
+      "loss": 0.0007,
+      "step": 17546
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.386751095957137e-08,
+      "loss": 0.0004,
+      "step": 17547
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.384315635655139e-08,
+      "loss": 0.0001,
+      "step": 17548
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 7.381880175353142e-08,
+      "loss": 0.0004,
+      "step": 17549
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.379444715051144e-08,
+      "loss": 0.0002,
+      "step": 17550
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.377009254749148e-08,
+      "loss": 0.0003,
+      "step": 17551
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.37457379444715e-08,
+      "loss": 0.0004,
+      "step": 17552
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.372138334145153e-08,
+      "loss": 0.0002,
+      "step": 17553
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.369702873843155e-08,
+      "loss": 0.0001,
+      "step": 17554
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.367267413541159e-08,
+      "loss": 0.0001,
+      "step": 17555
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.364831953239162e-08,
+      "loss": 0.0002,
+      "step": 17556
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.362396492937164e-08,
+      "loss": 0.0002,
+      "step": 17557
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.359961032635167e-08,
+      "loss": 0.0002,
+      "step": 17558
+    },
+    {
+      "epoch": 17.06,
+      "learning_rate": 7.357525572333171e-08,
+      "loss": 0.0002,
+      "step": 17559
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.355090112031173e-08,
+      "loss": 0.0005,
+      "step": 17560
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.352654651729176e-08,
+      "loss": 0.0002,
+      "step": 17561
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.350219191427178e-08,
+      "loss": 0.0002,
+      "step": 17562
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.347783731125182e-08,
+      "loss": 0.0002,
+      "step": 17563
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.345348270823186e-08,
+      "loss": 0.0002,
+      "step": 17564
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.342912810521188e-08,
+      "loss": 0.0002,
+      "step": 17565
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.340477350219191e-08,
+      "loss": 0.0003,
+      "step": 17566
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.338041889917195e-08,
+      "loss": 0.0002,
+      "step": 17567
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.335606429615197e-08,
+      "loss": 0.0003,
+      "step": 17568
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.3331709693132e-08,
+      "loss": 0.0003,
+      "step": 17569
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 7.330735509011204e-08,
+      "loss": 0.0001,
+      "step": 17570
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.328300048709206e-08,
+      "loss": 0.0003,
+      "step": 17571
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.325864588407209e-08,
+      "loss": 0.0002,
+      "step": 17572
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.323429128105211e-08,
+      "loss": 0.0002,
+      "step": 17573
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.320993667803215e-08,
+      "loss": 0.0005,
+      "step": 17574
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.318558207501218e-08,
+      "loss": 0.0002,
+      "step": 17575
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.31612274719922e-08,
+      "loss": 0.0005,
+      "step": 17576
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.313687286897223e-08,
+      "loss": 0.0002,
+      "step": 17577
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.311251826595226e-08,
+      "loss": 0.0002,
+      "step": 17578
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.308816366293229e-08,
+      "loss": 0.0002,
+      "step": 17579
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 7.306380905991231e-08,
+      "loss": 0.0003,
+      "step": 17580
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.303945445689234e-08,
+      "loss": 0.0004,
+      "step": 17581
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.301509985387238e-08,
+      "loss": 0.0002,
+      "step": 17582
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.299074525085242e-08,
+      "loss": 0.0004,
+      "step": 17583
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.296639064783244e-08,
+      "loss": 0.0002,
+      "step": 17584
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.294203604481247e-08,
+      "loss": 0.0002,
+      "step": 17585
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.29176814417925e-08,
+      "loss": 0.0002,
+      "step": 17586
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.289332683877253e-08,
+      "loss": 0.0001,
+      "step": 17587
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.286897223575256e-08,
+      "loss": 0.0003,
+      "step": 17588
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.284461763273258e-08,
+      "loss": 0.0001,
+      "step": 17589
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 7.282026302971262e-08,
+      "loss": 0.0001,
+      "step": 17590
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.279590842669264e-08,
+      "loss": 0.0002,
+      "step": 17591
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.277155382367267e-08,
+      "loss": 0.0004,
+      "step": 17592
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.27471992206527e-08,
+      "loss": 0.0002,
+      "step": 17593
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.272284461763273e-08,
+      "loss": 0.0002,
+      "step": 17594
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.269849001461276e-08,
+      "loss": 0.0003,
+      "step": 17595
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.267413541159278e-08,
+      "loss": 0.0002,
+      "step": 17596
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.264978080857281e-08,
+      "loss": 0.0001,
+      "step": 17597
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.262542620555285e-08,
+      "loss": 0.0005,
+      "step": 17598
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.260107160253287e-08,
+      "loss": 0.0004,
+      "step": 17599
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.25767169995129e-08,
+      "loss": 0.0002,
+      "step": 17600
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 7.255236239649292e-08,
+      "loss": 0.0002,
+      "step": 17601
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.252800779347296e-08,
+      "loss": 0.0002,
+      "step": 17602
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.2503653190453e-08,
+      "loss": 0.0001,
+      "step": 17603
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.247929858743302e-08,
+      "loss": 0.0002,
+      "step": 17604
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.245494398441305e-08,
+      "loss": 0.0003,
+      "step": 17605
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.243058938139309e-08,
+      "loss": 0.0012,
+      "step": 17606
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.240623477837311e-08,
+      "loss": 0.0003,
+      "step": 17607
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.238188017535314e-08,
+      "loss": 0.0002,
+      "step": 17608
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.235752557233316e-08,
+      "loss": 0.0002,
+      "step": 17609
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.23331709693132e-08,
+      "loss": 0.0001,
+      "step": 17610
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 7.230881636629323e-08,
+      "loss": 0.0002,
+      "step": 17611
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.228446176327325e-08,
+      "loss": 0.0002,
+      "step": 17612
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.226010716025328e-08,
+      "loss": 0.0001,
+      "step": 17613
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.223575255723332e-08,
+      "loss": 0.0002,
+      "step": 17614
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.221139795421334e-08,
+      "loss": 0.0002,
+      "step": 17615
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.218704335119337e-08,
+      "loss": 0.0003,
+      "step": 17616
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.216268874817339e-08,
+      "loss": 0.0002,
+      "step": 17617
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.213833414515343e-08,
+      "loss": 0.0002,
+      "step": 17618
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.211397954213346e-08,
+      "loss": 0.0002,
+      "step": 17619
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.208962493911348e-08,
+      "loss": 0.0002,
+      "step": 17620
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 7.20652703360935e-08,
+      "loss": 0.0002,
+      "step": 17621
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.204091573307356e-08,
+      "loss": 0.0003,
+      "step": 17622
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.201656113005358e-08,
+      "loss": 0.0002,
+      "step": 17623
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.199220652703361e-08,
+      "loss": 0.0002,
+      "step": 17624
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.196785192401365e-08,
+      "loss": 0.0002,
+      "step": 17625
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.194349732099367e-08,
+      "loss": 0.0004,
+      "step": 17626
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.19191427179737e-08,
+      "loss": 0.0003,
+      "step": 17627
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.189478811495372e-08,
+      "loss": 0.0002,
+      "step": 17628
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.187043351193376e-08,
+      "loss": 0.0002,
+      "step": 17629
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.184607890891378e-08,
+      "loss": 0.0001,
+      "step": 17630
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 7.182172430589381e-08,
+      "loss": 0.0002,
+      "step": 17631
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.179736970287384e-08,
+      "loss": 0.0003,
+      "step": 17632
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.177301509985387e-08,
+      "loss": 0.0003,
+      "step": 17633
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.17486604968339e-08,
+      "loss": 0.0003,
+      "step": 17634
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.172430589381392e-08,
+      "loss": 0.0001,
+      "step": 17635
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.169995129079395e-08,
+      "loss": 0.0004,
+      "step": 17636
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.167559668777399e-08,
+      "loss": 0.0002,
+      "step": 17637
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.165124208475401e-08,
+      "loss": 0.0003,
+      "step": 17638
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.162688748173404e-08,
+      "loss": 0.0002,
+      "step": 17639
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.160253287871406e-08,
+      "loss": 0.0003,
+      "step": 17640
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.157817827569411e-08,
+      "loss": 0.0003,
+      "step": 17641
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 7.155382367267414e-08,
+      "loss": 0.0001,
+      "step": 17642
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.152946906965416e-08,
+      "loss": 0.0002,
+      "step": 17643
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.150511446663419e-08,
+      "loss": 0.0002,
+      "step": 17644
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.148075986361423e-08,
+      "loss": 0.0002,
+      "step": 17645
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.145640526059425e-08,
+      "loss": 0.0002,
+      "step": 17646
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.143205065757428e-08,
+      "loss": 0.0002,
+      "step": 17647
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.14076960545543e-08,
+      "loss": 0.0001,
+      "step": 17648
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.138334145153434e-08,
+      "loss": 0.0002,
+      "step": 17649
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.135898684851437e-08,
+      "loss": 0.0002,
+      "step": 17650
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.133463224549439e-08,
+      "loss": 0.0002,
+      "step": 17651
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 7.131027764247442e-08,
+      "loss": 0.0003,
+      "step": 17652
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.128592303945446e-08,
+      "loss": 0.0002,
+      "step": 17653
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.126156843643448e-08,
+      "loss": 0.0001,
+      "step": 17654
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.123721383341451e-08,
+      "loss": 0.0001,
+      "step": 17655
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.121285923039453e-08,
+      "loss": 0.0002,
+      "step": 17656
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.118850462737457e-08,
+      "loss": 0.0002,
+      "step": 17657
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.11641500243546e-08,
+      "loss": 0.0003,
+      "step": 17658
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.113979542133462e-08,
+      "loss": 0.0001,
+      "step": 17659
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.111544081831466e-08,
+      "loss": 0.0003,
+      "step": 17660
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.10910862152947e-08,
+      "loss": 0.0002,
+      "step": 17661
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 7.106673161227472e-08,
+      "loss": 0.0002,
+      "step": 17662
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.104237700925475e-08,
+      "loss": 0.0003,
+      "step": 17663
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.101802240623477e-08,
+      "loss": 0.0004,
+      "step": 17664
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.099366780321481e-08,
+      "loss": 0.0002,
+      "step": 17665
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.096931320019484e-08,
+      "loss": 0.0002,
+      "step": 17666
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.094495859717486e-08,
+      "loss": 0.0002,
+      "step": 17667
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.092060399415489e-08,
+      "loss": 0.0005,
+      "step": 17668
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.089624939113493e-08,
+      "loss": 0.0002,
+      "step": 17669
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.087189478811495e-08,
+      "loss": 0.0003,
+      "step": 17670
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.084754018509498e-08,
+      "loss": 0.0002,
+      "step": 17671
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.0823185582075e-08,
+      "loss": 0.0003,
+      "step": 17672
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 7.079883097905504e-08,
+      "loss": 0.0001,
+      "step": 17673
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.077447637603506e-08,
+      "loss": 0.0003,
+      "step": 17674
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.075012177301509e-08,
+      "loss": 0.0002,
+      "step": 17675
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.072576716999511e-08,
+      "loss": 0.0003,
+      "step": 17676
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.070141256697515e-08,
+      "loss": 0.0002,
+      "step": 17677
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.067705796395518e-08,
+      "loss": 0.0002,
+      "step": 17678
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.065270336093522e-08,
+      "loss": 0.0002,
+      "step": 17679
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.062834875791525e-08,
+      "loss": 0.0002,
+      "step": 17680
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.060399415489528e-08,
+      "loss": 0.0002,
+      "step": 17681
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.05796395518753e-08,
+      "loss": 0.0002,
+      "step": 17682
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 7.055528494885533e-08,
+      "loss": 1.3696,
+      "step": 17683
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.053093034583537e-08,
+      "loss": 0.0002,
+      "step": 17684
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.05065757428154e-08,
+      "loss": 0.0002,
+      "step": 17685
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.048222113979542e-08,
+      "loss": 0.0003,
+      "step": 17686
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.045786653677544e-08,
+      "loss": 0.0003,
+      "step": 17687
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.043351193375548e-08,
+      "loss": 0.0004,
+      "step": 17688
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.040915733073551e-08,
+      "loss": 0.0002,
+      "step": 17689
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.038480272771553e-08,
+      "loss": 0.0003,
+      "step": 17690
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.036044812469556e-08,
+      "loss": 0.0003,
+      "step": 17691
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.03360935216756e-08,
+      "loss": 0.0004,
+      "step": 17692
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 7.031173891865562e-08,
+      "loss": 0.0003,
+      "step": 17693
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.028738431563565e-08,
+      "loss": 0.0003,
+      "step": 17694
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.026302971261567e-08,
+      "loss": 0.0003,
+      "step": 17695
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.023867510959571e-08,
+      "loss": 0.0003,
+      "step": 17696
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.021432050657574e-08,
+      "loss": 0.0005,
+      "step": 17697
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.018996590355577e-08,
+      "loss": 0.0003,
+      "step": 17698
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.01656113005358e-08,
+      "loss": 0.0002,
+      "step": 17699
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.014125669751584e-08,
+      "loss": 0.0002,
+      "step": 17700
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.011690209449586e-08,
+      "loss": 0.0001,
+      "step": 17701
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.009254749147589e-08,
+      "loss": 0.0002,
+      "step": 17702
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 7.006819288845591e-08,
+      "loss": 0.0002,
+      "step": 17703
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 7.004383828543595e-08,
+      "loss": 0.0002,
+      "step": 17704
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 7.001948368241598e-08,
+      "loss": 0.0001,
+      "step": 17705
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.9995129079396e-08,
+      "loss": 0.0002,
+      "step": 17706
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.997077447637603e-08,
+      "loss": 0.0004,
+      "step": 17707
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.994641987335607e-08,
+      "loss": 0.0001,
+      "step": 17708
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.992206527033609e-08,
+      "loss": 0.0002,
+      "step": 17709
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.989771066731612e-08,
+      "loss": 0.0002,
+      "step": 17710
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.987335606429614e-08,
+      "loss": 0.0002,
+      "step": 17711
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.984900146127618e-08,
+      "loss": 0.0002,
+      "step": 17712
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.98246468582562e-08,
+      "loss": 0.0002,
+      "step": 17713
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 6.980029225523623e-08,
+      "loss": 0.0001,
+      "step": 17714
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.977593765221625e-08,
+      "loss": 0.0002,
+      "step": 17715
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.975158304919629e-08,
+      "loss": 0.0002,
+      "step": 17716
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.972722844617633e-08,
+      "loss": 0.0002,
+      "step": 17717
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.970287384315636e-08,
+      "loss": 0.0003,
+      "step": 17718
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.967851924013638e-08,
+      "loss": 0.0001,
+      "step": 17719
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.965416463711642e-08,
+      "loss": 0.0003,
+      "step": 17720
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.962981003409645e-08,
+      "loss": 0.0002,
+      "step": 17721
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.960545543107647e-08,
+      "loss": 0.0003,
+      "step": 17722
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.95811008280565e-08,
+      "loss": 0.0004,
+      "step": 17723
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 6.955674622503653e-08,
+      "loss": 0.0003,
+      "step": 17724
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.953239162201656e-08,
+      "loss": 0.0003,
+      "step": 17725
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.950803701899658e-08,
+      "loss": 0.0001,
+      "step": 17726
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.948368241597661e-08,
+      "loss": 0.0003,
+      "step": 17727
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.945932781295665e-08,
+      "loss": 0.0001,
+      "step": 17728
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.943497320993667e-08,
+      "loss": 0.0002,
+      "step": 17729
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.94106186069167e-08,
+      "loss": 0.0004,
+      "step": 17730
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.938626400389672e-08,
+      "loss": 0.0002,
+      "step": 17731
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.936190940087676e-08,
+      "loss": 0.0003,
+      "step": 17732
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.933755479785679e-08,
+      "loss": 0.0002,
+      "step": 17733
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 6.931320019483681e-08,
+      "loss": 0.0003,
+      "step": 17734
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.928884559181684e-08,
+      "loss": 0.0003,
+      "step": 17735
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.926449098879688e-08,
+      "loss": 0.0003,
+      "step": 17736
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.924013638577691e-08,
+      "loss": 0.0003,
+      "step": 17737
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.921578178275694e-08,
+      "loss": 1.1603,
+      "step": 17738
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.919142717973698e-08,
+      "loss": 0.0002,
+      "step": 17739
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.9167072576717e-08,
+      "loss": 0.0003,
+      "step": 17740
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.914271797369703e-08,
+      "loss": 0.0003,
+      "step": 17741
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.911836337067705e-08,
+      "loss": 0.0003,
+      "step": 17742
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.909400876765709e-08,
+      "loss": 0.0004,
+      "step": 17743
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.906965416463712e-08,
+      "loss": 0.0003,
+      "step": 17744
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 6.904529956161714e-08,
+      "loss": 0.0003,
+      "step": 17745
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.902094495859717e-08,
+      "loss": 0.0002,
+      "step": 17746
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.89965903555772e-08,
+      "loss": 0.0001,
+      "step": 17747
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.897223575255723e-08,
+      "loss": 0.0002,
+      "step": 17748
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.894788114953726e-08,
+      "loss": 0.0002,
+      "step": 17749
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.892352654651728e-08,
+      "loss": 0.0003,
+      "step": 17750
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.889917194349732e-08,
+      "loss": 0.0001,
+      "step": 17751
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.887481734047734e-08,
+      "loss": 0.0002,
+      "step": 17752
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.885046273745737e-08,
+      "loss": 0.0002,
+      "step": 17753
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.88261081344374e-08,
+      "loss": 0.0002,
+      "step": 17754
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 6.880175353141743e-08,
+      "loss": 0.0004,
+      "step": 17755
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.877739892839747e-08,
+      "loss": 0.0002,
+      "step": 17756
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.87530443253775e-08,
+      "loss": 0.0001,
+      "step": 17757
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.872868972235752e-08,
+      "loss": 0.0003,
+      "step": 17758
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.870433511933756e-08,
+      "loss": 0.0003,
+      "step": 17759
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.867998051631759e-08,
+      "loss": 0.0002,
+      "step": 17760
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.865562591329761e-08,
+      "loss": 0.0002,
+      "step": 17761
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.863127131027764e-08,
+      "loss": 0.0003,
+      "step": 17762
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.860691670725767e-08,
+      "loss": 0.0002,
+      "step": 17763
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.85825621042377e-08,
+      "loss": 0.0002,
+      "step": 17764
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 6.855820750121772e-08,
+      "loss": 0.0003,
+      "step": 17765
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.853385289819775e-08,
+      "loss": 0.0002,
+      "step": 17766
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.850949829517779e-08,
+      "loss": 0.0003,
+      "step": 17767
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.848514369215781e-08,
+      "loss": 0.0002,
+      "step": 17768
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.846078908913784e-08,
+      "loss": 0.0004,
+      "step": 17769
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.843643448611786e-08,
+      "loss": 0.0002,
+      "step": 17770
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.84120798830979e-08,
+      "loss": 0.0002,
+      "step": 17771
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.838772528007793e-08,
+      "loss": 0.0002,
+      "step": 17772
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.836337067705795e-08,
+      "loss": 0.0003,
+      "step": 17773
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.833901607403798e-08,
+      "loss": 0.0003,
+      "step": 17774
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 6.831466147101803e-08,
+      "loss": 0.0002,
+      "step": 17775
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.829030686799805e-08,
+      "loss": 0.0004,
+      "step": 17776
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.826595226497808e-08,
+      "loss": 0.0002,
+      "step": 17777
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.82415976619581e-08,
+      "loss": 0.0005,
+      "step": 17778
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.821724305893814e-08,
+      "loss": 0.0003,
+      "step": 17779
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.819288845591817e-08,
+      "loss": 0.0004,
+      "step": 17780
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.81685338528982e-08,
+      "loss": 0.0002,
+      "step": 17781
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.814417924987822e-08,
+      "loss": 0.0002,
+      "step": 17782
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.811982464685826e-08,
+      "loss": 0.0002,
+      "step": 17783
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.809547004383828e-08,
+      "loss": 0.0002,
+      "step": 17784
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.807111544081831e-08,
+      "loss": 0.0003,
+      "step": 17785
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 6.804676083779833e-08,
+      "loss": 0.0002,
+      "step": 17786
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.802240623477837e-08,
+      "loss": 0.0002,
+      "step": 17787
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.79980516317584e-08,
+      "loss": 0.0004,
+      "step": 17788
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.797369702873842e-08,
+      "loss": 0.0003,
+      "step": 17789
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.794934242571846e-08,
+      "loss": 0.0001,
+      "step": 17790
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.792498782269849e-08,
+      "loss": 0.0002,
+      "step": 17791
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.790063321967851e-08,
+      "loss": 0.0004,
+      "step": 17792
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.787627861665854e-08,
+      "loss": 0.0002,
+      "step": 17793
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.785192401363859e-08,
+      "loss": 0.0003,
+      "step": 17794
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.782756941061861e-08,
+      "loss": 0.0003,
+      "step": 17795
+    },
+    {
+      "epoch": 17.29,
+      "learning_rate": 6.780321480759864e-08,
+      "loss": 0.0003,
+      "step": 17796
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.777886020457866e-08,
+      "loss": 0.0002,
+      "step": 17797
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.77545056015587e-08,
+      "loss": 0.0001,
+      "step": 17798
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.773015099853873e-08,
+      "loss": 0.0002,
+      "step": 17799
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.770579639551875e-08,
+      "loss": 0.0002,
+      "step": 17800
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.768144179249878e-08,
+      "loss": 0.0002,
+      "step": 17801
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.765708718947881e-08,
+      "loss": 0.0003,
+      "step": 17802
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.763273258645884e-08,
+      "loss": 0.0005,
+      "step": 17803
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.760837798343887e-08,
+      "loss": 0.0001,
+      "step": 17804
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.758402338041889e-08,
+      "loss": 0.0002,
+      "step": 17805
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 6.755966877739893e-08,
+      "loss": 0.0002,
+      "step": 17806
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.753531417437895e-08,
+      "loss": 0.0001,
+      "step": 17807
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.751095957135898e-08,
+      "loss": 0.0002,
+      "step": 17808
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.7486604968339e-08,
+      "loss": 0.0004,
+      "step": 17809
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.746225036531904e-08,
+      "loss": 0.0002,
+      "step": 17810
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.743789576229907e-08,
+      "loss": 0.0004,
+      "step": 17811
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.741354115927909e-08,
+      "loss": 0.0001,
+      "step": 17812
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.738918655625913e-08,
+      "loss": 0.0004,
+      "step": 17813
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.736483195323917e-08,
+      "loss": 0.0003,
+      "step": 17814
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.73404773502192e-08,
+      "loss": 0.0004,
+      "step": 17815
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.731612274719922e-08,
+      "loss": 0.0008,
+      "step": 17816
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 6.729176814417925e-08,
+      "loss": 0.0002,
+      "step": 17817
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.726741354115928e-08,
+      "loss": 0.0002,
+      "step": 17818
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.724305893813931e-08,
+      "loss": 0.0001,
+      "step": 17819
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.721870433511933e-08,
+      "loss": 0.0002,
+      "step": 17820
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.719434973209936e-08,
+      "loss": 0.0002,
+      "step": 17821
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.71699951290794e-08,
+      "loss": 0.0002,
+      "step": 17822
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.714564052605942e-08,
+      "loss": 0.0004,
+      "step": 17823
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.712128592303945e-08,
+      "loss": 0.0003,
+      "step": 17824
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.709693132001947e-08,
+      "loss": 0.0003,
+      "step": 17825
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.707257671699951e-08,
+      "loss": 0.0002,
+      "step": 17826
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 6.704822211397954e-08,
+      "loss": 0.0002,
+      "step": 17827
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.702386751095956e-08,
+      "loss": 0.0004,
+      "step": 17828
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.699951290793959e-08,
+      "loss": 0.0001,
+      "step": 17829
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.697515830491963e-08,
+      "loss": 0.0004,
+      "step": 17830
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.695080370189965e-08,
+      "loss": 0.0003,
+      "step": 17831
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.692644909887969e-08,
+      "loss": 0.0001,
+      "step": 17832
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.690209449585971e-08,
+      "loss": 0.0005,
+      "step": 17833
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.687773989283975e-08,
+      "loss": 0.0003,
+      "step": 17834
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.685338528981978e-08,
+      "loss": 0.0001,
+      "step": 17835
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.68290306867998e-08,
+      "loss": 0.0003,
+      "step": 17836
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 6.680467608377984e-08,
+      "loss": 0.0002,
+      "step": 17837
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.678032148075987e-08,
+      "loss": 0.0001,
+      "step": 17838
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.675596687773989e-08,
+      "loss": 0.0003,
+      "step": 17839
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.673161227471992e-08,
+      "loss": 0.0002,
+      "step": 17840
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.670725767169996e-08,
+      "loss": 0.0002,
+      "step": 17841
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.668290306867998e-08,
+      "loss": 0.0003,
+      "step": 17842
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.665854846566e-08,
+      "loss": 0.0001,
+      "step": 17843
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.663419386264003e-08,
+      "loss": 0.0002,
+      "step": 17844
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.660983925962007e-08,
+      "loss": 0.0002,
+      "step": 17845
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.65854846566001e-08,
+      "loss": 0.0005,
+      "step": 17846
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.656113005358012e-08,
+      "loss": 0.0003,
+      "step": 17847
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 6.653677545056014e-08,
+      "loss": 0.0001,
+      "step": 17848
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.651242084754018e-08,
+      "loss": 0.0001,
+      "step": 17849
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.648806624452021e-08,
+      "loss": 0.0004,
+      "step": 17850
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.646371164150025e-08,
+      "loss": 0.0003,
+      "step": 17851
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.643935703848027e-08,
+      "loss": 0.0004,
+      "step": 17852
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.641500243546031e-08,
+      "loss": 0.0002,
+      "step": 17853
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.639064783244034e-08,
+      "loss": 0.0001,
+      "step": 17854
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.636629322942036e-08,
+      "loss": 0.0004,
+      "step": 17855
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.634193862640039e-08,
+      "loss": 0.0004,
+      "step": 17856
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.631758402338042e-08,
+      "loss": 0.0002,
+      "step": 17857
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 6.629322942036045e-08,
+      "loss": 0.0002,
+      "step": 17858
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.626887481734047e-08,
+      "loss": 0.0001,
+      "step": 17859
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.62445202143205e-08,
+      "loss": 0.0002,
+      "step": 17860
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.622016561130054e-08,
+      "loss": 0.0001,
+      "step": 17861
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.619581100828056e-08,
+      "loss": 0.0001,
+      "step": 17862
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.617145640526059e-08,
+      "loss": 0.0001,
+      "step": 17863
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.614710180224061e-08,
+      "loss": 0.0004,
+      "step": 17864
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.612274719922065e-08,
+      "loss": 0.0003,
+      "step": 17865
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.609839259620068e-08,
+      "loss": 0.0002,
+      "step": 17866
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.60740379931807e-08,
+      "loss": 0.0003,
+      "step": 17867
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 6.604968339016073e-08,
+      "loss": 0.0001,
+      "step": 17868
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.602532878714077e-08,
+      "loss": 0.0002,
+      "step": 17869
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.600097418412079e-08,
+      "loss": 0.0004,
+      "step": 17870
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.597661958110083e-08,
+      "loss": 0.0002,
+      "step": 17871
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.595226497808085e-08,
+      "loss": 0.0003,
+      "step": 17872
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.592791037506089e-08,
+      "loss": 0.0004,
+      "step": 17873
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.590355577204092e-08,
+      "loss": 0.0004,
+      "step": 17874
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.587920116902094e-08,
+      "loss": 0.0002,
+      "step": 17875
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.585484656600097e-08,
+      "loss": 0.0002,
+      "step": 17876
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.5830491962981e-08,
+      "loss": 0.0003,
+      "step": 17877
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 6.580613735996103e-08,
+      "loss": 0.0002,
+      "step": 17878
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.578178275694106e-08,
+      "loss": 0.0003,
+      "step": 17879
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.575742815392108e-08,
+      "loss": 0.0002,
+      "step": 17880
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.573307355090112e-08,
+      "loss": 0.0003,
+      "step": 17881
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.570871894788115e-08,
+      "loss": 0.0002,
+      "step": 17882
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.568436434486117e-08,
+      "loss": 0.0002,
+      "step": 17883
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.56600097418412e-08,
+      "loss": 0.0003,
+      "step": 17884
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.563565513882123e-08,
+      "loss": 0.0002,
+      "step": 17885
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.561130053580126e-08,
+      "loss": 0.0001,
+      "step": 17886
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.558694593278128e-08,
+      "loss": 0.0002,
+      "step": 17887
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.556259132976131e-08,
+      "loss": 0.0002,
+      "step": 17888
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 6.553823672674135e-08,
+      "loss": 0.0001,
+      "step": 17889
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.551388212372139e-08,
+      "loss": 0.0002,
+      "step": 17890
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.548952752070141e-08,
+      "loss": 0.0002,
+      "step": 17891
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.546517291768145e-08,
+      "loss": 0.0002,
+      "step": 17892
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.544081831466148e-08,
+      "loss": 0.0003,
+      "step": 17893
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.54164637116415e-08,
+      "loss": 0.0003,
+      "step": 17894
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.539210910862153e-08,
+      "loss": 0.0003,
+      "step": 17895
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.536775450560156e-08,
+      "loss": 0.0001,
+      "step": 17896
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.534339990258159e-08,
+      "loss": 0.0003,
+      "step": 17897
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.531904529956161e-08,
+      "loss": 0.0002,
+      "step": 17898
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 6.529469069654164e-08,
+      "loss": 0.0001,
+      "step": 17899
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.527033609352168e-08,
+      "loss": 0.0002,
+      "step": 17900
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.52459814905017e-08,
+      "loss": 0.0002,
+      "step": 17901
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.522162688748173e-08,
+      "loss": 0.0002,
+      "step": 17902
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.519727228446175e-08,
+      "loss": 0.0002,
+      "step": 17903
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.517291768144179e-08,
+      "loss": 0.0002,
+      "step": 17904
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.514856307842182e-08,
+      "loss": 0.0002,
+      "step": 17905
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.512420847540184e-08,
+      "loss": 0.0005,
+      "step": 17906
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.509985387238187e-08,
+      "loss": 0.0002,
+      "step": 17907
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.50754992693619e-08,
+      "loss": 0.0001,
+      "step": 17908
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 6.505114466634194e-08,
+      "loss": 0.0002,
+      "step": 17909
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.502679006332197e-08,
+      "loss": 0.0002,
+      "step": 17910
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.5002435460302e-08,
+      "loss": 0.0002,
+      "step": 17911
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.497808085728203e-08,
+      "loss": 0.0002,
+      "step": 17912
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.495372625426206e-08,
+      "loss": 0.3339,
+      "step": 17913
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.492937165124208e-08,
+      "loss": 0.0002,
+      "step": 17914
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.490501704822211e-08,
+      "loss": 0.0003,
+      "step": 17915
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.488066244520215e-08,
+      "loss": 0.0002,
+      "step": 17916
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.485630784218217e-08,
+      "loss": 0.0001,
+      "step": 17917
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.48319532391622e-08,
+      "loss": 0.0002,
+      "step": 17918
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.480759863614222e-08,
+      "loss": 0.0003,
+      "step": 17919
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 6.478324403312226e-08,
+      "loss": 0.0001,
+      "step": 17920
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.475888943010229e-08,
+      "loss": 0.0002,
+      "step": 17921
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.473453482708231e-08,
+      "loss": 0.0002,
+      "step": 17922
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.471018022406234e-08,
+      "loss": 0.0003,
+      "step": 17923
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.468582562104237e-08,
+      "loss": 0.0001,
+      "step": 17924
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.46614710180224e-08,
+      "loss": 0.0003,
+      "step": 17925
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.463711641500242e-08,
+      "loss": 0.0003,
+      "step": 17926
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.461276181198245e-08,
+      "loss": 0.0002,
+      "step": 17927
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.45884072089625e-08,
+      "loss": 0.0001,
+      "step": 17928
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.456405260594253e-08,
+      "loss": 0.0003,
+      "step": 17929
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 6.453969800292255e-08,
+      "loss": 0.0002,
+      "step": 17930
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.451534339990258e-08,
+      "loss": 0.0001,
+      "step": 17931
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.449098879688262e-08,
+      "loss": 0.0004,
+      "step": 17932
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.446663419386264e-08,
+      "loss": 0.0003,
+      "step": 17933
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.444227959084267e-08,
+      "loss": 0.0002,
+      "step": 17934
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.441792498782269e-08,
+      "loss": 0.0002,
+      "step": 17935
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.439357038480273e-08,
+      "loss": 0.0002,
+      "step": 17936
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.436921578178275e-08,
+      "loss": 0.0001,
+      "step": 17937
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.434486117876278e-08,
+      "loss": 0.0001,
+      "step": 17938
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.43205065757428e-08,
+      "loss": 0.0001,
+      "step": 17939
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 6.429615197272284e-08,
+      "loss": 0.0001,
+      "step": 17940
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.427179736970287e-08,
+      "loss": 0.0001,
+      "step": 17941
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.42474427666829e-08,
+      "loss": 0.0002,
+      "step": 17942
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.422308816366292e-08,
+      "loss": 0.0002,
+      "step": 17943
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.419873356064296e-08,
+      "loss": 0.0002,
+      "step": 17944
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.417437895762298e-08,
+      "loss": 0.0002,
+      "step": 17945
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.415002435460301e-08,
+      "loss": 0.0003,
+      "step": 17946
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.412566975158306e-08,
+      "loss": 0.0003,
+      "step": 17947
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.410131514856308e-08,
+      "loss": 0.0002,
+      "step": 17948
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.407696054554311e-08,
+      "loss": 0.0002,
+      "step": 17949
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 6.405260594252313e-08,
+      "loss": 0.0003,
+      "step": 17950
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.402825133950317e-08,
+      "loss": 0.0002,
+      "step": 17951
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.40038967364832e-08,
+      "loss": 0.0002,
+      "step": 17952
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.397954213346322e-08,
+      "loss": 0.0002,
+      "step": 17953
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.395518753044325e-08,
+      "loss": 0.0002,
+      "step": 17954
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.393083292742329e-08,
+      "loss": 0.0003,
+      "step": 17955
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.390647832440331e-08,
+      "loss": 0.0001,
+      "step": 17956
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.388212372138334e-08,
+      "loss": 0.0002,
+      "step": 17957
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.385776911836336e-08,
+      "loss": 0.0002,
+      "step": 17958
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.38334145153434e-08,
+      "loss": 0.0002,
+      "step": 17959
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.380905991232343e-08,
+      "loss": 0.0003,
+      "step": 17960
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 6.378470530930345e-08,
+      "loss": 0.0003,
+      "step": 17961
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.376035070628348e-08,
+      "loss": 0.0002,
+      "step": 17962
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.373599610326351e-08,
+      "loss": 0.0002,
+      "step": 17963
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.371164150024354e-08,
+      "loss": 0.0002,
+      "step": 17964
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.368728689722357e-08,
+      "loss": 0.0002,
+      "step": 17965
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.36629322942036e-08,
+      "loss": 0.0003,
+      "step": 17966
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.363857769118364e-08,
+      "loss": 0.0002,
+      "step": 17967
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.361422308816367e-08,
+      "loss": 0.0002,
+      "step": 17968
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.358986848514369e-08,
+      "loss": 0.0003,
+      "step": 17969
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.356551388212372e-08,
+      "loss": 0.0003,
+      "step": 17970
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 6.354115927910376e-08,
+      "loss": 0.0002,
+      "step": 17971
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.351680467608378e-08,
+      "loss": 0.0002,
+      "step": 17972
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.34924500730638e-08,
+      "loss": 0.0003,
+      "step": 17973
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.346809547004383e-08,
+      "loss": 0.0004,
+      "step": 17974
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.344374086702387e-08,
+      "loss": 0.0003,
+      "step": 17975
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.34193862640039e-08,
+      "loss": 0.0003,
+      "step": 17976
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.339503166098392e-08,
+      "loss": 0.0003,
+      "step": 17977
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.337067705796395e-08,
+      "loss": 0.0001,
+      "step": 17978
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.334632245494398e-08,
+      "loss": 0.0002,
+      "step": 17979
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.332196785192401e-08,
+      "loss": 0.0002,
+      "step": 17980
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 6.329761324890403e-08,
+      "loss": 0.0003,
+      "step": 17981
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.327325864588406e-08,
+      "loss": 0.0002,
+      "step": 17982
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.32489040428641e-08,
+      "loss": 0.0002,
+      "step": 17983
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.322454943984412e-08,
+      "loss": 0.0003,
+      "step": 17984
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.320019483682415e-08,
+      "loss": 0.0003,
+      "step": 17985
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.317584023380419e-08,
+      "loss": 0.0002,
+      "step": 17986
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.315148563078422e-08,
+      "loss": 0.0004,
+      "step": 17987
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.312713102776425e-08,
+      "loss": 0.0003,
+      "step": 17988
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.310277642474428e-08,
+      "loss": 0.0002,
+      "step": 17989
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.30784218217243e-08,
+      "loss": 0.0002,
+      "step": 17990
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.305406721870434e-08,
+      "loss": 0.0003,
+      "step": 17991
+    },
+    {
+      "epoch": 17.48,
+      "learning_rate": 6.302971261568436e-08,
+      "loss": 0.0002,
+      "step": 17992
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.300535801266439e-08,
+      "loss": 0.0002,
+      "step": 17993
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.298100340964441e-08,
+      "loss": 0.0019,
+      "step": 17994
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.295664880662445e-08,
+      "loss": 0.0002,
+      "step": 17995
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.293229420360448e-08,
+      "loss": 0.0004,
+      "step": 17996
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.29079396005845e-08,
+      "loss": 0.0002,
+      "step": 17997
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.288358499756453e-08,
+      "loss": 0.0002,
+      "step": 17998
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.285923039454457e-08,
+      "loss": 0.0002,
+      "step": 17999
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.283487579152459e-08,
+      "loss": 0.0002,
+      "step": 18000
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.281052118850462e-08,
+      "loss": 0.0002,
+      "step": 18001
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 6.278616658548464e-08,
+      "loss": 0.0003,
+      "step": 18002
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.276181198246468e-08,
+      "loss": 0.0002,
+      "step": 18003
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.27374573794447e-08,
+      "loss": 0.0002,
+      "step": 18004
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.271310277642474e-08,
+      "loss": 0.0002,
+      "step": 18005
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.268874817340478e-08,
+      "loss": 0.0002,
+      "step": 18006
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.266439357038481e-08,
+      "loss": 0.0002,
+      "step": 18007
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.264003896736483e-08,
+      "loss": 0.0002,
+      "step": 18008
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.261568436434486e-08,
+      "loss": 0.0001,
+      "step": 18009
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.25913297613249e-08,
+      "loss": 0.0002,
+      "step": 18010
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.256697515830492e-08,
+      "loss": 0.0002,
+      "step": 18011
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 6.254262055528495e-08,
+      "loss": 0.0001,
+      "step": 18012
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.251826595226497e-08,
+      "loss": 0.0001,
+      "step": 18013
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.249391134924501e-08,
+      "loss": 0.0002,
+      "step": 18014
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.246955674622504e-08,
+      "loss": 0.0016,
+      "step": 18015
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.244520214320506e-08,
+      "loss": 0.0003,
+      "step": 18016
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.242084754018509e-08,
+      "loss": 0.0002,
+      "step": 18017
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.239649293716512e-08,
+      "loss": 0.0002,
+      "step": 18018
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.237213833414515e-08,
+      "loss": 0.0002,
+      "step": 18019
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.234778373112519e-08,
+      "loss": 0.0004,
+      "step": 18020
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.232342912810521e-08,
+      "loss": 0.0004,
+      "step": 18021
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 6.229907452508524e-08,
+      "loss": 0.0002,
+      "step": 18022
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.227471992206526e-08,
+      "loss": 0.0002,
+      "step": 18023
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.22503653190453e-08,
+      "loss": 0.0003,
+      "step": 18024
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.222601071602533e-08,
+      "loss": 0.0002,
+      "step": 18025
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.220165611300535e-08,
+      "loss": 0.0002,
+      "step": 18026
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.217730150998538e-08,
+      "loss": 0.0002,
+      "step": 18027
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.215294690696542e-08,
+      "loss": 0.0003,
+      "step": 18028
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.212859230394544e-08,
+      "loss": 0.0003,
+      "step": 18029
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.210423770092548e-08,
+      "loss": 0.9836,
+      "step": 18030
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.20798830979055e-08,
+      "loss": 0.0002,
+      "step": 18031
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.205552849488553e-08,
+      "loss": 0.0003,
+      "step": 18032
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 6.203117389186555e-08,
+      "loss": 0.0002,
+      "step": 18033
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.200681928884559e-08,
+      "loss": 0.0002,
+      "step": 18034
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.198246468582562e-08,
+      "loss": 0.0003,
+      "step": 18035
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.195811008280564e-08,
+      "loss": 0.0002,
+      "step": 18036
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.193375547978567e-08,
+      "loss": 0.0002,
+      "step": 18037
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.190940087676571e-08,
+      "loss": 0.0001,
+      "step": 18038
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.188504627374573e-08,
+      "loss": 0.0002,
+      "step": 18039
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.186069167072577e-08,
+      "loss": 0.0005,
+      "step": 18040
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.18363370677058e-08,
+      "loss": 0.0003,
+      "step": 18041
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.181198246468582e-08,
+      "loss": 0.0001,
+      "step": 18042
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 6.178762786166585e-08,
+      "loss": 0.0002,
+      "step": 18043
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.176327325864588e-08,
+      "loss": 0.0003,
+      "step": 18044
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.173891865562591e-08,
+      "loss": 0.0002,
+      "step": 18045
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.171456405260593e-08,
+      "loss": 0.0003,
+      "step": 18046
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.169020944958596e-08,
+      "loss": 0.0002,
+      "step": 18047
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.1665854846566e-08,
+      "loss": 0.0004,
+      "step": 18048
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.164150024354602e-08,
+      "loss": 0.0003,
+      "step": 18049
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.161714564052606e-08,
+      "loss": 0.0003,
+      "step": 18050
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.159279103750609e-08,
+      "loss": 0.0002,
+      "step": 18051
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.156843643448611e-08,
+      "loss": 0.0001,
+      "step": 18052
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 6.154408183146614e-08,
+      "loss": 0.0002,
+      "step": 18053
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.151972722844618e-08,
+      "loss": 0.0002,
+      "step": 18054
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.14953726254262e-08,
+      "loss": 0.0004,
+      "step": 18055
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.147101802240623e-08,
+      "loss": 0.0001,
+      "step": 18056
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.144666341938626e-08,
+      "loss": 0.0002,
+      "step": 18057
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.142230881636629e-08,
+      "loss": 0.0002,
+      "step": 18058
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.139795421334633e-08,
+      "loss": 0.0004,
+      "step": 18059
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.137359961032635e-08,
+      "loss": 0.0002,
+      "step": 18060
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.134924500730638e-08,
+      "loss": 0.0002,
+      "step": 18061
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.13248904042864e-08,
+      "loss": 0.0001,
+      "step": 18062
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.130053580126644e-08,
+      "loss": 0.0002,
+      "step": 18063
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 6.127618119824647e-08,
+      "loss": 0.0002,
+      "step": 18064
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.125182659522649e-08,
+      "loss": 0.0004,
+      "step": 18065
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.122747199220652e-08,
+      "loss": 0.0003,
+      "step": 18066
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.120311738918656e-08,
+      "loss": 0.0004,
+      "step": 18067
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.117876278616658e-08,
+      "loss": 0.0003,
+      "step": 18068
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.115440818314662e-08,
+      "loss": 0.0003,
+      "step": 18069
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.113005358012664e-08,
+      "loss": 0.0002,
+      "step": 18070
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.110569897710667e-08,
+      "loss": 0.0001,
+      "step": 18071
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.10813443740867e-08,
+      "loss": 0.0002,
+      "step": 18072
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.105698977106673e-08,
+      "loss": 0.0003,
+      "step": 18073
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 6.103263516804676e-08,
+      "loss": 0.0001,
+      "step": 18074
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.100828056502678e-08,
+      "loss": 0.088,
+      "step": 18075
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.098392596200682e-08,
+      "loss": 0.0004,
+      "step": 18076
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.095957135898685e-08,
+      "loss": 0.0004,
+      "step": 18077
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.093521675596687e-08,
+      "loss": 0.0002,
+      "step": 18078
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.091086215294691e-08,
+      "loss": 0.0002,
+      "step": 18079
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.088650754992694e-08,
+      "loss": 0.0003,
+      "step": 18080
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.086215294690696e-08,
+      "loss": 0.2521,
+      "step": 18081
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.083779834388699e-08,
+      "loss": 0.0002,
+      "step": 18082
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.081344374086702e-08,
+      "loss": 0.0002,
+      "step": 18083
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 6.078908913784705e-08,
+      "loss": 0.0002,
+      "step": 18084
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.076473453482707e-08,
+      "loss": 0.0001,
+      "step": 18085
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.074037993180711e-08,
+      "loss": 0.0002,
+      "step": 18086
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.071602532878714e-08,
+      "loss": 0.0003,
+      "step": 18087
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.069167072576716e-08,
+      "loss": 0.0002,
+      "step": 18088
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.06673161227472e-08,
+      "loss": 0.0003,
+      "step": 18089
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.064296151972723e-08,
+      "loss": 0.0003,
+      "step": 18090
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.061860691670725e-08,
+      "loss": 0.0009,
+      "step": 18091
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.059425231368728e-08,
+      "loss": 0.0002,
+      "step": 18092
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.056989771066732e-08,
+      "loss": 0.0002,
+      "step": 18093
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 6.054554310764734e-08,
+      "loss": 0.0002,
+      "step": 18094
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.052118850462737e-08,
+      "loss": 0.0004,
+      "step": 18095
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.04968339016074e-08,
+      "loss": 0.0002,
+      "step": 18096
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.047247929858743e-08,
+      "loss": 0.0002,
+      "step": 18097
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.044812469556745e-08,
+      "loss": 0.0003,
+      "step": 18098
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.042377009254749e-08,
+      "loss": 0.0003,
+      "step": 18099
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.039941548952752e-08,
+      "loss": 0.0002,
+      "step": 18100
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.037506088650754e-08,
+      "loss": 0.0044,
+      "step": 18101
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.035070628348757e-08,
+      "loss": 0.0002,
+      "step": 18102
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.032635168046761e-08,
+      "loss": 0.0002,
+      "step": 18103
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.030199707744763e-08,
+      "loss": 0.0002,
+      "step": 18104
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 6.027764247442767e-08,
+      "loss": 0.0001,
+      "step": 18105
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.02532878714077e-08,
+      "loss": 0.0002,
+      "step": 18106
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.022893326838772e-08,
+      "loss": 0.0002,
+      "step": 18107
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.020457866536776e-08,
+      "loss": 0.0003,
+      "step": 18108
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.018022406234778e-08,
+      "loss": 0.0002,
+      "step": 18109
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.015586945932781e-08,
+      "loss": 0.0003,
+      "step": 18110
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.013151485630783e-08,
+      "loss": 0.0003,
+      "step": 18111
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.010716025328787e-08,
+      "loss": 0.0002,
+      "step": 18112
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.00828056502679e-08,
+      "loss": 0.0004,
+      "step": 18113
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.005845104724792e-08,
+      "loss": 0.0004,
+      "step": 18114
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 6.003409644422796e-08,
+      "loss": 0.0003,
+      "step": 18115
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 6.000974184120799e-08,
+      "loss": 0.0002,
+      "step": 18116
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.998538723818801e-08,
+      "loss": 0.0002,
+      "step": 18117
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.996103263516805e-08,
+      "loss": 0.0003,
+      "step": 18118
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.993667803214808e-08,
+      "loss": 0.0002,
+      "step": 18119
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.99123234291281e-08,
+      "loss": 0.0004,
+      "step": 18120
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.988796882610813e-08,
+      "loss": 0.0002,
+      "step": 18121
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.986361422308816e-08,
+      "loss": 0.0003,
+      "step": 18122
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.983925962006819e-08,
+      "loss": 0.0003,
+      "step": 18123
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.981490501704823e-08,
+      "loss": 0.0002,
+      "step": 18124
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 5.979055041402825e-08,
+      "loss": 0.0002,
+      "step": 18125
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.976619581100828e-08,
+      "loss": 0.0002,
+      "step": 18126
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.97418412079883e-08,
+      "loss": 0.0002,
+      "step": 18127
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.971748660496834e-08,
+      "loss": 0.0001,
+      "step": 18128
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.969313200194837e-08,
+      "loss": 0.0001,
+      "step": 18129
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.966877739892839e-08,
+      "loss": 0.0002,
+      "step": 18130
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.964442279590842e-08,
+      "loss": 0.0002,
+      "step": 18131
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.962006819288846e-08,
+      "loss": 0.0005,
+      "step": 18132
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.959571358986848e-08,
+      "loss": 0.0003,
+      "step": 18133
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.957135898684851e-08,
+      "loss": 0.0003,
+      "step": 18134
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.9547004383828545e-08,
+      "loss": 0.0001,
+      "step": 18135
+    },
+    {
+      "epoch": 17.62,
+      "learning_rate": 5.952264978080857e-08,
+      "loss": 0.0002,
+      "step": 18136
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.94982951777886e-08,
+      "loss": 0.0003,
+      "step": 18137
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.947394057476863e-08,
+      "loss": 0.0002,
+      "step": 18138
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.944958597174866e-08,
+      "loss": 0.0027,
+      "step": 18139
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.9425231368728684e-08,
+      "loss": 0.0002,
+      "step": 18140
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.9400876765708716e-08,
+      "loss": 0.0002,
+      "step": 18141
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.937652216268874e-08,
+      "loss": 0.0002,
+      "step": 18142
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.935216755966877e-08,
+      "loss": 0.0002,
+      "step": 18143
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.9327812956648804e-08,
+      "loss": 0.0003,
+      "step": 18144
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.9303458353628836e-08,
+      "loss": 0.0001,
+      "step": 18145
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 5.927910375060886e-08,
+      "loss": 0.0001,
+      "step": 18146
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.925474914758889e-08,
+      "loss": 0.0003,
+      "step": 18147
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.923039454456892e-08,
+      "loss": 0.0002,
+      "step": 18148
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.920603994154895e-08,
+      "loss": 0.0003,
+      "step": 18149
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.9181685338528975e-08,
+      "loss": 0.0002,
+      "step": 18150
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.915733073550901e-08,
+      "loss": 0.0008,
+      "step": 18151
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.913297613248903e-08,
+      "loss": 0.0004,
+      "step": 18152
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.910862152946907e-08,
+      "loss": 0.0003,
+      "step": 18153
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.9084266926449096e-08,
+      "loss": 0.0002,
+      "step": 18154
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.905991232342913e-08,
+      "loss": 0.0003,
+      "step": 18155
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 5.903555772040915e-08,
+      "loss": 0.0002,
+      "step": 18156
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.9011203117389184e-08,
+      "loss": 0.0002,
+      "step": 18157
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.898684851436921e-08,
+      "loss": 0.0004,
+      "step": 18158
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.896249391134924e-08,
+      "loss": 0.0004,
+      "step": 18159
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.8938139308329267e-08,
+      "loss": 0.0001,
+      "step": 18160
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.89137847053093e-08,
+      "loss": 0.0003,
+      "step": 18161
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.8889430102289323e-08,
+      "loss": 0.0004,
+      "step": 18162
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.886507549926936e-08,
+      "loss": 0.0002,
+      "step": 18163
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.8840720896249394e-08,
+      "loss": 0.0002,
+      "step": 18164
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.881636629322942e-08,
+      "loss": 0.0003,
+      "step": 18165
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 5.879201169020945e-08,
+      "loss": 0.0001,
+      "step": 18166
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.8767657087189476e-08,
+      "loss": 0.0001,
+      "step": 18167
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.874330248416951e-08,
+      "loss": 0.0003,
+      "step": 18168
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.871894788114953e-08,
+      "loss": 0.0003,
+      "step": 18169
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.8694593278129565e-08,
+      "loss": 0.0001,
+      "step": 18170
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.867023867510959e-08,
+      "loss": 0.0002,
+      "step": 18171
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.864588407208963e-08,
+      "loss": 0.0002,
+      "step": 18172
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.862152946906965e-08,
+      "loss": 0.0003,
+      "step": 18173
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.8597174866049685e-08,
+      "loss": 0.0003,
+      "step": 18174
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.857282026302971e-08,
+      "loss": 0.0003,
+      "step": 18175
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.854846566000974e-08,
+      "loss": 1.8607,
+      "step": 18176
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 5.852411105698977e-08,
+      "loss": 0.0004,
+      "step": 18177
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.84997564539698e-08,
+      "loss": 0.0003,
+      "step": 18178
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.8475401850949824e-08,
+      "loss": 0.0003,
+      "step": 18179
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.8451047247929856e-08,
+      "loss": 0.0003,
+      "step": 18180
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.842669264490988e-08,
+      "loss": 0.0002,
+      "step": 18181
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.840233804188992e-08,
+      "loss": 0.0002,
+      "step": 18182
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.8377983438869945e-08,
+      "loss": 0.0001,
+      "step": 18183
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.8353628835849976e-08,
+      "loss": 0.0002,
+      "step": 18184
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.832927423283e-08,
+      "loss": 0.0002,
+      "step": 18185
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.8304919629810033e-08,
+      "loss": 0.0002,
+      "step": 18186
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 5.828056502679006e-08,
+      "loss": 0.0001,
+      "step": 18187
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.825621042377009e-08,
+      "loss": 0.0002,
+      "step": 18188
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.8231855820750115e-08,
+      "loss": 0.0002,
+      "step": 18189
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.820750121773015e-08,
+      "loss": 0.0003,
+      "step": 18190
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.818314661471018e-08,
+      "loss": 0.0002,
+      "step": 18191
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.815879201169021e-08,
+      "loss": 0.0004,
+      "step": 18192
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.8134437408670236e-08,
+      "loss": 0.0004,
+      "step": 18193
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.811008280565027e-08,
+      "loss": 0.0003,
+      "step": 18194
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.808572820263029e-08,
+      "loss": 0.0002,
+      "step": 18195
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.8061373599610325e-08,
+      "loss": 0.0005,
+      "step": 18196
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 5.803701899659035e-08,
+      "loss": 0.0003,
+      "step": 18197
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.801266439357038e-08,
+      "loss": 0.0002,
+      "step": 18198
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.798830979055041e-08,
+      "loss": 0.0002,
+      "step": 18199
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.796395518753044e-08,
+      "loss": 0.0002,
+      "step": 18200
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.793960058451047e-08,
+      "loss": 0.0003,
+      "step": 18201
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.79152459814905e-08,
+      "loss": 0.0002,
+      "step": 18202
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.789089137847053e-08,
+      "loss": 0.0002,
+      "step": 18203
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.786653677545056e-08,
+      "loss": 0.0004,
+      "step": 18204
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.7842182172430584e-08,
+      "loss": 0.0003,
+      "step": 18205
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.7817827569410616e-08,
+      "loss": 0.0002,
+      "step": 18206
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.779347296639064e-08,
+      "loss": 0.0003,
+      "step": 18207
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 5.776911836337067e-08,
+      "loss": 0.0002,
+      "step": 18208
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.77447637603507e-08,
+      "loss": 0.0003,
+      "step": 18209
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.772040915733073e-08,
+      "loss": 0.0001,
+      "step": 18210
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.769605455431076e-08,
+      "loss": 0.0002,
+      "step": 18211
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.7671699951290794e-08,
+      "loss": 0.0002,
+      "step": 18212
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.764734534827082e-08,
+      "loss": 0.0002,
+      "step": 18213
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.762299074525085e-08,
+      "loss": 0.0002,
+      "step": 18214
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.7598636142230876e-08,
+      "loss": 0.0004,
+      "step": 18215
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.757428153921091e-08,
+      "loss": 0.001,
+      "step": 18216
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.754992693619094e-08,
+      "loss": 0.0003,
+      "step": 18217
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 5.7525572333170964e-08,
+      "loss": 0.0003,
+      "step": 18218
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.7501217730150996e-08,
+      "loss": 0.0003,
+      "step": 18219
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.747686312713103e-08,
+      "loss": 0.0002,
+      "step": 18220
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.745250852411106e-08,
+      "loss": 0.0002,
+      "step": 18221
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.7428153921091085e-08,
+      "loss": 1.7421,
+      "step": 18222
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.7403799318071117e-08,
+      "loss": 0.0004,
+      "step": 18223
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.737944471505114e-08,
+      "loss": 0.0002,
+      "step": 18224
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.7355090112031174e-08,
+      "loss": 0.0002,
+      "step": 18225
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.73307355090112e-08,
+      "loss": 0.0003,
+      "step": 18226
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.730638090599123e-08,
+      "loss": 0.0002,
+      "step": 18227
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 5.7282026302971256e-08,
+      "loss": 0.0001,
+      "step": 18228
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.725767169995129e-08,
+      "loss": 0.0002,
+      "step": 18229
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.723331709693132e-08,
+      "loss": 0.0002,
+      "step": 18230
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.720896249391135e-08,
+      "loss": 0.0002,
+      "step": 18231
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.7184607890891376e-08,
+      "loss": 0.0004,
+      "step": 18232
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.716025328787141e-08,
+      "loss": 0.0002,
+      "step": 18233
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.713589868485143e-08,
+      "loss": 0.0002,
+      "step": 18234
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.7111544081831465e-08,
+      "loss": 0.0002,
+      "step": 18235
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.708718947881149e-08,
+      "loss": 0.0002,
+      "step": 18236
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.706283487579152e-08,
+      "loss": 0.0004,
+      "step": 18237
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.703848027277155e-08,
+      "loss": 0.0002,
+      "step": 18238
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 5.7014125669751585e-08,
+      "loss": 0.0002,
+      "step": 18239
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.698977106673161e-08,
+      "loss": 0.0002,
+      "step": 18240
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.696541646371164e-08,
+      "loss": 0.0003,
+      "step": 18241
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.694106186069167e-08,
+      "loss": 0.0003,
+      "step": 18242
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.69167072576717e-08,
+      "loss": 0.0002,
+      "step": 18243
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.6892352654651725e-08,
+      "loss": 0.0003,
+      "step": 18244
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.6867998051631756e-08,
+      "loss": 0.0003,
+      "step": 18245
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.684364344861178e-08,
+      "loss": 0.0002,
+      "step": 18246
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.681928884559181e-08,
+      "loss": 0.0001,
+      "step": 18247
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.679493424257184e-08,
+      "loss": 0.0002,
+      "step": 18248
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 5.677057963955188e-08,
+      "loss": 0.0003,
+      "step": 18249
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.67462250365319e-08,
+      "loss": 0.0003,
+      "step": 18250
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.6721870433511934e-08,
+      "loss": 0.0002,
+      "step": 18251
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.669751583049196e-08,
+      "loss": 0.0002,
+      "step": 18252
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.667316122747199e-08,
+      "loss": 0.0002,
+      "step": 18253
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.6648806624452016e-08,
+      "loss": 0.0002,
+      "step": 18254
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.662445202143205e-08,
+      "loss": 0.0002,
+      "step": 18255
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.660009741841207e-08,
+      "loss": 0.0003,
+      "step": 18256
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.6575742815392105e-08,
+      "loss": 0.0003,
+      "step": 18257
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.6551388212372136e-08,
+      "loss": 0.0002,
+      "step": 18258
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 5.652703360935217e-08,
+      "loss": 0.0087,
+      "step": 18259
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.6502679006332193e-08,
+      "loss": 0.0003,
+      "step": 18260
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.6478324403312225e-08,
+      "loss": 0.0002,
+      "step": 18261
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.645396980029225e-08,
+      "loss": 0.0004,
+      "step": 18262
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.642961519727228e-08,
+      "loss": 0.0003,
+      "step": 18263
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.640526059425231e-08,
+      "loss": 0.0002,
+      "step": 18264
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.638090599123234e-08,
+      "loss": 0.0001,
+      "step": 18265
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.6356551388212364e-08,
+      "loss": 0.0002,
+      "step": 18266
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.6332196785192396e-08,
+      "loss": 0.0004,
+      "step": 18267
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.6307842182172434e-08,
+      "loss": 0.0002,
+      "step": 18268
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 5.628348757915246e-08,
+      "loss": 0.001,
+      "step": 18269
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.625913297613249e-08,
+      "loss": 0.0002,
+      "step": 18270
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.6234778373112516e-08,
+      "loss": 0.0002,
+      "step": 18271
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.621042377009255e-08,
+      "loss": 0.0003,
+      "step": 18272
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.6186069167072573e-08,
+      "loss": 0.0002,
+      "step": 18273
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.6161714564052605e-08,
+      "loss": 0.0002,
+      "step": 18274
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.613735996103263e-08,
+      "loss": 0.0003,
+      "step": 18275
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.611300535801266e-08,
+      "loss": 0.0001,
+      "step": 18276
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.608865075499269e-08,
+      "loss": 0.0002,
+      "step": 18277
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.6064296151972726e-08,
+      "loss": 0.0001,
+      "step": 18278
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.603994154895275e-08,
+      "loss": 0.0003,
+      "step": 18279
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 5.601558694593278e-08,
+      "loss": 0.0002,
+      "step": 18280
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.599123234291281e-08,
+      "loss": 0.0002,
+      "step": 18281
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.596687773989284e-08,
+      "loss": 0.0002,
+      "step": 18282
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.5942523136872865e-08,
+      "loss": 0.0002,
+      "step": 18283
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.5918168533852897e-08,
+      "loss": 0.0002,
+      "step": 18284
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.589381393083292e-08,
+      "loss": 0.0003,
+      "step": 18285
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.5869459327812954e-08,
+      "loss": 0.0003,
+      "step": 18286
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.5845104724792985e-08,
+      "loss": 0.0002,
+      "step": 18287
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.582075012177302e-08,
+      "loss": 0.0005,
+      "step": 18288
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.579639551875304e-08,
+      "loss": 0.0004,
+      "step": 18289
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 5.5772040915733074e-08,
+      "loss": 0.0001,
+      "step": 18290
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.57476863127131e-08,
+      "loss": 0.0003,
+      "step": 18291
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.572333170969313e-08,
+      "loss": 0.0002,
+      "step": 18292
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.5698977106673156e-08,
+      "loss": 0.0001,
+      "step": 18293
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.567462250365319e-08,
+      "loss": 0.0002,
+      "step": 18294
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.565026790063321e-08,
+      "loss": 0.0002,
+      "step": 18295
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.5625913297613245e-08,
+      "loss": 0.0004,
+      "step": 18296
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.5601558694593277e-08,
+      "loss": 0.0003,
+      "step": 18297
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.557720409157331e-08,
+      "loss": 0.0003,
+      "step": 18298
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.5552849488553334e-08,
+      "loss": 0.0002,
+      "step": 18299
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 5.5528494885533365e-08,
+      "loss": 0.0003,
+      "step": 18300
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.550414028251339e-08,
+      "loss": 0.0001,
+      "step": 18301
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.547978567949342e-08,
+      "loss": 0.0002,
+      "step": 18302
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.545543107647345e-08,
+      "loss": 0.0003,
+      "step": 18303
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.543107647345348e-08,
+      "loss": 0.0003,
+      "step": 18304
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.5406721870433504e-08,
+      "loss": 1.7769,
+      "step": 18305
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.538236726741354e-08,
+      "loss": 0.0001,
+      "step": 18306
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.535801266439357e-08,
+      "loss": 0.0004,
+      "step": 18307
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.53336580613736e-08,
+      "loss": 0.0002,
+      "step": 18308
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.5309303458353625e-08,
+      "loss": 0.0001,
+      "step": 18309
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.528494885533366e-08,
+      "loss": 0.0004,
+      "step": 18310
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 5.526059425231368e-08,
+      "loss": 0.0004,
+      "step": 18311
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.5236239649293714e-08,
+      "loss": 0.0003,
+      "step": 18312
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.521188504627374e-08,
+      "loss": 0.0003,
+      "step": 18313
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.518753044325377e-08,
+      "loss": 0.0003,
+      "step": 18314
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.5163175840233796e-08,
+      "loss": 0.0002,
+      "step": 18315
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.5138821237213834e-08,
+      "loss": 0.0002,
+      "step": 18316
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.511446663419386e-08,
+      "loss": 0.0001,
+      "step": 18317
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.509011203117389e-08,
+      "loss": 0.0002,
+      "step": 18318
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.5065757428153916e-08,
+      "loss": 0.0002,
+      "step": 18319
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.504140282513395e-08,
+      "loss": 0.0003,
+      "step": 18320
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 5.501704822211397e-08,
+      "loss": 0.0002,
+      "step": 18321
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.4992693619094005e-08,
+      "loss": 0.0003,
+      "step": 18322
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.496833901607403e-08,
+      "loss": 0.0004,
+      "step": 18323
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.494398441305406e-08,
+      "loss": 0.0003,
+      "step": 18324
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.49196298100341e-08,
+      "loss": 0.0003,
+      "step": 18325
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.4895275207014126e-08,
+      "loss": 0.0002,
+      "step": 18326
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.487092060399416e-08,
+      "loss": 0.0001,
+      "step": 18327
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.484656600097418e-08,
+      "loss": 0.0003,
+      "step": 18328
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.4822211397954214e-08,
+      "loss": 0.0002,
+      "step": 18329
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.479785679493424e-08,
+      "loss": 0.0001,
+      "step": 18330
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 5.477350219191427e-08,
+      "loss": 0.0003,
+      "step": 18331
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.4749147588894296e-08,
+      "loss": 0.0002,
+      "step": 18332
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.472479298587433e-08,
+      "loss": 0.0002,
+      "step": 18333
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.4700438382854353e-08,
+      "loss": 0.0002,
+      "step": 18334
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.467608377983439e-08,
+      "loss": 0.0002,
+      "step": 18335
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.465172917681442e-08,
+      "loss": 0.0001,
+      "step": 18336
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.462737457379445e-08,
+      "loss": 0.0001,
+      "step": 18337
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.4603019970774474e-08,
+      "loss": 0.0005,
+      "step": 18338
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.4578665367754506e-08,
+      "loss": 0.0002,
+      "step": 18339
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.455431076473453e-08,
+      "loss": 0.0003,
+      "step": 18340
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 5.452995616171456e-08,
+      "loss": 0.0001,
+      "step": 18341
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.450560155869459e-08,
+      "loss": 0.0003,
+      "step": 18342
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.448124695567462e-08,
+      "loss": 0.0002,
+      "step": 18343
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.4456892352654645e-08,
+      "loss": 0.0002,
+      "step": 18344
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.443253774963468e-08,
+      "loss": 0.0002,
+      "step": 18345
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.440818314661471e-08,
+      "loss": 0.0002,
+      "step": 18346
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.438382854359474e-08,
+      "loss": 0.0004,
+      "step": 18347
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.4359473940574765e-08,
+      "loss": 0.0001,
+      "step": 18348
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.43351193375548e-08,
+      "loss": 0.0004,
+      "step": 18349
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.431076473453482e-08,
+      "loss": 0.0002,
+      "step": 18350
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.4286410131514854e-08,
+      "loss": 0.0002,
+      "step": 18351
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 5.426205552849488e-08,
+      "loss": 0.0001,
+      "step": 18352
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.423770092547491e-08,
+      "loss": 0.0003,
+      "step": 18353
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.421334632245494e-08,
+      "loss": 0.0003,
+      "step": 18354
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.4188991719434974e-08,
+      "loss": 0.0005,
+      "step": 18355
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.4164637116415e-08,
+      "loss": 0.0004,
+      "step": 18356
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.414028251339503e-08,
+      "loss": 0.0002,
+      "step": 18357
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.4115927910375057e-08,
+      "loss": 0.0001,
+      "step": 18358
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.409157330735509e-08,
+      "loss": 0.0002,
+      "step": 18359
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.4067218704335114e-08,
+      "loss": 0.0001,
+      "step": 18360
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.4042864101315145e-08,
+      "loss": 0.0002,
+      "step": 18361
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 5.401850949829517e-08,
+      "loss": 0.0004,
+      "step": 18362
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.39941548952752e-08,
+      "loss": 0.0002,
+      "step": 18363
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.3969800292255234e-08,
+      "loss": 0.0003,
+      "step": 18364
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.3945445689235266e-08,
+      "loss": 0.0003,
+      "step": 18365
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.392109108621529e-08,
+      "loss": 0.0002,
+      "step": 18366
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.389673648319532e-08,
+      "loss": 0.0002,
+      "step": 18367
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.387238188017535e-08,
+      "loss": 0.0002,
+      "step": 18368
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.384802727715538e-08,
+      "loss": 0.0003,
+      "step": 18369
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.3823672674135405e-08,
+      "loss": 0.0003,
+      "step": 18370
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.3799318071115437e-08,
+      "loss": 0.0001,
+      "step": 18371
+    },
+    {
+      "epoch": 17.85,
+      "learning_rate": 5.377496346809546e-08,
+      "loss": 0.0002,
+      "step": 18372
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.37506088650755e-08,
+      "loss": 0.0002,
+      "step": 18373
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.3726254262055525e-08,
+      "loss": 0.0004,
+      "step": 18374
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.370189965903556e-08,
+      "loss": 0.0002,
+      "step": 18375
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.367754505601559e-08,
+      "loss": 0.0002,
+      "step": 18376
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.3653190452995614e-08,
+      "loss": 0.0002,
+      "step": 18377
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.3628835849975646e-08,
+      "loss": 0.0003,
+      "step": 18378
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.360448124695567e-08,
+      "loss": 0.0003,
+      "step": 18379
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.35801266439357e-08,
+      "loss": 0.0001,
+      "step": 18380
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.355577204091573e-08,
+      "loss": 0.0001,
+      "step": 18381
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.353141743789576e-08,
+      "loss": 0.0002,
+      "step": 18382
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 5.350706283487579e-08,
+      "loss": 0.0002,
+      "step": 18383
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.3482708231855823e-08,
+      "loss": 0.0001,
+      "step": 18384
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.345835362883585e-08,
+      "loss": 0.0002,
+      "step": 18385
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.343399902581588e-08,
+      "loss": 0.0002,
+      "step": 18386
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.3409644422795906e-08,
+      "loss": 2.5061,
+      "step": 18387
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.338528981977594e-08,
+      "loss": 0.0002,
+      "step": 18388
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.336093521675596e-08,
+      "loss": 0.0003,
+      "step": 18389
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.3336580613735994e-08,
+      "loss": 0.0001,
+      "step": 18390
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.331222601071602e-08,
+      "loss": 0.0003,
+      "step": 18391
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.328787140769606e-08,
+      "loss": 0.0002,
+      "step": 18392
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 5.326351680467608e-08,
+      "loss": 0.0002,
+      "step": 18393
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.3239162201656115e-08,
+      "loss": 0.0002,
+      "step": 18394
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.321480759863614e-08,
+      "loss": 0.0004,
+      "step": 18395
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.319045299561617e-08,
+      "loss": 0.0002,
+      "step": 18396
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.31660983925962e-08,
+      "loss": 0.0003,
+      "step": 18397
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.314174378957623e-08,
+      "loss": 0.0002,
+      "step": 18398
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.3117389186556254e-08,
+      "loss": 0.0002,
+      "step": 18399
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.3093034583536286e-08,
+      "loss": 0.0003,
+      "step": 18400
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.306867998051631e-08,
+      "loss": 0.0002,
+      "step": 18401
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.304432537749635e-08,
+      "loss": 0.0002,
+      "step": 18402
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 5.3019970774476374e-08,
+      "loss": 0.0003,
+      "step": 18403
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.2995616171456406e-08,
+      "loss": 0.0003,
+      "step": 18404
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.297126156843643e-08,
+      "loss": 0.0002,
+      "step": 18405
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.294690696541646e-08,
+      "loss": 0.0002,
+      "step": 18406
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.292255236239649e-08,
+      "loss": 0.0003,
+      "step": 18407
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.289819775937652e-08,
+      "loss": 0.0004,
+      "step": 18408
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.2873843156356545e-08,
+      "loss": 0.0004,
+      "step": 18409
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.284948855333658e-08,
+      "loss": 0.0003,
+      "step": 18410
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.28251339503166e-08,
+      "loss": 0.0001,
+      "step": 18411
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.280077934729664e-08,
+      "loss": 0.0003,
+      "step": 18412
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 5.2776424744276666e-08,
+      "loss": 0.0006,
+      "step": 18413
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.27520701412567e-08,
+      "loss": 0.0003,
+      "step": 18414
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.272771553823672e-08,
+      "loss": 0.0002,
+      "step": 18415
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.2703360935216754e-08,
+      "loss": 0.0005,
+      "step": 18416
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.267900633219678e-08,
+      "loss": 0.0002,
+      "step": 18417
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.265465172917681e-08,
+      "loss": 0.0003,
+      "step": 18418
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.2630297126156837e-08,
+      "loss": 0.0002,
+      "step": 18419
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.260594252313687e-08,
+      "loss": 0.0002,
+      "step": 18420
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.25815879201169e-08,
+      "loss": 0.0001,
+      "step": 18421
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.255723331709693e-08,
+      "loss": 0.0002,
+      "step": 18422
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.253287871407696e-08,
+      "loss": 0.0003,
+      "step": 18423
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 5.250852411105699e-08,
+      "loss": 0.0002,
+      "step": 18424
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.2484169508037014e-08,
+      "loss": 0.0003,
+      "step": 18425
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.2459814905017046e-08,
+      "loss": 0.0003,
+      "step": 18426
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.243546030199707e-08,
+      "loss": 0.0002,
+      "step": 18427
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.24111056989771e-08,
+      "loss": 0.0002,
+      "step": 18428
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.238675109595713e-08,
+      "loss": 0.0003,
+      "step": 18429
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.236239649293716e-08,
+      "loss": 0.0003,
+      "step": 18430
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.23380418899172e-08,
+      "loss": 0.0003,
+      "step": 18431
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.231368728689722e-08,
+      "loss": 0.0002,
+      "step": 18432
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.2289332683877255e-08,
+      "loss": 0.0002,
+      "step": 18433
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 5.226497808085728e-08,
+      "loss": 0.0002,
+      "step": 18434
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.224062347783731e-08,
+      "loss": 0.0001,
+      "step": 18435
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.221626887481734e-08,
+      "loss": 0.0003,
+      "step": 18436
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.219191427179737e-08,
+      "loss": 0.0001,
+      "step": 18437
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.2167559668777394e-08,
+      "loss": 0.0003,
+      "step": 18438
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.2143205065757426e-08,
+      "loss": 0.0002,
+      "step": 18439
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.211885046273746e-08,
+      "loss": 0.0002,
+      "step": 18440
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.209449585971749e-08,
+      "loss": 0.0004,
+      "step": 18441
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.2070141256697515e-08,
+      "loss": 0.0002,
+      "step": 18442
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.2045786653677546e-08,
+      "loss": 0.0001,
+      "step": 18443
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 5.202143205065757e-08,
+      "loss": 0.0003,
+      "step": 18444
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.1997077447637603e-08,
+      "loss": 0.0002,
+      "step": 18445
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.197272284461763e-08,
+      "loss": 0.0002,
+      "step": 18446
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.194836824159766e-08,
+      "loss": 0.0004,
+      "step": 18447
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.1924013638577685e-08,
+      "loss": 0.0004,
+      "step": 18448
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.189965903555772e-08,
+      "loss": 0.0002,
+      "step": 18449
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.187530443253775e-08,
+      "loss": 0.0004,
+      "step": 18450
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.185094982951778e-08,
+      "loss": 0.0003,
+      "step": 18451
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.1826595226497806e-08,
+      "loss": 0.0003,
+      "step": 18452
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.180224062347784e-08,
+      "loss": 0.0002,
+      "step": 18453
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.177788602045786e-08,
+      "loss": 0.0001,
+      "step": 18454
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 5.1753531417437895e-08,
+      "loss": 0.0002,
+      "step": 18455
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.172917681441792e-08,
+      "loss": 0.0004,
+      "step": 18456
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.170482221139795e-08,
+      "loss": 0.0002,
+      "step": 18457
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.168046760837798e-08,
+      "loss": 0.0004,
+      "step": 18458
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.165611300535801e-08,
+      "loss": 0.0004,
+      "step": 18459
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.163175840233804e-08,
+      "loss": 0.0002,
+      "step": 18460
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.160740379931807e-08,
+      "loss": 0.0002,
+      "step": 18461
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.15830491962981e-08,
+      "loss": 0.0002,
+      "step": 18462
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.155869459327813e-08,
+      "loss": 0.0002,
+      "step": 18463
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.1534339990258154e-08,
+      "loss": 0.0002,
+      "step": 18464
+    },
+    {
+      "epoch": 17.94,
+      "learning_rate": 5.1509985387238186e-08,
+      "loss": 0.0002,
+      "step": 18465
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.148563078421821e-08,
+      "loss": 0.0002,
+      "step": 18466
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.146127618119824e-08,
+      "loss": 0.0002,
+      "step": 18467
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.143692157817827e-08,
+      "loss": 0.0004,
+      "step": 18468
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.1412566975158307e-08,
+      "loss": 0.0002,
+      "step": 18469
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.138821237213833e-08,
+      "loss": 0.0002,
+      "step": 18470
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.1363857769118363e-08,
+      "loss": 0.0003,
+      "step": 18471
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.133950316609839e-08,
+      "loss": 0.0002,
+      "step": 18472
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.131514856307842e-08,
+      "loss": 0.0003,
+      "step": 18473
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.1290793960058446e-08,
+      "loss": 0.0002,
+      "step": 18474
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 5.126643935703848e-08,
+      "loss": 0.0003,
+      "step": 18475
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.12420847540185e-08,
+      "loss": 0.0002,
+      "step": 18476
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.1217730150998534e-08,
+      "loss": 0.0004,
+      "step": 18477
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.119337554797856e-08,
+      "loss": 0.0001,
+      "step": 18478
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.11690209449586e-08,
+      "loss": 0.0003,
+      "step": 18479
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.114466634193862e-08,
+      "loss": 0.0003,
+      "step": 18480
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.1120311738918655e-08,
+      "loss": 0.0004,
+      "step": 18481
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.109595713589868e-08,
+      "loss": 0.0002,
+      "step": 18482
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.107160253287871e-08,
+      "loss": 0.0001,
+      "step": 18483
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.104724792985874e-08,
+      "loss": 0.0003,
+      "step": 18484
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 5.102289332683877e-08,
+      "loss": 0.0003,
+      "step": 18485
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.09985387238188e-08,
+      "loss": 0.0003,
+      "step": 18486
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.0974184120798826e-08,
+      "loss": 0.0001,
+      "step": 18487
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.0949829517778864e-08,
+      "loss": 0.0002,
+      "step": 18488
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.092547491475889e-08,
+      "loss": 0.0002,
+      "step": 18489
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.090112031173892e-08,
+      "loss": 0.0003,
+      "step": 18490
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.0876765708718946e-08,
+      "loss": 0.0002,
+      "step": 18491
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.085241110569898e-08,
+      "loss": 1.2203,
+      "step": 18492
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.0828056502679e-08,
+      "loss": 0.0002,
+      "step": 18493
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.0803701899659035e-08,
+      "loss": 0.0001,
+      "step": 18494
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.077934729663906e-08,
+      "loss": 0.0002,
+      "step": 18495
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 5.075499269361909e-08,
+      "loss": 0.0003,
+      "step": 18496
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.073063809059912e-08,
+      "loss": 0.0001,
+      "step": 18497
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.0706283487579155e-08,
+      "loss": 0.0002,
+      "step": 18498
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.068192888455918e-08,
+      "loss": 0.0004,
+      "step": 18499
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.065757428153921e-08,
+      "loss": 0.0002,
+      "step": 18500
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.063321967851924e-08,
+      "loss": 0.0004,
+      "step": 18501
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.060886507549927e-08,
+      "loss": 0.0002,
+      "step": 18502
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.0584510472479295e-08,
+      "loss": 0.0001,
+      "step": 18503
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.0560155869459326e-08,
+      "loss": 0.0001,
+      "step": 18504
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.053580126643935e-08,
+      "loss": 0.0002,
+      "step": 18505
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 5.051144666341938e-08,
+      "loss": 2.1535,
+      "step": 18506
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.0487092060399415e-08,
+      "loss": 0.0002,
+      "step": 18507
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.046273745737945e-08,
+      "loss": 0.0003,
+      "step": 18508
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.043838285435947e-08,
+      "loss": 0.0002,
+      "step": 18509
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.0414028251339504e-08,
+      "loss": 0.0003,
+      "step": 18510
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.038967364831953e-08,
+      "loss": 0.0002,
+      "step": 18511
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.036531904529956e-08,
+      "loss": 0.0003,
+      "step": 18512
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.0340964442279586e-08,
+      "loss": 0.0002,
+      "step": 18513
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.031660983925962e-08,
+      "loss": 0.0004,
+      "step": 18514
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.029225523623964e-08,
+      "loss": 0.0002,
+      "step": 18515
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 5.0267900633219675e-08,
+      "loss": 0.0002,
+      "step": 18516
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.0243546030199706e-08,
+      "loss": 0.0002,
+      "step": 18517
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.021919142717974e-08,
+      "loss": 0.0004,
+      "step": 18518
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.0194836824159763e-08,
+      "loss": 0.0002,
+      "step": 18519
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.0170482221139795e-08,
+      "loss": 0.0004,
+      "step": 18520
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.014612761811982e-08,
+      "loss": 0.0001,
+      "step": 18521
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.012177301509985e-08,
+      "loss": 0.0001,
+      "step": 18522
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.9990277102576568,
+      "eval_loss": 0.014007431454956532,
+      "eval_runtime": 2.7862,
+      "eval_samples_per_second": 738.289,
+      "eval_steps_per_second": 11.844,
+      "step": 18522
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.009741841207988e-08,
+      "loss": 0.0003,
+      "step": 18523
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.007306380905991e-08,
+      "loss": 0.0002,
+      "step": 18524
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.0048709206039934e-08,
+      "loss": 0.0002,
+      "step": 18525
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5.0024354603019966e-08,
+      "loss": 0.0005,
+      "step": 18526
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 5e-08,
+      "loss": 0.0002,
+      "step": 18527
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.997564539698003e-08,
+      "loss": 0.0002,
+      "step": 18528
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.9951290793960055e-08,
+      "loss": 0.0003,
+      "step": 18529
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.9926936190940086e-08,
+      "loss": 0.0002,
+      "step": 18530
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.990258158792011e-08,
+      "loss": 0.0003,
+      "step": 18531
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.9878226984900143e-08,
+      "loss": 0.0002,
+      "step": 18532
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.985387238188017e-08,
+      "loss": 0.0002,
+      "step": 18533
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.98295177788602e-08,
+      "loss": 0.0002,
+      "step": 18534
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.9805163175840226e-08,
+      "loss": 0.0002,
+      "step": 18535
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.9780808572820264e-08,
+      "loss": 0.0003,
+      "step": 18536
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 4.9756453969800296e-08,
+      "loss": 0.0003,
+      "step": 18537
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.973209936678032e-08,
+      "loss": 0.0002,
+      "step": 18538
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.970774476376035e-08,
+      "loss": 0.0002,
+      "step": 18539
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.968339016074038e-08,
+      "loss": 0.0002,
+      "step": 18540
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.965903555772041e-08,
+      "loss": 0.0003,
+      "step": 18541
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.9634680954700435e-08,
+      "loss": 0.0002,
+      "step": 18542
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.9610326351680467e-08,
+      "loss": 0.0002,
+      "step": 18543
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.958597174866049e-08,
+      "loss": 0.0001,
+      "step": 18544
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.9561617145640524e-08,
+      "loss": 0.0002,
+      "step": 18545
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.9537262542620555e-08,
+      "loss": 0.0003,
+      "step": 18546
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 4.951290793960059e-08,
+      "loss": 0.0003,
+      "step": 18547
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.948855333658061e-08,
+      "loss": 0.0003,
+      "step": 18548
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.9464198733560644e-08,
+      "loss": 0.0003,
+      "step": 18549
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.943984413054067e-08,
+      "loss": 0.0002,
+      "step": 18550
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.94154895275207e-08,
+      "loss": 0.0003,
+      "step": 18551
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.9391134924500726e-08,
+      "loss": 0.0002,
+      "step": 18552
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.936678032148076e-08,
+      "loss": 0.0001,
+      "step": 18553
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.934242571846078e-08,
+      "loss": 0.0007,
+      "step": 18554
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.931807111544082e-08,
+      "loss": 0.0001,
+      "step": 18555
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.9293716512420847e-08,
+      "loss": 0.0003,
+      "step": 18556
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.926936190940088e-08,
+      "loss": 0.0003,
+      "step": 18557
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 4.9245007306380904e-08,
+      "loss": 0.0001,
+      "step": 18558
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.9220652703360935e-08,
+      "loss": 0.0002,
+      "step": 18559
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.919629810034096e-08,
+      "loss": 0.0001,
+      "step": 18560
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.917194349732099e-08,
+      "loss": 0.0003,
+      "step": 18561
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.914758889430102e-08,
+      "loss": 0.0002,
+      "step": 18562
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.912323429128105e-08,
+      "loss": 0.0002,
+      "step": 18563
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.9098879688261074e-08,
+      "loss": 0.0004,
+      "step": 18564
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.907452508524111e-08,
+      "loss": 0.0002,
+      "step": 18565
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.905017048222114e-08,
+      "loss": 0.0004,
+      "step": 18566
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.902581587920117e-08,
+      "loss": 0.0003,
+      "step": 18567
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 4.9001461276181195e-08,
+      "loss": 0.0004,
+      "step": 18568
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.897710667316123e-08,
+      "loss": 0.0002,
+      "step": 18569
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.895275207014125e-08,
+      "loss": 0.0001,
+      "step": 18570
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.8928397467121284e-08,
+      "loss": 0.0011,
+      "step": 18571
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.890404286410131e-08,
+      "loss": 0.0003,
+      "step": 18572
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.887968826108134e-08,
+      "loss": 0.0004,
+      "step": 18573
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.885533365806137e-08,
+      "loss": 0.0001,
+      "step": 18574
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.8830979055041404e-08,
+      "loss": 0.0002,
+      "step": 18575
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.880662445202143e-08,
+      "loss": 0.0001,
+      "step": 18576
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.878226984900146e-08,
+      "loss": 0.0002,
+      "step": 18577
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 4.8757915245981486e-08,
+      "loss": 0.0002,
+      "step": 18578
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.873356064296152e-08,
+      "loss": 0.0003,
+      "step": 18579
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.870920603994154e-08,
+      "loss": 0.0012,
+      "step": 18580
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.8684851436921575e-08,
+      "loss": 0.0003,
+      "step": 18581
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.86604968339016e-08,
+      "loss": 0.0005,
+      "step": 18582
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.863614223088163e-08,
+      "loss": 0.0001,
+      "step": 18583
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.8611787627861664e-08,
+      "loss": 1.6427,
+      "step": 18584
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.8587433024841696e-08,
+      "loss": 0.0003,
+      "step": 18585
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.856307842182172e-08,
+      "loss": 0.0002,
+      "step": 18586
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.853872381880175e-08,
+      "loss": 0.0002,
+      "step": 18587
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 4.851436921578178e-08,
+      "loss": 0.0002,
+      "step": 18588
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.849001461276181e-08,
+      "loss": 0.0003,
+      "step": 18589
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.8465660009741835e-08,
+      "loss": 0.0002,
+      "step": 18590
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.8441305406721866e-08,
+      "loss": 0.0002,
+      "step": 18591
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.841695080370189e-08,
+      "loss": 0.0002,
+      "step": 18592
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.8392596200681923e-08,
+      "loss": 0.0002,
+      "step": 18593
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.836824159766196e-08,
+      "loss": 0.0002,
+      "step": 18594
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.834388699464199e-08,
+      "loss": 0.0002,
+      "step": 18595
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.831953239162202e-08,
+      "loss": 0.0002,
+      "step": 18596
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.8295177788602044e-08,
+      "loss": 0.0004,
+      "step": 18597
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.8270823185582076e-08,
+      "loss": 0.0005,
+      "step": 18598
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 4.82464685825621e-08,
+      "loss": 0.0003,
+      "step": 18599
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.822211397954213e-08,
+      "loss": 0.0002,
+      "step": 18600
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.819775937652216e-08,
+      "loss": 0.0002,
+      "step": 18601
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.817340477350219e-08,
+      "loss": 0.0002,
+      "step": 18602
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.814905017048222e-08,
+      "loss": 0.0004,
+      "step": 18603
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.812469556746225e-08,
+      "loss": 0.0004,
+      "step": 18604
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.810034096444228e-08,
+      "loss": 0.0002,
+      "step": 18605
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.807598636142231e-08,
+      "loss": 0.0002,
+      "step": 18606
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.8051631758402335e-08,
+      "loss": 0.0002,
+      "step": 18607
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.802727715538237e-08,
+      "loss": 0.0002,
+      "step": 18608
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 4.800292255236239e-08,
+      "loss": 0.0005,
+      "step": 18609
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.7978567949342424e-08,
+      "loss": 0.0002,
+      "step": 18610
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.795421334632245e-08,
+      "loss": 0.0002,
+      "step": 18611
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.792985874330248e-08,
+      "loss": 0.0004,
+      "step": 18612
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.790550414028251e-08,
+      "loss": 0.0004,
+      "step": 18613
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.7881149537262544e-08,
+      "loss": 0.0001,
+      "step": 18614
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.785679493424257e-08,
+      "loss": 0.0002,
+      "step": 18615
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.78324403312226e-08,
+      "loss": 0.0003,
+      "step": 18616
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.7808085728202627e-08,
+      "loss": 0.0001,
+      "step": 18617
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.778373112518266e-08,
+      "loss": 0.0003,
+      "step": 18618
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 4.7759376522162684e-08,
+      "loss": 0.0004,
+      "step": 18619
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.7735021919142715e-08,
+      "loss": 0.0004,
+      "step": 18620
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.771066731612274e-08,
+      "loss": 0.0003,
+      "step": 18621
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.768631271310278e-08,
+      "loss": 0.0002,
+      "step": 18622
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.7661958110082804e-08,
+      "loss": 0.0002,
+      "step": 18623
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.7637603507062836e-08,
+      "loss": 0.0004,
+      "step": 18624
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.761324890404286e-08,
+      "loss": 0.0003,
+      "step": 18625
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.758889430102289e-08,
+      "loss": 0.0002,
+      "step": 18626
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.756453969800292e-08,
+      "loss": 0.0004,
+      "step": 18627
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.754018509498295e-08,
+      "loss": 0.0003,
+      "step": 18628
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.7515830491962975e-08,
+      "loss": 0.0003,
+      "step": 18629
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 4.7491475888943007e-08,
+      "loss": 0.0002,
+      "step": 18630
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.746712128592303e-08,
+      "loss": 0.0002,
+      "step": 18631
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.744276668290307e-08,
+      "loss": 0.0003,
+      "step": 18632
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.7418412079883095e-08,
+      "loss": 0.0002,
+      "step": 18633
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.739405747686313e-08,
+      "loss": 0.0003,
+      "step": 18634
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.736970287384315e-08,
+      "loss": 0.0002,
+      "step": 18635
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.7345348270823184e-08,
+      "loss": 0.0002,
+      "step": 18636
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.732099366780321e-08,
+      "loss": 0.0001,
+      "step": 18637
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.729663906478324e-08,
+      "loss": 0.0004,
+      "step": 18638
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.7272284461763266e-08,
+      "loss": 0.0004,
+      "step": 18639
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 4.72479298587433e-08,
+      "loss": 0.0003,
+      "step": 18640
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.722357525572333e-08,
+      "loss": 0.0002,
+      "step": 18641
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.719922065270336e-08,
+      "loss": 0.0002,
+      "step": 18642
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.717486604968339e-08,
+      "loss": 0.0001,
+      "step": 18643
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.715051144666342e-08,
+      "loss": 0.0002,
+      "step": 18644
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.712615684364345e-08,
+      "loss": 0.0003,
+      "step": 18645
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.7101802240623475e-08,
+      "loss": 0.0003,
+      "step": 18646
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.707744763760351e-08,
+      "loss": 0.0002,
+      "step": 18647
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.705309303458353e-08,
+      "loss": 0.0006,
+      "step": 18648
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.7028738431563564e-08,
+      "loss": 0.0003,
+      "step": 18649
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 4.700438382854359e-08,
+      "loss": 0.0003,
+      "step": 18650
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.698002922552363e-08,
+      "loss": 0.0003,
+      "step": 18651
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.695567462250365e-08,
+      "loss": 0.0003,
+      "step": 18652
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.6931320019483685e-08,
+      "loss": 0.0003,
+      "step": 18653
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.690696541646371e-08,
+      "loss": 0.0003,
+      "step": 18654
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.688261081344374e-08,
+      "loss": 0.0002,
+      "step": 18655
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.685825621042377e-08,
+      "loss": 0.0002,
+      "step": 18656
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.68339016074038e-08,
+      "loss": 0.3579,
+      "step": 18657
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.6809547004383824e-08,
+      "loss": 0.0002,
+      "step": 18658
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.6785192401363856e-08,
+      "loss": 0.0002,
+      "step": 18659
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 4.676083779834388e-08,
+      "loss": 0.0002,
+      "step": 18660
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.673648319532392e-08,
+      "loss": 0.0005,
+      "step": 18661
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.6712128592303944e-08,
+      "loss": 0.0002,
+      "step": 18662
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.6687773989283976e-08,
+      "loss": 0.0002,
+      "step": 18663
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.6663419386264e-08,
+      "loss": 0.0001,
+      "step": 18664
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.663906478324403e-08,
+      "loss": 0.0003,
+      "step": 18665
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.661471018022406e-08,
+      "loss": 0.0003,
+      "step": 18666
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.659035557720409e-08,
+      "loss": 0.0001,
+      "step": 18667
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.6566000974184115e-08,
+      "loss": 0.0003,
+      "step": 18668
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.654164637116415e-08,
+      "loss": 0.0002,
+      "step": 18669
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.651729176814418e-08,
+      "loss": 0.0002,
+      "step": 18670
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 4.649293716512421e-08,
+      "loss": 0.0002,
+      "step": 18671
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.6468582562104236e-08,
+      "loss": 0.0003,
+      "step": 18672
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.644422795908427e-08,
+      "loss": 0.0002,
+      "step": 18673
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.641987335606429e-08,
+      "loss": 0.0004,
+      "step": 18674
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.6395518753044324e-08,
+      "loss": 0.0007,
+      "step": 18675
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.637116415002435e-08,
+      "loss": 0.0001,
+      "step": 18676
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.634680954700438e-08,
+      "loss": 0.0001,
+      "step": 18677
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.6322454943984406e-08,
+      "loss": 0.0003,
+      "step": 18678
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.629810034096444e-08,
+      "loss": 0.0002,
+      "step": 18679
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.627374573794447e-08,
+      "loss": 0.0002,
+      "step": 18680
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 4.62493911349245e-08,
+      "loss": 0.0002,
+      "step": 18681
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.622503653190453e-08,
+      "loss": 0.0002,
+      "step": 18682
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.620068192888456e-08,
+      "loss": 0.0002,
+      "step": 18683
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.6176327325864584e-08,
+      "loss": 0.0002,
+      "step": 18684
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.6151972722844616e-08,
+      "loss": 0.0002,
+      "step": 18685
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.612761811982464e-08,
+      "loss": 0.0002,
+      "step": 18686
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.610326351680467e-08,
+      "loss": 0.0002,
+      "step": 18687
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.60789089137847e-08,
+      "loss": 0.0002,
+      "step": 18688
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.6054554310764736e-08,
+      "loss": 0.0004,
+      "step": 18689
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.603019970774476e-08,
+      "loss": 0.0002,
+      "step": 18690
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 4.600584510472479e-08,
+      "loss": 0.0002,
+      "step": 18691
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.598149050170482e-08,
+      "loss": 0.0002,
+      "step": 18692
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.595713589868485e-08,
+      "loss": 0.0001,
+      "step": 18693
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.5932781295664875e-08,
+      "loss": 0.0002,
+      "step": 18694
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.590842669264491e-08,
+      "loss": 0.0002,
+      "step": 18695
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.588407208962493e-08,
+      "loss": 0.0002,
+      "step": 18696
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.5859717486604964e-08,
+      "loss": 0.0001,
+      "step": 18697
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.583536288358499e-08,
+      "loss": 0.0003,
+      "step": 18698
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.581100828056503e-08,
+      "loss": 2.4129,
+      "step": 18699
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.578665367754506e-08,
+      "loss": 0.0002,
+      "step": 18700
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.5762299074525085e-08,
+      "loss": 0.0002,
+      "step": 18701
+    },
+    {
+      "epoch": 18.17,
+      "learning_rate": 4.5737944471505116e-08,
+      "loss": 0.0003,
+      "step": 18702
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.571358986848514e-08,
+      "loss": 0.0002,
+      "step": 18703
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.568923526546517e-08,
+      "loss": 0.0002,
+      "step": 18704
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.56648806624452e-08,
+      "loss": 0.0002,
+      "step": 18705
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.564052605942523e-08,
+      "loss": 0.0003,
+      "step": 18706
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.5616171456405255e-08,
+      "loss": 0.0002,
+      "step": 18707
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.5591816853385294e-08,
+      "loss": 0.0003,
+      "step": 18708
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.556746225036532e-08,
+      "loss": 0.0002,
+      "step": 18709
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.554310764734535e-08,
+      "loss": 0.0004,
+      "step": 18710
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.5518753044325376e-08,
+      "loss": 0.0002,
+      "step": 18711
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 4.549439844130541e-08,
+      "loss": 0.0002,
+      "step": 18712
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.547004383828543e-08,
+      "loss": 0.0003,
+      "step": 18713
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.5445689235265465e-08,
+      "loss": 0.0002,
+      "step": 18714
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.542133463224549e-08,
+      "loss": 0.0003,
+      "step": 18715
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.539698002922552e-08,
+      "loss": 0.0001,
+      "step": 18716
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.537262542620555e-08,
+      "loss": 0.0001,
+      "step": 18717
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.5348270823185585e-08,
+      "loss": 0.0002,
+      "step": 18718
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.532391622016561e-08,
+      "loss": 0.0003,
+      "step": 18719
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.529956161714564e-08,
+      "loss": 0.0002,
+      "step": 18720
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.527520701412567e-08,
+      "loss": 0.0003,
+      "step": 18721
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 4.52508524111057e-08,
+      "loss": 0.0003,
+      "step": 18722
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.5226497808085724e-08,
+      "loss": 0.0002,
+      "step": 18723
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.5202143205065756e-08,
+      "loss": 0.0003,
+      "step": 18724
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.517778860204578e-08,
+      "loss": 0.0003,
+      "step": 18725
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.515343399902581e-08,
+      "loss": 0.0002,
+      "step": 18726
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.512907939600584e-08,
+      "loss": 0.0001,
+      "step": 18727
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.5104724792985877e-08,
+      "loss": 0.0003,
+      "step": 18728
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.50803701899659e-08,
+      "loss": 0.0002,
+      "step": 18729
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.5056015586945933e-08,
+      "loss": 0.0002,
+      "step": 18730
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.503166098392596e-08,
+      "loss": 0.0002,
+      "step": 18731
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 4.500730638090599e-08,
+      "loss": 0.0002,
+      "step": 18732
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.4982951777886016e-08,
+      "loss": 0.0003,
+      "step": 18733
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.495859717486605e-08,
+      "loss": 0.0003,
+      "step": 18734
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.493424257184607e-08,
+      "loss": 0.0004,
+      "step": 18735
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.4909887968826104e-08,
+      "loss": 0.0002,
+      "step": 18736
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.4885533365806136e-08,
+      "loss": 0.0002,
+      "step": 18737
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.486117876278617e-08,
+      "loss": 0.0004,
+      "step": 18738
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.483682415976619e-08,
+      "loss": 0.0002,
+      "step": 18739
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.4812469556746225e-08,
+      "loss": 0.0003,
+      "step": 18740
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.478811495372625e-08,
+      "loss": 0.0002,
+      "step": 18741
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.476376035070628e-08,
+      "loss": 0.0002,
+      "step": 18742
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 4.473940574768631e-08,
+      "loss": 0.0003,
+      "step": 18743
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.471505114466634e-08,
+      "loss": 0.0002,
+      "step": 18744
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.4690696541646364e-08,
+      "loss": 0.0002,
+      "step": 18745
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.4666341938626396e-08,
+      "loss": 0.0003,
+      "step": 18746
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.464198733560643e-08,
+      "loss": 0.0002,
+      "step": 18747
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.461763273258646e-08,
+      "loss": 0.0008,
+      "step": 18748
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.4593278129566484e-08,
+      "loss": 0.0002,
+      "step": 18749
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.4568923526546516e-08,
+      "loss": 0.0002,
+      "step": 18750
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.454456892352654e-08,
+      "loss": 0.0004,
+      "step": 18751
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.452021432050657e-08,
+      "loss": 0.0002,
+      "step": 18752
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 4.44958597174866e-08,
+      "loss": 0.0002,
+      "step": 18753
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.447150511446663e-08,
+      "loss": 0.0002,
+      "step": 18754
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.444715051144666e-08,
+      "loss": 0.0002,
+      "step": 18755
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.4422795908426694e-08,
+      "loss": 0.0004,
+      "step": 18756
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.4398441305406725e-08,
+      "loss": 0.0002,
+      "step": 18757
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.437408670238675e-08,
+      "loss": 0.0004,
+      "step": 18758
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.434973209936678e-08,
+      "loss": 0.0002,
+      "step": 18759
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.432537749634681e-08,
+      "loss": 0.0003,
+      "step": 18760
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.430102289332684e-08,
+      "loss": 0.0003,
+      "step": 18761
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.4276668290306864e-08,
+      "loss": 0.0002,
+      "step": 18762
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 4.4252313687286896e-08,
+      "loss": 0.0003,
+      "step": 18763
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.422795908426692e-08,
+      "loss": 0.0002,
+      "step": 18764
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.420360448124695e-08,
+      "loss": 0.0003,
+      "step": 18765
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.4179249878226985e-08,
+      "loss": 0.0002,
+      "step": 18766
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.415489527520702e-08,
+      "loss": 0.0004,
+      "step": 18767
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.413054067218704e-08,
+      "loss": 0.0004,
+      "step": 18768
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.4106186069167074e-08,
+      "loss": 0.0002,
+      "step": 18769
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.40818314661471e-08,
+      "loss": 0.0034,
+      "step": 18770
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.405747686312713e-08,
+      "loss": 0.0003,
+      "step": 18771
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.4033122260107156e-08,
+      "loss": 0.0002,
+      "step": 18772
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.400876765708719e-08,
+      "loss": 0.0002,
+      "step": 18773
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 4.398441305406721e-08,
+      "loss": 0.0001,
+      "step": 18774
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.3960058451047245e-08,
+      "loss": 0.0001,
+      "step": 18775
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.3935703848027276e-08,
+      "loss": 0.0003,
+      "step": 18776
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.391134924500731e-08,
+      "loss": 0.0002,
+      "step": 18777
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.388699464198733e-08,
+      "loss": 0.0002,
+      "step": 18778
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.3862640038967365e-08,
+      "loss": 0.0004,
+      "step": 18779
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.383828543594739e-08,
+      "loss": 0.0002,
+      "step": 18780
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.381393083292742e-08,
+      "loss": 0.0002,
+      "step": 18781
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.378957622990745e-08,
+      "loss": 0.0002,
+      "step": 18782
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.376522162688748e-08,
+      "loss": 0.0002,
+      "step": 18783
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 4.3740867023867504e-08,
+      "loss": 0.0002,
+      "step": 18784
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.371651242084754e-08,
+      "loss": 0.0002,
+      "step": 18785
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.369215781782757e-08,
+      "loss": 0.1986,
+      "step": 18786
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.36678032148076e-08,
+      "loss": 0.0001,
+      "step": 18787
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.3643448611787625e-08,
+      "loss": 0.0003,
+      "step": 18788
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.3619094008767656e-08,
+      "loss": 0.0002,
+      "step": 18789
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.359473940574768e-08,
+      "loss": 0.0003,
+      "step": 18790
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.3570384802727713e-08,
+      "loss": 0.0003,
+      "step": 18791
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.354603019970774e-08,
+      "loss": 0.0002,
+      "step": 18792
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.352167559668777e-08,
+      "loss": 0.0002,
+      "step": 18793
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 4.3497320993667796e-08,
+      "loss": 0.0002,
+      "step": 18794
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.3472966390647834e-08,
+      "loss": 0.0002,
+      "step": 18795
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.344861178762786e-08,
+      "loss": 0.0002,
+      "step": 18796
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.342425718460789e-08,
+      "loss": 0.0002,
+      "step": 18797
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.3399902581587916e-08,
+      "loss": 0.0003,
+      "step": 18798
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.337554797856795e-08,
+      "loss": 0.0003,
+      "step": 18799
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.335119337554797e-08,
+      "loss": 0.0001,
+      "step": 18800
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.3326838772528005e-08,
+      "loss": 0.0003,
+      "step": 18801
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.330248416950803e-08,
+      "loss": 0.0004,
+      "step": 18802
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.327812956648806e-08,
+      "loss": 0.0002,
+      "step": 18803
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 4.3253774963468093e-08,
+      "loss": 0.0002,
+      "step": 18804
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.3229420360448125e-08,
+      "loss": 0.0003,
+      "step": 18805
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.320506575742816e-08,
+      "loss": 0.0001,
+      "step": 18806
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.318071115440818e-08,
+      "loss": 0.0003,
+      "step": 18807
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.3156356551388214e-08,
+      "loss": 0.0003,
+      "step": 18808
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.313200194836824e-08,
+      "loss": 0.0002,
+      "step": 18809
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.310764734534827e-08,
+      "loss": 0.0003,
+      "step": 18810
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.3083292742328296e-08,
+      "loss": 0.0003,
+      "step": 18811
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.305893813930833e-08,
+      "loss": 0.0003,
+      "step": 18812
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.303458353628835e-08,
+      "loss": 0.0004,
+      "step": 18813
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.301022893326839e-08,
+      "loss": 0.0002,
+      "step": 18814
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 4.2985874330248417e-08,
+      "loss": 0.0001,
+      "step": 18815
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.296151972722845e-08,
+      "loss": 0.0003,
+      "step": 18816
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.2937165124208474e-08,
+      "loss": 0.0002,
+      "step": 18817
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.2912810521188505e-08,
+      "loss": 0.0002,
+      "step": 18818
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.288845591816853e-08,
+      "loss": 0.0002,
+      "step": 18819
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.286410131514856e-08,
+      "loss": 0.0003,
+      "step": 18820
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.283974671212859e-08,
+      "loss": 0.0004,
+      "step": 18821
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.281539210910862e-08,
+      "loss": 0.0002,
+      "step": 18822
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.279103750608865e-08,
+      "loss": 0.0002,
+      "step": 18823
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.276668290306868e-08,
+      "loss": 0.0003,
+      "step": 18824
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 4.274232830004871e-08,
+      "loss": 0.0002,
+      "step": 18825
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.271797369702874e-08,
+      "loss": 0.0001,
+      "step": 18826
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.2693619094008765e-08,
+      "loss": 0.0002,
+      "step": 18827
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.26692644909888e-08,
+      "loss": 0.0002,
+      "step": 18828
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.264490988796882e-08,
+      "loss": 0.0002,
+      "step": 18829
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.2620555284948854e-08,
+      "loss": 0.0002,
+      "step": 18830
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.259620068192888e-08,
+      "loss": 0.0003,
+      "step": 18831
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.257184607890891e-08,
+      "loss": 0.0003,
+      "step": 18832
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.254749147588894e-08,
+      "loss": 0.0002,
+      "step": 18833
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.2523136872868974e-08,
+      "loss": 0.0002,
+      "step": 18834
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 4.2498782269849e-08,
+      "loss": 0.0001,
+      "step": 18835
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.247442766682903e-08,
+      "loss": 0.0005,
+      "step": 18836
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.2450073063809056e-08,
+      "loss": 0.0002,
+      "step": 18837
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.242571846078909e-08,
+      "loss": 0.0002,
+      "step": 18838
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.240136385776911e-08,
+      "loss": 0.0002,
+      "step": 18839
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.2377009254749145e-08,
+      "loss": 0.0003,
+      "step": 18840
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.235265465172917e-08,
+      "loss": 0.0003,
+      "step": 18841
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.23283000487092e-08,
+      "loss": 0.0014,
+      "step": 18842
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.2303945445689234e-08,
+      "loss": 0.0002,
+      "step": 18843
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.2279590842669266e-08,
+      "loss": 0.0004,
+      "step": 18844
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.225523623964929e-08,
+      "loss": 0.0003,
+      "step": 18845
+    },
+    {
+      "epoch": 18.31,
+      "learning_rate": 4.223088163662932e-08,
+      "loss": 0.0005,
+      "step": 18846
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.220652703360935e-08,
+      "loss": 0.0003,
+      "step": 18847
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.218217243058938e-08,
+      "loss": 0.0002,
+      "step": 18848
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.2157817827569405e-08,
+      "loss": 0.0002,
+      "step": 18849
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.2133463224549436e-08,
+      "loss": 0.0002,
+      "step": 18850
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.210910862152946e-08,
+      "loss": 0.0003,
+      "step": 18851
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.20847540185095e-08,
+      "loss": 0.0002,
+      "step": 18852
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.2060399415489525e-08,
+      "loss": 0.0002,
+      "step": 18853
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.203604481246956e-08,
+      "loss": 0.0002,
+      "step": 18854
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.201169020944958e-08,
+      "loss": 0.0002,
+      "step": 18855
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 4.1987335606429614e-08,
+      "loss": 0.0003,
+      "step": 18856
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.196298100340964e-08,
+      "loss": 0.0001,
+      "step": 18857
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.193862640038967e-08,
+      "loss": 0.0004,
+      "step": 18858
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.1914271797369696e-08,
+      "loss": 0.0001,
+      "step": 18859
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.188991719434973e-08,
+      "loss": 0.0002,
+      "step": 18860
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.186556259132975e-08,
+      "loss": 0.0004,
+      "step": 18861
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.184120798830979e-08,
+      "loss": 0.0002,
+      "step": 18862
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.181685338528982e-08,
+      "loss": 0.0003,
+      "step": 18863
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.179249878226985e-08,
+      "loss": 0.0002,
+      "step": 18864
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.176814417924988e-08,
+      "loss": 0.0001,
+      "step": 18865
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 4.1743789576229905e-08,
+      "loss": 0.0002,
+      "step": 18866
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.171943497320994e-08,
+      "loss": 0.0004,
+      "step": 18867
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.169508037018996e-08,
+      "loss": 0.0002,
+      "step": 18868
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.1670725767169994e-08,
+      "loss": 0.0002,
+      "step": 18869
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.164637116415002e-08,
+      "loss": 0.0002,
+      "step": 18870
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.162201656113006e-08,
+      "loss": 0.0002,
+      "step": 18871
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.159766195811008e-08,
+      "loss": 0.0003,
+      "step": 18872
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.1573307355090114e-08,
+      "loss": 0.0002,
+      "step": 18873
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.154895275207014e-08,
+      "loss": 0.0003,
+      "step": 18874
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.152459814905017e-08,
+      "loss": 0.0001,
+      "step": 18875
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.1500243546030197e-08,
+      "loss": 0.0001,
+      "step": 18876
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 4.147588894301023e-08,
+      "loss": 0.0002,
+      "step": 18877
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.1451534339990253e-08,
+      "loss": 0.0002,
+      "step": 18878
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.1427179736970285e-08,
+      "loss": 0.0002,
+      "step": 18879
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.140282513395031e-08,
+      "loss": 0.0003,
+      "step": 18880
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.137847053093035e-08,
+      "loss": 0.0003,
+      "step": 18881
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.1354115927910374e-08,
+      "loss": 0.0003,
+      "step": 18882
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.1329761324890406e-08,
+      "loss": 0.0002,
+      "step": 18883
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.130540672187043e-08,
+      "loss": 0.0003,
+      "step": 18884
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.128105211885046e-08,
+      "loss": 0.0004,
+      "step": 18885
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.125669751583049e-08,
+      "loss": 0.0002,
+      "step": 18886
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 4.123234291281052e-08,
+      "loss": 0.0003,
+      "step": 18887
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.1207988309790545e-08,
+      "loss": 0.0001,
+      "step": 18888
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.1183633706770577e-08,
+      "loss": 0.0004,
+      "step": 18889
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.115927910375061e-08,
+      "loss": 0.0004,
+      "step": 18890
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.113492450073064e-08,
+      "loss": 0.0003,
+      "step": 18891
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.1110569897710665e-08,
+      "loss": 0.0002,
+      "step": 18892
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.10862152946907e-08,
+      "loss": 0.0002,
+      "step": 18893
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.106186069167072e-08,
+      "loss": 0.0001,
+      "step": 18894
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.1037506088650754e-08,
+      "loss": 0.0003,
+      "step": 18895
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.101315148563078e-08,
+      "loss": 0.0004,
+      "step": 18896
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 4.098879688261081e-08,
+      "loss": 0.0003,
+      "step": 18897
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.0964442279590836e-08,
+      "loss": 0.0003,
+      "step": 18898
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.094008767657087e-08,
+      "loss": 0.0002,
+      "step": 18899
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.09157330735509e-08,
+      "loss": 0.0002,
+      "step": 18900
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.089137847053093e-08,
+      "loss": 0.0002,
+      "step": 18901
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.086702386751096e-08,
+      "loss": 0.0004,
+      "step": 18902
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.084266926449099e-08,
+      "loss": 0.0004,
+      "step": 18903
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.0818314661471014e-08,
+      "loss": 0.0003,
+      "step": 18904
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.0793960058451045e-08,
+      "loss": 0.0003,
+      "step": 18905
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.076960545543107e-08,
+      "loss": 0.0002,
+      "step": 18906
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 4.07452508524111e-08,
+      "loss": 0.0002,
+      "step": 18907
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.072089624939113e-08,
+      "loss": 0.0003,
+      "step": 18908
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.069654164637116e-08,
+      "loss": 0.0003,
+      "step": 18909
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.067218704335119e-08,
+      "loss": 0.0005,
+      "step": 18910
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.064783244033122e-08,
+      "loss": 0.0002,
+      "step": 18911
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.062347783731125e-08,
+      "loss": 0.0003,
+      "step": 18912
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.059912323429128e-08,
+      "loss": 0.0004,
+      "step": 18913
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.0574768631271305e-08,
+      "loss": 0.0002,
+      "step": 18914
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.055041402825134e-08,
+      "loss": 0.0002,
+      "step": 18915
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.052605942523137e-08,
+      "loss": 0.0002,
+      "step": 18916
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.0501704822211394e-08,
+      "loss": 0.0002,
+      "step": 18917
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 4.0477350219191426e-08,
+      "loss": 0.0002,
+      "step": 18918
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.045299561617146e-08,
+      "loss": 0.0002,
+      "step": 18919
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.042864101315149e-08,
+      "loss": 0.0003,
+      "step": 18920
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.0404286410131514e-08,
+      "loss": 0.0003,
+      "step": 18921
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.0379931807111546e-08,
+      "loss": 0.0004,
+      "step": 18922
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.035557720409157e-08,
+      "loss": 0.0002,
+      "step": 18923
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.03312226010716e-08,
+      "loss": 0.0003,
+      "step": 18924
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.030686799805163e-08,
+      "loss": 0.0001,
+      "step": 18925
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.028251339503166e-08,
+      "loss": 0.0001,
+      "step": 18926
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.0258158792011685e-08,
+      "loss": 0.0002,
+      "step": 18927
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 4.023380418899172e-08,
+      "loss": 0.0002,
+      "step": 18928
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.020944958597175e-08,
+      "loss": 0.0002,
+      "step": 18929
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.018509498295178e-08,
+      "loss": 0.0004,
+      "step": 18930
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.0160740379931806e-08,
+      "loss": 0.0002,
+      "step": 18931
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.013638577691184e-08,
+      "loss": 0.0001,
+      "step": 18932
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.011203117389186e-08,
+      "loss": 0.0001,
+      "step": 18933
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.0087676570871894e-08,
+      "loss": 0.74,
+      "step": 18934
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.006332196785192e-08,
+      "loss": 0.0002,
+      "step": 18935
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.003896736483195e-08,
+      "loss": 0.0002,
+      "step": 18936
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 4.0014612761811976e-08,
+      "loss": 0.0003,
+      "step": 18937
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 3.9990258158792015e-08,
+      "loss": 0.0001,
+      "step": 18938
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.996590355577204e-08,
+      "loss": 0.0003,
+      "step": 18939
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.994154895275207e-08,
+      "loss": 0.0002,
+      "step": 18940
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.99171943497321e-08,
+      "loss": 0.0002,
+      "step": 18941
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.989283974671213e-08,
+      "loss": 0.0002,
+      "step": 18942
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.9868485143692154e-08,
+      "loss": 0.0002,
+      "step": 18943
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.9844130540672186e-08,
+      "loss": 0.0002,
+      "step": 18944
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.981977593765221e-08,
+      "loss": 0.0002,
+      "step": 18945
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.979542133463224e-08,
+      "loss": 0.0003,
+      "step": 18946
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.977106673161227e-08,
+      "loss": 0.0002,
+      "step": 18947
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.9746712128592306e-08,
+      "loss": 0.0003,
+      "step": 18948
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 3.972235752557233e-08,
+      "loss": 0.0002,
+      "step": 18949
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.969800292255236e-08,
+      "loss": 0.0003,
+      "step": 18950
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.967364831953239e-08,
+      "loss": 0.0002,
+      "step": 18951
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.964929371651242e-08,
+      "loss": 0.0003,
+      "step": 18952
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.9624939113492445e-08,
+      "loss": 0.0002,
+      "step": 18953
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.960058451047248e-08,
+      "loss": 0.0001,
+      "step": 18954
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.95762299074525e-08,
+      "loss": 0.0002,
+      "step": 18955
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.9551875304432534e-08,
+      "loss": 0.0002,
+      "step": 18956
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.9527520701412566e-08,
+      "loss": 0.0002,
+      "step": 18957
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.95031660983926e-08,
+      "loss": 0.0002,
+      "step": 18958
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 3.947881149537262e-08,
+      "loss": 0.0001,
+      "step": 18959
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.9454456892352655e-08,
+      "loss": 0.0001,
+      "step": 18960
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.943010228933268e-08,
+      "loss": 0.0003,
+      "step": 18961
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.940574768631271e-08,
+      "loss": 0.0002,
+      "step": 18962
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.9381393083292737e-08,
+      "loss": 0.0003,
+      "step": 18963
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.935703848027277e-08,
+      "loss": 0.0002,
+      "step": 18964
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.9332683877252794e-08,
+      "loss": 0.0003,
+      "step": 18965
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.9308329274232825e-08,
+      "loss": 0.0003,
+      "step": 18966
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.9283974671212864e-08,
+      "loss": 0.0001,
+      "step": 18967
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.925962006819289e-08,
+      "loss": 0.0003,
+      "step": 18968
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 3.923526546517292e-08,
+      "loss": 0.0002,
+      "step": 18969
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.9210910862152946e-08,
+      "loss": 0.0003,
+      "step": 18970
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.918655625913298e-08,
+      "loss": 0.0003,
+      "step": 18971
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.9162201656113e-08,
+      "loss": 0.0002,
+      "step": 18972
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.9137847053093035e-08,
+      "loss": 0.0001,
+      "step": 18973
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.911349245007306e-08,
+      "loss": 0.0003,
+      "step": 18974
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.908913784705309e-08,
+      "loss": 0.0001,
+      "step": 18975
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.906478324403312e-08,
+      "loss": 0.0005,
+      "step": 18976
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.9040428641013155e-08,
+      "loss": 0.0002,
+      "step": 18977
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.901607403799318e-08,
+      "loss": 0.0164,
+      "step": 18978
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 3.899171943497321e-08,
+      "loss": 0.0002,
+      "step": 18979
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.896736483195324e-08,
+      "loss": 0.0002,
+      "step": 18980
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.894301022893327e-08,
+      "loss": 0.0002,
+      "step": 18981
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.8918655625913294e-08,
+      "loss": 2.0236,
+      "step": 18982
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.8894301022893326e-08,
+      "loss": 0.0002,
+      "step": 18983
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.886994641987335e-08,
+      "loss": 0.0002,
+      "step": 18984
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.884559181685338e-08,
+      "loss": 1.0072,
+      "step": 18985
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.8821237213833415e-08,
+      "loss": 0.0001,
+      "step": 18986
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.8796882610813446e-08,
+      "loss": 0.0001,
+      "step": 18987
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.877252800779347e-08,
+      "loss": 0.0051,
+      "step": 18988
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.8748173404773503e-08,
+      "loss": 0.0002,
+      "step": 18989
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 3.872381880175353e-08,
+      "loss": 0.0002,
+      "step": 18990
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.869946419873356e-08,
+      "loss": 0.0002,
+      "step": 18991
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.8675109595713586e-08,
+      "loss": 0.0003,
+      "step": 18992
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.865075499269362e-08,
+      "loss": 0.0003,
+      "step": 18993
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.862640038967364e-08,
+      "loss": 0.0003,
+      "step": 18994
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.8602045786653674e-08,
+      "loss": 0.0001,
+      "step": 18995
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.8577691183633706e-08,
+      "loss": 0.0002,
+      "step": 18996
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.855333658061374e-08,
+      "loss": 0.0002,
+      "step": 18997
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.852898197759376e-08,
+      "loss": 0.0004,
+      "step": 18998
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.8504627374573795e-08,
+      "loss": 0.0001,
+      "step": 18999
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 3.848027277155382e-08,
+      "loss": 0.0002,
+      "step": 19000
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.845591816853385e-08,
+      "loss": 0.0002,
+      "step": 19001
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.843156356551388e-08,
+      "loss": 0.0003,
+      "step": 19002
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.840720896249391e-08,
+      "loss": 0.0003,
+      "step": 19003
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.8382854359473934e-08,
+      "loss": 0.0002,
+      "step": 19004
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.835849975645397e-08,
+      "loss": 0.0003,
+      "step": 19005
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.8334145153434e-08,
+      "loss": 0.0002,
+      "step": 19006
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.830979055041403e-08,
+      "loss": 0.0002,
+      "step": 19007
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.8285435947394054e-08,
+      "loss": 0.0002,
+      "step": 19008
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.8261081344374086e-08,
+      "loss": 0.0003,
+      "step": 19009
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 3.823672674135411e-08,
+      "loss": 0.0003,
+      "step": 19010
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.821237213833414e-08,
+      "loss": 0.0001,
+      "step": 19011
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.818801753531417e-08,
+      "loss": 0.0002,
+      "step": 19012
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.81636629322942e-08,
+      "loss": 0.0002,
+      "step": 19013
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.8139308329274225e-08,
+      "loss": 0.0002,
+      "step": 19014
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.8114953726254264e-08,
+      "loss": 0.0002,
+      "step": 19015
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.809059912323429e-08,
+      "loss": 0.0003,
+      "step": 19016
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.806624452021432e-08,
+      "loss": 0.0002,
+      "step": 19017
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.8041889917194346e-08,
+      "loss": 0.0001,
+      "step": 19018
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.801753531417438e-08,
+      "loss": 0.0004,
+      "step": 19019
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.79931807111544e-08,
+      "loss": 0.0002,
+      "step": 19020
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 3.7968826108134434e-08,
+      "loss": 0.0001,
+      "step": 19021
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.794447150511446e-08,
+      "loss": 0.0003,
+      "step": 19022
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.792011690209449e-08,
+      "loss": 0.0003,
+      "step": 19023
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.789576229907453e-08,
+      "loss": 0.0002,
+      "step": 19024
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.7871407696054555e-08,
+      "loss": 0.0002,
+      "step": 19025
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.784705309303459e-08,
+      "loss": 0.0002,
+      "step": 19026
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.782269849001461e-08,
+      "loss": 0.0001,
+      "step": 19027
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.7798343886994644e-08,
+      "loss": 0.0003,
+      "step": 19028
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.777398928397467e-08,
+      "loss": 0.0002,
+      "step": 19029
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.77496346809547e-08,
+      "loss": 0.0002,
+      "step": 19030
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 3.7725280077934726e-08,
+      "loss": 0.0002,
+      "step": 19031
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.770092547491476e-08,
+      "loss": 0.0001,
+      "step": 19032
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.767657087189478e-08,
+      "loss": 0.0001,
+      "step": 19033
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.765221626887482e-08,
+      "loss": 0.0002,
+      "step": 19034
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.7627861665854846e-08,
+      "loss": 0.0001,
+      "step": 19035
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.760350706283488e-08,
+      "loss": 0.0002,
+      "step": 19036
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.75791524598149e-08,
+      "loss": 0.0002,
+      "step": 19037
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.7554797856794935e-08,
+      "loss": 0.0003,
+      "step": 19038
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.753044325377496e-08,
+      "loss": 0.0002,
+      "step": 19039
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.750608865075499e-08,
+      "loss": 0.0004,
+      "step": 19040
+    },
+    {
+      "epoch": 18.5,
+      "learning_rate": 3.748173404773502e-08,
+      "loss": 0.0002,
+      "step": 19041
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.745737944471505e-08,
+      "loss": 0.0002,
+      "step": 19042
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.7433024841695074e-08,
+      "loss": 0.0003,
+      "step": 19043
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.740867023867511e-08,
+      "loss": 0.0005,
+      "step": 19044
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.738431563565514e-08,
+      "loss": 0.0002,
+      "step": 19045
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.735996103263517e-08,
+      "loss": 0.0003,
+      "step": 19046
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.7335606429615195e-08,
+      "loss": 0.0002,
+      "step": 19047
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.7311251826595226e-08,
+      "loss": 0.0002,
+      "step": 19048
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.728689722357525e-08,
+      "loss": 0.0004,
+      "step": 19049
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.7262542620555283e-08,
+      "loss": 0.0002,
+      "step": 19050
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 3.723818801753531e-08,
+      "loss": 0.0003,
+      "step": 19051
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.721383341451534e-08,
+      "loss": 0.0001,
+      "step": 19052
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.718947881149537e-08,
+      "loss": 0.0005,
+      "step": 19053
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.7165124208475404e-08,
+      "loss": 0.0002,
+      "step": 19054
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.714076960545543e-08,
+      "loss": 0.0002,
+      "step": 19055
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.711641500243546e-08,
+      "loss": 0.0002,
+      "step": 19056
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.7092060399415486e-08,
+      "loss": 0.0001,
+      "step": 19057
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.706770579639552e-08,
+      "loss": 0.0002,
+      "step": 19058
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.704335119337554e-08,
+      "loss": 0.0002,
+      "step": 19059
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.7018996590355575e-08,
+      "loss": 0.0002,
+      "step": 19060
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.69946419873356e-08,
+      "loss": 0.0005,
+      "step": 19061
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 3.697028738431563e-08,
+      "loss": 0.0003,
+      "step": 19062
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.6945932781295663e-08,
+      "loss": 0.0003,
+      "step": 19063
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.6921578178275695e-08,
+      "loss": 0.0003,
+      "step": 19064
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.689722357525572e-08,
+      "loss": 0.0003,
+      "step": 19065
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.687286897223575e-08,
+      "loss": 0.0002,
+      "step": 19066
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.684851436921578e-08,
+      "loss": 0.0003,
+      "step": 19067
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.682415976619581e-08,
+      "loss": 0.0002,
+      "step": 19068
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.6799805163175834e-08,
+      "loss": 0.0002,
+      "step": 19069
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.6775450560155866e-08,
+      "loss": 0.0003,
+      "step": 19070
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.675109595713589e-08,
+      "loss": 0.0002,
+      "step": 19071
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 3.672674135411593e-08,
+      "loss": 0.0001,
+      "step": 19072
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.6702386751095955e-08,
+      "loss": 0.0002,
+      "step": 19073
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.6678032148075987e-08,
+      "loss": 0.0002,
+      "step": 19074
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.665367754505602e-08,
+      "loss": 0.0002,
+      "step": 19075
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.6629322942036044e-08,
+      "loss": 0.0002,
+      "step": 19076
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.6604968339016075e-08,
+      "loss": 0.0002,
+      "step": 19077
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.65806137359961e-08,
+      "loss": 0.0004,
+      "step": 19078
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.655625913297613e-08,
+      "loss": 0.0004,
+      "step": 19079
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.653190452995616e-08,
+      "loss": 0.0002,
+      "step": 19080
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.650754992693619e-08,
+      "loss": 0.0004,
+      "step": 19081
+    },
+    {
+      "epoch": 18.54,
+      "learning_rate": 3.648319532391622e-08,
+      "loss": 0.0002,
+      "step": 19082
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.645884072089625e-08,
+      "loss": 0.0002,
+      "step": 19083
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.643448611787628e-08,
+      "loss": 0.0002,
+      "step": 19084
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.641013151485631e-08,
+      "loss": 0.0003,
+      "step": 19085
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.6385776911836335e-08,
+      "loss": 0.0003,
+      "step": 19086
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.6361422308816367e-08,
+      "loss": 0.0001,
+      "step": 19087
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.633706770579639e-08,
+      "loss": 0.0002,
+      "step": 19088
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.6312713102776424e-08,
+      "loss": 0.0001,
+      "step": 19089
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.628835849975645e-08,
+      "loss": 0.0002,
+      "step": 19090
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.626400389673648e-08,
+      "loss": 0.0001,
+      "step": 19091
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.623964929371651e-08,
+      "loss": 0.0003,
+      "step": 19092
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 3.6215294690696544e-08,
+      "loss": 0.0002,
+      "step": 19093
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.619094008767657e-08,
+      "loss": 0.0003,
+      "step": 19094
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.61665854846566e-08,
+      "loss": 0.0002,
+      "step": 19095
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.6142230881636626e-08,
+      "loss": 0.0004,
+      "step": 19096
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.611787627861666e-08,
+      "loss": 0.0046,
+      "step": 19097
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.609352167559668e-08,
+      "loss": 0.0002,
+      "step": 19098
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.6069167072576715e-08,
+      "loss": 0.0001,
+      "step": 19099
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.604481246955674e-08,
+      "loss": 0.0002,
+      "step": 19100
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.602045786653678e-08,
+      "loss": 0.0002,
+      "step": 19101
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.5996103263516804e-08,
+      "loss": 0.0003,
+      "step": 19102
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 3.5971748660496836e-08,
+      "loss": 0.0002,
+      "step": 19103
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.594739405747686e-08,
+      "loss": 0.0002,
+      "step": 19104
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.592303945445689e-08,
+      "loss": 0.0004,
+      "step": 19105
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.589868485143692e-08,
+      "loss": 0.0003,
+      "step": 19106
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.587433024841695e-08,
+      "loss": 0.0002,
+      "step": 19107
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.5849975645396975e-08,
+      "loss": 0.0003,
+      "step": 19108
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.5825621042377006e-08,
+      "loss": 0.0004,
+      "step": 19109
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.580126643935703e-08,
+      "loss": 0.0002,
+      "step": 19110
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.577691183633707e-08,
+      "loss": 0.0003,
+      "step": 19111
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.5752557233317095e-08,
+      "loss": 0.0002,
+      "step": 19112
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 3.572820263029713e-08,
+      "loss": 0.0004,
+      "step": 19113
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.570384802727715e-08,
+      "loss": 0.0004,
+      "step": 19114
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.5679493424257184e-08,
+      "loss": 0.0002,
+      "step": 19115
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.565513882123721e-08,
+      "loss": 0.0001,
+      "step": 19116
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.563078421821724e-08,
+      "loss": 0.0001,
+      "step": 19117
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.5606429615197266e-08,
+      "loss": 0.0003,
+      "step": 19118
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.55820750121773e-08,
+      "loss": 0.0002,
+      "step": 19119
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.555772040915733e-08,
+      "loss": 0.0003,
+      "step": 19120
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.553336580613736e-08,
+      "loss": 0.0002,
+      "step": 19121
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.5509011203117386e-08,
+      "loss": 0.0001,
+      "step": 19122
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 3.548465660009742e-08,
+      "loss": 0.0003,
+      "step": 19123
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.5460301997077443e-08,
+      "loss": 0.0004,
+      "step": 19124
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.5435947394057475e-08,
+      "loss": 0.0003,
+      "step": 19125
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.54115927910375e-08,
+      "loss": 0.0002,
+      "step": 19126
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.538723818801753e-08,
+      "loss": 0.0003,
+      "step": 19127
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.536288358499756e-08,
+      "loss": 0.0002,
+      "step": 19128
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.533852898197759e-08,
+      "loss": 0.0003,
+      "step": 19129
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.531417437895763e-08,
+      "loss": 0.0002,
+      "step": 19130
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.528981977593765e-08,
+      "loss": 0.0001,
+      "step": 19131
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.5265465172917684e-08,
+      "loss": 0.0003,
+      "step": 19132
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.524111056989771e-08,
+      "loss": 0.0002,
+      "step": 19133
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 3.521675596687774e-08,
+      "loss": 0.0005,
+      "step": 19134
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.5192401363857767e-08,
+      "loss": 0.0002,
+      "step": 19135
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.51680467608378e-08,
+      "loss": 0.0001,
+      "step": 19136
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.5143692157817823e-08,
+      "loss": 0.0003,
+      "step": 19137
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.5119337554797855e-08,
+      "loss": 0.0002,
+      "step": 19138
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.509498295177789e-08,
+      "loss": 0.0001,
+      "step": 19139
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.507062834875792e-08,
+      "loss": 0.0002,
+      "step": 19140
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.5046273745737944e-08,
+      "loss": 0.0002,
+      "step": 19141
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.5021919142717976e-08,
+      "loss": 0.0002,
+      "step": 19142
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.4997564539698e-08,
+      "loss": 0.0003,
+      "step": 19143
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 3.497320993667803e-08,
+      "loss": 0.0002,
+      "step": 19144
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.494885533365806e-08,
+      "loss": 0.0002,
+      "step": 19145
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.492450073063809e-08,
+      "loss": 0.0001,
+      "step": 19146
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.4900146127618115e-08,
+      "loss": 0.0003,
+      "step": 19147
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.4875791524598147e-08,
+      "loss": 0.0003,
+      "step": 19148
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.485143692157818e-08,
+      "loss": 0.0003,
+      "step": 19149
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.482708231855821e-08,
+      "loss": 0.0002,
+      "step": 19150
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.4802727715538235e-08,
+      "loss": 1.6499,
+      "step": 19151
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.477837311251827e-08,
+      "loss": 0.0002,
+      "step": 19152
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.475401850949829e-08,
+      "loss": 0.0001,
+      "step": 19153
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 3.4729663906478324e-08,
+      "loss": 0.0002,
+      "step": 19154
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.470530930345835e-08,
+      "loss": 0.0003,
+      "step": 19155
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.468095470043838e-08,
+      "loss": 0.0003,
+      "step": 19156
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.4656600097418406e-08,
+      "loss": 0.0002,
+      "step": 19157
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.463224549439844e-08,
+      "loss": 0.0001,
+      "step": 19158
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.460789089137847e-08,
+      "loss": 0.0003,
+      "step": 19159
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.45835362883585e-08,
+      "loss": 0.0002,
+      "step": 19160
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.455918168533853e-08,
+      "loss": 0.0001,
+      "step": 19161
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.453482708231856e-08,
+      "loss": 0.0001,
+      "step": 19162
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.4510472479298584e-08,
+      "loss": 0.0002,
+      "step": 19163
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.4486117876278615e-08,
+      "loss": 0.0001,
+      "step": 19164
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 3.446176327325864e-08,
+      "loss": 0.0002,
+      "step": 19165
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.443740867023867e-08,
+      "loss": 0.0003,
+      "step": 19166
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.44130540672187e-08,
+      "loss": 0.0004,
+      "step": 19167
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.4388699464198736e-08,
+      "loss": 0.0002,
+      "step": 19168
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.436434486117876e-08,
+      "loss": 0.0004,
+      "step": 19169
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.433999025815879e-08,
+      "loss": 0.0003,
+      "step": 19170
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.431563565513882e-08,
+      "loss": 0.0001,
+      "step": 19171
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.429128105211885e-08,
+      "loss": 0.0003,
+      "step": 19172
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.4266926449098875e-08,
+      "loss": 0.0007,
+      "step": 19173
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.424257184607891e-08,
+      "loss": 0.0002,
+      "step": 19174
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 3.421821724305893e-08,
+      "loss": 0.0003,
+      "step": 19175
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.4193862640038964e-08,
+      "loss": 0.0002,
+      "step": 19176
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.416950803701899e-08,
+      "loss": 1.7528,
+      "step": 19177
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.414515343399903e-08,
+      "loss": 0.0003,
+      "step": 19178
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.412079883097905e-08,
+      "loss": 0.0002,
+      "step": 19179
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.4096444227959084e-08,
+      "loss": 0.0003,
+      "step": 19180
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.407208962493911e-08,
+      "loss": 0.0002,
+      "step": 19181
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.404773502191914e-08,
+      "loss": 0.0003,
+      "step": 19182
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.4023380418899166e-08,
+      "loss": 0.0003,
+      "step": 19183
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.39990258158792e-08,
+      "loss": 0.0002,
+      "step": 19184
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 3.397467121285923e-08,
+      "loss": 0.0003,
+      "step": 19185
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.3950316609839255e-08,
+      "loss": 0.0003,
+      "step": 19186
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.3925962006819293e-08,
+      "loss": 0.0004,
+      "step": 19187
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.390160740379932e-08,
+      "loss": 0.0002,
+      "step": 19188
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.387725280077935e-08,
+      "loss": 0.0002,
+      "step": 19189
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.3852898197759376e-08,
+      "loss": 0.0002,
+      "step": 19190
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.382854359473941e-08,
+      "loss": 0.0002,
+      "step": 19191
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.380418899171943e-08,
+      "loss": 0.0003,
+      "step": 19192
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.3779834388699464e-08,
+      "loss": 0.0001,
+      "step": 19193
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.375547978567949e-08,
+      "loss": 0.0004,
+      "step": 19194
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 3.373112518265952e-08,
+      "loss": 0.0002,
+      "step": 19195
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.3706770579639546e-08,
+      "loss": 0.0002,
+      "step": 19196
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.3682415976619585e-08,
+      "loss": 0.0001,
+      "step": 19197
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.365806137359961e-08,
+      "loss": 0.0002,
+      "step": 19198
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.363370677057964e-08,
+      "loss": 0.0001,
+      "step": 19199
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.360935216755967e-08,
+      "loss": 0.0001,
+      "step": 19200
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.35849975645397e-08,
+      "loss": 0.0002,
+      "step": 19201
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.3560642961519724e-08,
+      "loss": 0.0003,
+      "step": 19202
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.3536288358499756e-08,
+      "loss": 0.0003,
+      "step": 19203
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.351193375547978e-08,
+      "loss": 0.0002,
+      "step": 19204
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.348757915245981e-08,
+      "loss": 0.0001,
+      "step": 19205
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 3.3463224549439844e-08,
+      "loss": 0.0002,
+      "step": 19206
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.3438869946419876e-08,
+      "loss": 0.0001,
+      "step": 19207
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.34145153433999e-08,
+      "loss": 0.0004,
+      "step": 19208
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.339016074037993e-08,
+      "loss": 0.0002,
+      "step": 19209
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.336580613735996e-08,
+      "loss": 0.0002,
+      "step": 19210
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.334145153433999e-08,
+      "loss": 0.0002,
+      "step": 19211
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.3317096931320015e-08,
+      "loss": 0.0002,
+      "step": 19212
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.329274232830005e-08,
+      "loss": 0.0002,
+      "step": 19213
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.326838772528007e-08,
+      "loss": 0.0001,
+      "step": 19214
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.3244033122260104e-08,
+      "loss": 0.0001,
+      "step": 19215
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 3.3219678519240136e-08,
+      "loss": 0.0004,
+      "step": 19216
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.319532391622017e-08,
+      "loss": 0.0004,
+      "step": 19217
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.317096931320019e-08,
+      "loss": 0.0003,
+      "step": 19218
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.3146614710180225e-08,
+      "loss": 0.0003,
+      "step": 19219
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.312226010716025e-08,
+      "loss": 0.0002,
+      "step": 19220
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.309790550414028e-08,
+      "loss": 0.0003,
+      "step": 19221
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.3073550901120307e-08,
+      "loss": 0.0002,
+      "step": 19222
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.304919629810034e-08,
+      "loss": 0.0002,
+      "step": 19223
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.3024841695080364e-08,
+      "loss": 0.0002,
+      "step": 19224
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.3000487092060395e-08,
+      "loss": 0.0002,
+      "step": 19225
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 3.297613248904043e-08,
+      "loss": 0.0003,
+      "step": 19226
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.295177788602046e-08,
+      "loss": 0.0002,
+      "step": 19227
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.2927423283000484e-08,
+      "loss": 0.0002,
+      "step": 19228
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.2903068679980516e-08,
+      "loss": 0.0002,
+      "step": 19229
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.287871407696054e-08,
+      "loss": 0.0002,
+      "step": 19230
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.285435947394057e-08,
+      "loss": 0.0003,
+      "step": 19231
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.28300048709206e-08,
+      "loss": 0.0002,
+      "step": 19232
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.280565026790063e-08,
+      "loss": 0.0002,
+      "step": 19233
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.2781295664880655e-08,
+      "loss": 0.0001,
+      "step": 19234
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.2756941061860693e-08,
+      "loss": 0.0001,
+      "step": 19235
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.2732586458840725e-08,
+      "loss": 0.0002,
+      "step": 19236
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 3.270823185582075e-08,
+      "loss": 0.0002,
+      "step": 19237
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.268387725280078e-08,
+      "loss": 0.0002,
+      "step": 19238
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.265952264978081e-08,
+      "loss": 0.0002,
+      "step": 19239
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.263516804676084e-08,
+      "loss": 0.0002,
+      "step": 19240
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.2610813443740864e-08,
+      "loss": 0.0012,
+      "step": 19241
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.2586458840720896e-08,
+      "loss": 0.0002,
+      "step": 19242
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.256210423770092e-08,
+      "loss": 0.0002,
+      "step": 19243
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.253774963468095e-08,
+      "loss": 0.0002,
+      "step": 19244
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.2513395031660985e-08,
+      "loss": 0.0002,
+      "step": 19245
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.2489040428641016e-08,
+      "loss": 0.0003,
+      "step": 19246
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 3.246468582562104e-08,
+      "loss": 0.0001,
+      "step": 19247
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.2440331222601073e-08,
+      "loss": 0.0003,
+      "step": 19248
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.24159766195811e-08,
+      "loss": 0.0003,
+      "step": 19249
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.239162201656113e-08,
+      "loss": 0.0003,
+      "step": 19250
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.2367267413541156e-08,
+      "loss": 0.0002,
+      "step": 19251
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.234291281052119e-08,
+      "loss": 0.0002,
+      "step": 19252
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.231855820750121e-08,
+      "loss": 0.0002,
+      "step": 19253
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.229420360448125e-08,
+      "loss": 0.0003,
+      "step": 19254
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.2269849001461276e-08,
+      "loss": 0.0002,
+      "step": 19255
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.224549439844131e-08,
+      "loss": 0.0003,
+      "step": 19256
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 3.222113979542133e-08,
+      "loss": 0.0003,
+      "step": 19257
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.2196785192401365e-08,
+      "loss": 0.0003,
+      "step": 19258
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.217243058938139e-08,
+      "loss": 0.0002,
+      "step": 19259
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.214807598636142e-08,
+      "loss": 0.0002,
+      "step": 19260
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.212372138334145e-08,
+      "loss": 0.0003,
+      "step": 19261
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.209936678032148e-08,
+      "loss": 0.0137,
+      "step": 19262
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.2075012177301504e-08,
+      "loss": 0.0001,
+      "step": 19263
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.205065757428154e-08,
+      "loss": 0.0002,
+      "step": 19264
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.202630297126157e-08,
+      "loss": 0.0003,
+      "step": 19265
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.20019483682416e-08,
+      "loss": 0.0001,
+      "step": 19266
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.1977593765221624e-08,
+      "loss": 0.0002,
+      "step": 19267
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 3.1953239162201656e-08,
+      "loss": 0.0003,
+      "step": 19268
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.192888455918168e-08,
+      "loss": 0.0002,
+      "step": 19269
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.190452995616171e-08,
+      "loss": 0.0002,
+      "step": 19270
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.188017535314174e-08,
+      "loss": 0.0001,
+      "step": 19271
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.185582075012177e-08,
+      "loss": 0.0007,
+      "step": 19272
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.18314661471018e-08,
+      "loss": 0.0004,
+      "step": 19273
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.1807111544081834e-08,
+      "loss": 0.0003,
+      "step": 19274
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.178275694106186e-08,
+      "loss": 0.0001,
+      "step": 19275
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.175840233804189e-08,
+      "loss": 0.0002,
+      "step": 19276
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.1734047735021916e-08,
+      "loss": 0.0002,
+      "step": 19277
+    },
+    {
+      "epoch": 18.73,
+      "learning_rate": 3.170969313200195e-08,
+      "loss": 0.0003,
+      "step": 19278
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.168533852898197e-08,
+      "loss": 0.0003,
+      "step": 19279
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.1660983925962004e-08,
+      "loss": 0.0002,
+      "step": 19280
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.163662932294203e-08,
+      "loss": 0.0002,
+      "step": 19281
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.161227471992206e-08,
+      "loss": 0.0002,
+      "step": 19282
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.158792011690209e-08,
+      "loss": 0.0002,
+      "step": 19283
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.1563565513882125e-08,
+      "loss": 0.0003,
+      "step": 19284
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.153921091086215e-08,
+      "loss": 0.0001,
+      "step": 19285
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.151485630784218e-08,
+      "loss": 0.0001,
+      "step": 19286
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.149050170482221e-08,
+      "loss": 0.0002,
+      "step": 19287
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 3.146614710180224e-08,
+      "loss": 0.0003,
+      "step": 19288
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.1441792498782264e-08,
+      "loss": 0.0002,
+      "step": 19289
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.1417437895762296e-08,
+      "loss": 0.0002,
+      "step": 19290
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.139308329274232e-08,
+      "loss": 0.0003,
+      "step": 19291
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.136872868972235e-08,
+      "loss": 0.0003,
+      "step": 19292
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.134437408670239e-08,
+      "loss": 1.097,
+      "step": 19293
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.1320019483682416e-08,
+      "loss": 0.0002,
+      "step": 19294
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.129566488066245e-08,
+      "loss": 0.0002,
+      "step": 19295
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.127131027764247e-08,
+      "loss": 0.0002,
+      "step": 19296
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.1246955674622505e-08,
+      "loss": 0.0003,
+      "step": 19297
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 3.122260107160253e-08,
+      "loss": 0.0003,
+      "step": 19298
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.119824646858256e-08,
+      "loss": 0.0003,
+      "step": 19299
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.1173891865562594e-08,
+      "loss": 0.0004,
+      "step": 19300
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.114953726254262e-08,
+      "loss": 0.0003,
+      "step": 19301
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.112518265952265e-08,
+      "loss": 0.0003,
+      "step": 19302
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.1100828056502676e-08,
+      "loss": 0.0003,
+      "step": 19303
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.107647345348271e-08,
+      "loss": 0.0007,
+      "step": 19304
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.105211885046274e-08,
+      "loss": 0.0002,
+      "step": 19305
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.1027764247442765e-08,
+      "loss": 0.0002,
+      "step": 19306
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.1003409644422796e-08,
+      "loss": 0.0003,
+      "step": 19307
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.097905504140282e-08,
+      "loss": 0.0003,
+      "step": 19308
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 3.0954700438382853e-08,
+      "loss": 0.0002,
+      "step": 19309
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.0930345835362885e-08,
+      "loss": 0.0002,
+      "step": 19310
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.090599123234291e-08,
+      "loss": 0.0002,
+      "step": 19311
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.088163662932294e-08,
+      "loss": 0.0002,
+      "step": 19312
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.085728202630297e-08,
+      "loss": 0.0002,
+      "step": 19313
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.0832927423283e-08,
+      "loss": 0.0002,
+      "step": 19314
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.080857282026303e-08,
+      "loss": 0.0002,
+      "step": 19315
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.0784218217243056e-08,
+      "loss": 0.0002,
+      "step": 19316
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.075986361422309e-08,
+      "loss": 0.0003,
+      "step": 19317
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.073550901120311e-08,
+      "loss": 0.0002,
+      "step": 19318
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 3.0711154408183145e-08,
+      "loss": 0.0002,
+      "step": 19319
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.0686799805163176e-08,
+      "loss": 0.0004,
+      "step": 19320
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.06624452021432e-08,
+      "loss": 0.0002,
+      "step": 19321
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.0638090599123233e-08,
+      "loss": 0.0002,
+      "step": 19322
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.061373599610326e-08,
+      "loss": 0.0002,
+      "step": 19323
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.058938139308329e-08,
+      "loss": 0.0002,
+      "step": 19324
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.056502679006332e-08,
+      "loss": 0.0004,
+      "step": 19325
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.054067218704335e-08,
+      "loss": 0.0003,
+      "step": 19326
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.051631758402338e-08,
+      "loss": 0.0003,
+      "step": 19327
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.049196298100341e-08,
+      "loss": 0.0002,
+      "step": 19328
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 3.0467608377983436e-08,
+      "loss": 0.0004,
+      "step": 19329
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.044325377496347e-08,
+      "loss": 0.0003,
+      "step": 19330
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.041889917194349e-08,
+      "loss": 0.0003,
+      "step": 19331
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.0394544568923525e-08,
+      "loss": 0.0002,
+      "step": 19332
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.0370189965903557e-08,
+      "loss": 0.0002,
+      "step": 19333
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.034583536288358e-08,
+      "loss": 0.0002,
+      "step": 19334
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.0321480759863614e-08,
+      "loss": 0.0004,
+      "step": 19335
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.029712615684364e-08,
+      "loss": 0.0004,
+      "step": 19336
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.027277155382367e-08,
+      "loss": 0.0002,
+      "step": 19337
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.02484169508037e-08,
+      "loss": 0.0002,
+      "step": 19338
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.022406234778373e-08,
+      "loss": 0.0002,
+      "step": 19339
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 3.019970774476376e-08,
+      "loss": 0.0002,
+      "step": 19340
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 3.0175353141743784e-08,
+      "loss": 0.0002,
+      "step": 19341
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 3.0150998538723816e-08,
+      "loss": 0.0003,
+      "step": 19342
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 3.012664393570385e-08,
+      "loss": 0.0002,
+      "step": 19343
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 3.010228933268388e-08,
+      "loss": 0.0002,
+      "step": 19344
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 3.0077934729663905e-08,
+      "loss": 0.0001,
+      "step": 19345
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 3.0053580126643937e-08,
+      "loss": 0.0003,
+      "step": 19346
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 3.002922552362396e-08,
+      "loss": 0.0002,
+      "step": 19347
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 3.0004870920603994e-08,
+      "loss": 0.0002,
+      "step": 19348
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 2.9980516317584025e-08,
+      "loss": 0.0001,
+      "step": 19349
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 2.995616171456405e-08,
+      "loss": 0.0003,
+      "step": 19350
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.993180711154408e-08,
+      "loss": 0.0003,
+      "step": 19351
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.9907452508524114e-08,
+      "loss": 0.0001,
+      "step": 19352
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.988309790550414e-08,
+      "loss": 0.0002,
+      "step": 19353
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.985874330248417e-08,
+      "loss": 0.0002,
+      "step": 19354
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.9834388699464196e-08,
+      "loss": 0.0003,
+      "step": 19355
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.981003409644423e-08,
+      "loss": 0.0002,
+      "step": 19356
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.9785679493424256e-08,
+      "loss": 0.0004,
+      "step": 19357
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.9761324890404285e-08,
+      "loss": 0.0001,
+      "step": 19358
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.9736970287384313e-08,
+      "loss": 0.0004,
+      "step": 19359
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 2.9712615684364342e-08,
+      "loss": 0.0001,
+      "step": 19360
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.968826108134437e-08,
+      "loss": 0.3633,
+      "step": 19361
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.9663906478324402e-08,
+      "loss": 0.0002,
+      "step": 19362
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.963955187530443e-08,
+      "loss": 0.0003,
+      "step": 19363
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.961519727228446e-08,
+      "loss": 0.0002,
+      "step": 19364
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.9590842669264488e-08,
+      "loss": 0.0002,
+      "step": 19365
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.9566488066244516e-08,
+      "loss": 0.0005,
+      "step": 19366
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.9542133463224548e-08,
+      "loss": 0.0002,
+      "step": 19367
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.9517778860204576e-08,
+      "loss": 0.0003,
+      "step": 19368
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.9493424257184605e-08,
+      "loss": 0.0003,
+      "step": 19369
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 2.9469069654164633e-08,
+      "loss": 0.0003,
+      "step": 19370
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.9444715051144662e-08,
+      "loss": 0.0002,
+      "step": 19371
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.9420360448124697e-08,
+      "loss": 0.0003,
+      "step": 19372
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.9396005845104725e-08,
+      "loss": 0.0003,
+      "step": 19373
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.9371651242084754e-08,
+      "loss": 0.0003,
+      "step": 19374
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.9347296639064782e-08,
+      "loss": 0.0003,
+      "step": 19375
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.9322942036044814e-08,
+      "loss": 0.0005,
+      "step": 19376
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.9298587433024843e-08,
+      "loss": 0.0001,
+      "step": 19377
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.927423283000487e-08,
+      "loss": 0.0001,
+      "step": 19378
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.92498782269849e-08,
+      "loss": 0.0002,
+      "step": 19379
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.9225523623964928e-08,
+      "loss": 0.0002,
+      "step": 19380
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 2.920116902094496e-08,
+      "loss": 0.0004,
+      "step": 19381
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.9176814417924988e-08,
+      "loss": 0.0002,
+      "step": 19382
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.9152459814905017e-08,
+      "loss": 0.0003,
+      "step": 19383
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.9128105211885045e-08,
+      "loss": 0.0003,
+      "step": 19384
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.9103750608865074e-08,
+      "loss": 0.0002,
+      "step": 19385
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.9079396005845105e-08,
+      "loss": 0.0002,
+      "step": 19386
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.9055041402825134e-08,
+      "loss": 0.0002,
+      "step": 19387
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.9030686799805162e-08,
+      "loss": 0.0003,
+      "step": 19388
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.900633219678519e-08,
+      "loss": 0.0003,
+      "step": 19389
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.898197759376522e-08,
+      "loss": 0.0003,
+      "step": 19390
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 2.895762299074525e-08,
+      "loss": 0.0005,
+      "step": 19391
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.893326838772528e-08,
+      "loss": 0.0002,
+      "step": 19392
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8908913784705308e-08,
+      "loss": 0.0002,
+      "step": 19393
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8884559181685336e-08,
+      "loss": 0.0002,
+      "step": 19394
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8860204578665365e-08,
+      "loss": 0.0003,
+      "step": 19395
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8835849975645397e-08,
+      "loss": 0.0002,
+      "step": 19396
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8811495372625425e-08,
+      "loss": 0.0003,
+      "step": 19397
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8787140769605454e-08,
+      "loss": 0.0002,
+      "step": 19398
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8762786166585482e-08,
+      "loss": 0.0002,
+      "step": 19399
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8738431563565514e-08,
+      "loss": 0.0004,
+      "step": 19400
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 2.8714076960545542e-08,
+      "loss": 0.0003,
+      "step": 19401
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.868972235752557e-08,
+      "loss": 0.0002,
+      "step": 19402
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.86653677545056e-08,
+      "loss": 0.0002,
+      "step": 19403
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.8641013151485628e-08,
+      "loss": 0.0003,
+      "step": 19404
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.861665854846566e-08,
+      "loss": 0.0002,
+      "step": 19405
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.8592303945445688e-08,
+      "loss": 0.0004,
+      "step": 19406
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.8567949342425717e-08,
+      "loss": 0.0002,
+      "step": 19407
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.8543594739405745e-08,
+      "loss": 0.0003,
+      "step": 19408
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.8519240136385774e-08,
+      "loss": 0.0002,
+      "step": 19409
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.8494885533365805e-08,
+      "loss": 0.0002,
+      "step": 19410
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.8470530930345834e-08,
+      "loss": 0.0003,
+      "step": 19411
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 2.8446176327325862e-08,
+      "loss": 0.0003,
+      "step": 19412
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.842182172430589e-08,
+      "loss": 0.0002,
+      "step": 19413
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.839746712128592e-08,
+      "loss": 0.0002,
+      "step": 19414
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.837311251826595e-08,
+      "loss": 0.0004,
+      "step": 19415
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.834875791524598e-08,
+      "loss": 0.0002,
+      "step": 19416
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.8324403312226008e-08,
+      "loss": 0.0002,
+      "step": 19417
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.8300048709206036e-08,
+      "loss": 0.0002,
+      "step": 19418
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.8275694106186068e-08,
+      "loss": 0.0002,
+      "step": 19419
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.8251339503166097e-08,
+      "loss": 0.0001,
+      "step": 19420
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.8226984900146125e-08,
+      "loss": 0.0003,
+      "step": 19421
+    },
+    {
+      "epoch": 18.87,
+      "learning_rate": 2.8202630297126154e-08,
+      "loss": 0.0002,
+      "step": 19422
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.8178275694106182e-08,
+      "loss": 0.0003,
+      "step": 19423
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.8153921091086217e-08,
+      "loss": 0.0002,
+      "step": 19424
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.8129566488066246e-08,
+      "loss": 0.0004,
+      "step": 19425
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.8105211885046274e-08,
+      "loss": 0.0001,
+      "step": 19426
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.8080857282026303e-08,
+      "loss": 0.0003,
+      "step": 19427
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.805650267900633e-08,
+      "loss": 0.0001,
+      "step": 19428
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.8032148075986363e-08,
+      "loss": 0.0002,
+      "step": 19429
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.800779347296639e-08,
+      "loss": 0.0001,
+      "step": 19430
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.798343886994642e-08,
+      "loss": 0.0003,
+      "step": 19431
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 2.7959084266926448e-08,
+      "loss": 0.0001,
+      "step": 19432
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.7934729663906477e-08,
+      "loss": 0.0002,
+      "step": 19433
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.791037506088651e-08,
+      "loss": 0.0001,
+      "step": 19434
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.7886020457866537e-08,
+      "loss": 0.0001,
+      "step": 19435
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.7861665854846565e-08,
+      "loss": 0.0002,
+      "step": 19436
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.7837311251826594e-08,
+      "loss": 0.0002,
+      "step": 19437
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.7812956648806622e-08,
+      "loss": 0.0003,
+      "step": 19438
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.7788602045786654e-08,
+      "loss": 0.0002,
+      "step": 19439
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.7764247442766683e-08,
+      "loss": 0.0002,
+      "step": 19440
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.773989283974671e-08,
+      "loss": 0.0002,
+      "step": 19441
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 2.771553823672674e-08,
+      "loss": 0.0002,
+      "step": 19442
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.769118363370677e-08,
+      "loss": 0.0003,
+      "step": 19443
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.76668290306868e-08,
+      "loss": 0.0002,
+      "step": 19444
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.764247442766683e-08,
+      "loss": 0.0001,
+      "step": 19445
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.7618119824646857e-08,
+      "loss": 0.0002,
+      "step": 19446
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.7593765221626885e-08,
+      "loss": 0.0003,
+      "step": 19447
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.7569410618606917e-08,
+      "loss": 0.0002,
+      "step": 19448
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.7545056015586946e-08,
+      "loss": 0.0002,
+      "step": 19449
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.7520701412566974e-08,
+      "loss": 0.0004,
+      "step": 19450
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.7496346809547003e-08,
+      "loss": 0.0003,
+      "step": 19451
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.747199220652703e-08,
+      "loss": 0.0002,
+      "step": 19452
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 2.7447637603507063e-08,
+      "loss": 0.0002,
+      "step": 19453
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.742328300048709e-08,
+      "loss": 0.0001,
+      "step": 19454
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.739892839746712e-08,
+      "loss": 0.0001,
+      "step": 19455
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.7374573794447148e-08,
+      "loss": 0.0002,
+      "step": 19456
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.7350219191427177e-08,
+      "loss": 0.0003,
+      "step": 19457
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.732586458840721e-08,
+      "loss": 0.0004,
+      "step": 19458
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.7301509985387237e-08,
+      "loss": 0.0002,
+      "step": 19459
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.7277155382367265e-08,
+      "loss": 0.0002,
+      "step": 19460
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.7252800779347294e-08,
+      "loss": 0.0003,
+      "step": 19461
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.7228446176327322e-08,
+      "loss": 0.0001,
+      "step": 19462
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 2.7204091573307354e-08,
+      "loss": 0.0002,
+      "step": 19463
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.7179736970287383e-08,
+      "loss": 0.0002,
+      "step": 19464
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.715538236726741e-08,
+      "loss": 0.0004,
+      "step": 19465
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.713102776424744e-08,
+      "loss": 0.0003,
+      "step": 19466
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.710667316122747e-08,
+      "loss": 0.0003,
+      "step": 19467
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.70823185582075e-08,
+      "loss": 0.0002,
+      "step": 19468
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.7057963955187528e-08,
+      "loss": 0.0001,
+      "step": 19469
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.7033609352167557e-08,
+      "loss": 0.0001,
+      "step": 19470
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.7009254749147585e-08,
+      "loss": 0.001,
+      "step": 19471
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.6984900146127617e-08,
+      "loss": 0.0002,
+      "step": 19472
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 2.6960545543107646e-08,
+      "loss": 0.0002,
+      "step": 19473
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.6936190940087674e-08,
+      "loss": 0.0002,
+      "step": 19474
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.6911836337067702e-08,
+      "loss": 0.0002,
+      "step": 19475
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.688748173404773e-08,
+      "loss": 0.0002,
+      "step": 19476
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.6863127131027763e-08,
+      "loss": 0.0003,
+      "step": 19477
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.6838772528007794e-08,
+      "loss": 0.0001,
+      "step": 19478
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.6814417924987823e-08,
+      "loss": 0.0002,
+      "step": 19479
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.679006332196785e-08,
+      "loss": 0.0002,
+      "step": 19480
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.676570871894788e-08,
+      "loss": 0.0002,
+      "step": 19481
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.6741354115927912e-08,
+      "loss": 0.0001,
+      "step": 19482
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.671699951290794e-08,
+      "loss": 0.0003,
+      "step": 19483
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 2.669264490988797e-08,
+      "loss": 0.0002,
+      "step": 19484
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.6668290306867997e-08,
+      "loss": 0.0002,
+      "step": 19485
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.664393570384803e-08,
+      "loss": 0.0005,
+      "step": 19486
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.6619581100828057e-08,
+      "loss": 0.0026,
+      "step": 19487
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.6595226497808086e-08,
+      "loss": 0.0002,
+      "step": 19488
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.6570871894788114e-08,
+      "loss": 0.0002,
+      "step": 19489
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.6546517291768143e-08,
+      "loss": 0.0002,
+      "step": 19490
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.6522162688748175e-08,
+      "loss": 0.0004,
+      "step": 19491
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.6497808085728203e-08,
+      "loss": 0.0002,
+      "step": 19492
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.647345348270823e-08,
+      "loss": 0.0002,
+      "step": 19493
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 2.644909887968826e-08,
+      "loss": 0.0002,
+      "step": 19494
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.642474427666829e-08,
+      "loss": 0.0001,
+      "step": 19495
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.640038967364832e-08,
+      "loss": 0.0001,
+      "step": 19496
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.637603507062835e-08,
+      "loss": 0.0002,
+      "step": 19497
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.6351680467608377e-08,
+      "loss": 0.0001,
+      "step": 19498
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.6327325864588406e-08,
+      "loss": 0.0006,
+      "step": 19499
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.6302971261568434e-08,
+      "loss": 0.0003,
+      "step": 19500
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.6278616658548466e-08,
+      "loss": 0.0002,
+      "step": 19501
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.6254262055528494e-08,
+      "loss": 0.0002,
+      "step": 19502
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.6229907452508523e-08,
+      "loss": 0.0002,
+      "step": 19503
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 2.620555284948855e-08,
+      "loss": 0.0004,
+      "step": 19504
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.618119824646858e-08,
+      "loss": 0.0002,
+      "step": 19505
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.615684364344861e-08,
+      "loss": 0.0002,
+      "step": 19506
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.613248904042864e-08,
+      "loss": 0.0001,
+      "step": 19507
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.610813443740867e-08,
+      "loss": 0.0002,
+      "step": 19508
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.6083779834388697e-08,
+      "loss": 0.0002,
+      "step": 19509
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.605942523136873e-08,
+      "loss": 0.0003,
+      "step": 19510
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.6035070628348757e-08,
+      "loss": 0.0002,
+      "step": 19511
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.6010716025328786e-08,
+      "loss": 0.0003,
+      "step": 19512
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.5986361422308814e-08,
+      "loss": 0.0002,
+      "step": 19513
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 2.5962006819288843e-08,
+      "loss": 0.0002,
+      "step": 19514
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.5937652216268874e-08,
+      "loss": 0.0002,
+      "step": 19515
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.5913297613248903e-08,
+      "loss": 0.0003,
+      "step": 19516
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.588894301022893e-08,
+      "loss": 0.0002,
+      "step": 19517
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.586458840720896e-08,
+      "loss": 0.0003,
+      "step": 19518
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.584023380418899e-08,
+      "loss": 0.0003,
+      "step": 19519
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.581587920116902e-08,
+      "loss": 0.0002,
+      "step": 19520
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.579152459814905e-08,
+      "loss": 0.0003,
+      "step": 19521
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.5767169995129077e-08,
+      "loss": 0.0003,
+      "step": 19522
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.5742815392109106e-08,
+      "loss": 0.0002,
+      "step": 19523
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.5718460789089134e-08,
+      "loss": 0.0002,
+      "step": 19524
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 2.5694106186069166e-08,
+      "loss": 0.0001,
+      "step": 19525
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.5669751583049194e-08,
+      "loss": 0.0004,
+      "step": 19526
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.5645396980029223e-08,
+      "loss": 0.0002,
+      "step": 19527
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.562104237700925e-08,
+      "loss": 0.0002,
+      "step": 19528
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.559668777398928e-08,
+      "loss": 0.0002,
+      "step": 19529
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.557233317096931e-08,
+      "loss": 0.0003,
+      "step": 19530
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.554797856794934e-08,
+      "loss": 0.0001,
+      "step": 19531
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.552362396492937e-08,
+      "loss": 0.0001,
+      "step": 19532
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.54992693619094e-08,
+      "loss": 0.0002,
+      "step": 19533
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.5474914758889432e-08,
+      "loss": 0.0002,
+      "step": 19534
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 2.545056015586946e-08,
+      "loss": 0.0002,
+      "step": 19535
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.542620555284949e-08,
+      "loss": 0.0005,
+      "step": 19536
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.5401850949829517e-08,
+      "loss": 0.0002,
+      "step": 19537
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.5377496346809546e-08,
+      "loss": 0.0002,
+      "step": 19538
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.5353141743789578e-08,
+      "loss": 0.0003,
+      "step": 19539
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.5328787140769606e-08,
+      "loss": 0.0002,
+      "step": 19540
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.5304432537749635e-08,
+      "loss": 0.0002,
+      "step": 19541
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.5280077934729663e-08,
+      "loss": 0.0002,
+      "step": 19542
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.525572333170969e-08,
+      "loss": 0.0002,
+      "step": 19543
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.5231368728689723e-08,
+      "loss": 0.0002,
+      "step": 19544
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 2.5207014125669752e-08,
+      "loss": 0.0002,
+      "step": 19545
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.518265952264978e-08,
+      "loss": 0.0158,
+      "step": 19546
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.515830491962981e-08,
+      "loss": 0.0002,
+      "step": 19547
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.5133950316609837e-08,
+      "loss": 0.0004,
+      "step": 19548
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.510959571358987e-08,
+      "loss": 0.0002,
+      "step": 19549
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.5085241110569898e-08,
+      "loss": 0.0002,
+      "step": 19550
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.5060886507549926e-08,
+      "loss": 0.0005,
+      "step": 19551
+    },
+    {
+      "epoch": 19.0,
+      "eval_accuracy": 0.9990277102576568,
+      "eval_loss": 0.0139386635273695,
+      "eval_runtime": 2.7878,
+      "eval_samples_per_second": 737.845,
+      "eval_steps_per_second": 11.837,
+      "step": 19551
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.5036531904529955e-08,
+      "loss": 0.0002,
+      "step": 19552
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.5012177301509983e-08,
+      "loss": 0.0002,
+      "step": 19553
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.4987822698490015e-08,
+      "loss": 0.0002,
+      "step": 19554
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.4963468095470043e-08,
+      "loss": 0.0003,
+      "step": 19555
+    },
+    {
+      "epoch": 19.0,
+      "learning_rate": 2.4939113492450072e-08,
+      "loss": 0.0002,
+      "step": 19556
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.49147588894301e-08,
+      "loss": 0.0004,
+      "step": 19557
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.4890404286410132e-08,
+      "loss": 0.0003,
+      "step": 19558
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.486604968339016e-08,
+      "loss": 0.0005,
+      "step": 19559
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.484169508037019e-08,
+      "loss": 0.0003,
+      "step": 19560
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.4817340477350217e-08,
+      "loss": 0.0003,
+      "step": 19561
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.4792985874330246e-08,
+      "loss": 0.0002,
+      "step": 19562
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.4768631271310278e-08,
+      "loss": 0.0002,
+      "step": 19563
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.4744276668290306e-08,
+      "loss": 0.0002,
+      "step": 19564
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.4719922065270335e-08,
+      "loss": 0.0003,
+      "step": 19565
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 2.4695567462250363e-08,
+      "loss": 0.0001,
+      "step": 19566
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.467121285923039e-08,
+      "loss": 0.0005,
+      "step": 19567
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4646858256210423e-08,
+      "loss": 0.0001,
+      "step": 19568
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4622503653190452e-08,
+      "loss": 0.0002,
+      "step": 19569
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.459814905017048e-08,
+      "loss": 0.0003,
+      "step": 19570
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.457379444715051e-08,
+      "loss": 0.0002,
+      "step": 19571
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4549439844130537e-08,
+      "loss": 0.0002,
+      "step": 19572
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.452508524111057e-08,
+      "loss": 0.0004,
+      "step": 19573
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4500730638090597e-08,
+      "loss": 0.0003,
+      "step": 19574
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4476376035070626e-08,
+      "loss": 0.0001,
+      "step": 19575
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 2.4452021432050654e-08,
+      "loss": 0.0004,
+      "step": 19576
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.4427666829030686e-08,
+      "loss": 0.0003,
+      "step": 19577
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.4403312226010715e-08,
+      "loss": 0.0001,
+      "step": 19578
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.4378957622990743e-08,
+      "loss": 0.0003,
+      "step": 19579
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.435460301997077e-08,
+      "loss": 0.0001,
+      "step": 19580
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.43302484169508e-08,
+      "loss": 0.0003,
+      "step": 19581
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.4305893813930832e-08,
+      "loss": 0.0002,
+      "step": 19582
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.428153921091086e-08,
+      "loss": 0.0004,
+      "step": 19583
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.425718460789089e-08,
+      "loss": 0.0002,
+      "step": 19584
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.4232830004870917e-08,
+      "loss": 0.0002,
+      "step": 19585
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.4208475401850946e-08,
+      "loss": 0.0002,
+      "step": 19586
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 2.418412079883098e-08,
+      "loss": 0.0002,
+      "step": 19587
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.415976619581101e-08,
+      "loss": 0.0003,
+      "step": 19588
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.4135411592791038e-08,
+      "loss": 0.0002,
+      "step": 19589
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.4111056989771066e-08,
+      "loss": 0.0004,
+      "step": 19590
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.4086702386751095e-08,
+      "loss": 0.0002,
+      "step": 19591
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.4062347783731127e-08,
+      "loss": 0.0003,
+      "step": 19592
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.4037993180711155e-08,
+      "loss": 0.0003,
+      "step": 19593
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.4013638577691183e-08,
+      "loss": 0.0003,
+      "step": 19594
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.3989283974671212e-08,
+      "loss": 0.0001,
+      "step": 19595
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.396492937165124e-08,
+      "loss": 0.0003,
+      "step": 19596
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 2.3940574768631272e-08,
+      "loss": 0.0001,
+      "step": 19597
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.39162201656113e-08,
+      "loss": 0.0001,
+      "step": 19598
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.389186556259133e-08,
+      "loss": 0.0003,
+      "step": 19599
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.3867510959571358e-08,
+      "loss": 0.0002,
+      "step": 19600
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.384315635655139e-08,
+      "loss": 0.0002,
+      "step": 19601
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.3818801753531418e-08,
+      "loss": 0.0003,
+      "step": 19602
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.3794447150511446e-08,
+      "loss": 0.0002,
+      "step": 19603
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.3770092547491475e-08,
+      "loss": 0.0002,
+      "step": 19604
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.3745737944471503e-08,
+      "loss": 0.0003,
+      "step": 19605
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.3721383341451535e-08,
+      "loss": 0.0002,
+      "step": 19606
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 2.3697028738431564e-08,
+      "loss": 0.0004,
+      "step": 19607
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.3672674135411592e-08,
+      "loss": 0.0003,
+      "step": 19608
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.364831953239162e-08,
+      "loss": 0.0002,
+      "step": 19609
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.362396492937165e-08,
+      "loss": 0.0003,
+      "step": 19610
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.359961032635168e-08,
+      "loss": 0.0002,
+      "step": 19611
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.357525572333171e-08,
+      "loss": 0.0006,
+      "step": 19612
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.3550901120311738e-08,
+      "loss": 0.0002,
+      "step": 19613
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.3526546517291766e-08,
+      "loss": 0.0002,
+      "step": 19614
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.3502191914271795e-08,
+      "loss": 0.0004,
+      "step": 19615
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.3477837311251826e-08,
+      "loss": 0.0002,
+      "step": 19616
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 2.3453482708231855e-08,
+      "loss": 0.0001,
+      "step": 19617
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.3429128105211883e-08,
+      "loss": 0.0002,
+      "step": 19618
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.3404773502191912e-08,
+      "loss": 0.0002,
+      "step": 19619
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.338041889917194e-08,
+      "loss": 0.0002,
+      "step": 19620
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.3356064296151972e-08,
+      "loss": 0.0001,
+      "step": 19621
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.3331709693132e-08,
+      "loss": 0.0002,
+      "step": 19622
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.330735509011203e-08,
+      "loss": 0.0002,
+      "step": 19623
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.3283000487092058e-08,
+      "loss": 0.0002,
+      "step": 19624
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.325864588407209e-08,
+      "loss": 0.0002,
+      "step": 19625
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.3234291281052118e-08,
+      "loss": 0.0003,
+      "step": 19626
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.3209936678032146e-08,
+      "loss": 0.0002,
+      "step": 19627
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 2.3185582075012175e-08,
+      "loss": 0.0003,
+      "step": 19628
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.3161227471992203e-08,
+      "loss": 0.0002,
+      "step": 19629
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.3136872868972235e-08,
+      "loss": 0.0003,
+      "step": 19630
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.3112518265952264e-08,
+      "loss": 0.0001,
+      "step": 19631
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.3088163662932292e-08,
+      "loss": 0.0003,
+      "step": 19632
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.306380905991232e-08,
+      "loss": 0.0002,
+      "step": 19633
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.303945445689235e-08,
+      "loss": 0.0002,
+      "step": 19634
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.301509985387238e-08,
+      "loss": 0.0003,
+      "step": 19635
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.299074525085241e-08,
+      "loss": 0.0003,
+      "step": 19636
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.2966390647832438e-08,
+      "loss": 0.0002,
+      "step": 19637
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 2.2942036044812466e-08,
+      "loss": 0.0002,
+      "step": 19638
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.2917681441792495e-08,
+      "loss": 0.0001,
+      "step": 19639
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.289332683877253e-08,
+      "loss": 0.0002,
+      "step": 19640
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.2868972235752558e-08,
+      "loss": 0.0003,
+      "step": 19641
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.2844617632732587e-08,
+      "loss": 0.0003,
+      "step": 19642
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.2820263029712615e-08,
+      "loss": 0.0004,
+      "step": 19643
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.2795908426692647e-08,
+      "loss": 0.0008,
+      "step": 19644
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.2771553823672675e-08,
+      "loss": 0.0003,
+      "step": 19645
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.2747199220652704e-08,
+      "loss": 0.0003,
+      "step": 19646
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.2722844617632732e-08,
+      "loss": 0.0002,
+      "step": 19647
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 2.269849001461276e-08,
+      "loss": 0.0002,
+      "step": 19648
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2674135411592793e-08,
+      "loss": 0.0002,
+      "step": 19649
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.264978080857282e-08,
+      "loss": 0.0002,
+      "step": 19650
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.262542620555285e-08,
+      "loss": 0.0003,
+      "step": 19651
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2601071602532878e-08,
+      "loss": 0.0002,
+      "step": 19652
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2576716999512906e-08,
+      "loss": 0.0002,
+      "step": 19653
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2552362396492938e-08,
+      "loss": 0.0002,
+      "step": 19654
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2528007793472967e-08,
+      "loss": 0.0004,
+      "step": 19655
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2503653190452995e-08,
+      "loss": 0.0002,
+      "step": 19656
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2479298587433024e-08,
+      "loss": 0.0002,
+      "step": 19657
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2454943984413052e-08,
+      "loss": 0.0004,
+      "step": 19658
+    },
+    {
+      "epoch": 19.1,
+      "learning_rate": 2.2430589381393084e-08,
+      "loss": 0.0002,
+      "step": 19659
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.2406234778373112e-08,
+      "loss": 0.0001,
+      "step": 19660
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.238188017535314e-08,
+      "loss": 0.0002,
+      "step": 19661
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.235752557233317e-08,
+      "loss": 0.0003,
+      "step": 19662
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.2333170969313198e-08,
+      "loss": 0.0001,
+      "step": 19663
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.230881636629323e-08,
+      "loss": 0.0003,
+      "step": 19664
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.2284461763273258e-08,
+      "loss": 0.0002,
+      "step": 19665
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.2260107160253287e-08,
+      "loss": 0.0002,
+      "step": 19666
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.2235752557233315e-08,
+      "loss": 0.0004,
+      "step": 19667
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.2211397954213347e-08,
+      "loss": 0.0003,
+      "step": 19668
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 2.2187043351193375e-08,
+      "loss": 0.0002,
+      "step": 19669
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.2162688748173404e-08,
+      "loss": 0.0003,
+      "step": 19670
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.2138334145153432e-08,
+      "loss": 0.0002,
+      "step": 19671
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.211397954213346e-08,
+      "loss": 0.0003,
+      "step": 19672
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.2089624939113492e-08,
+      "loss": 0.0002,
+      "step": 19673
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.206527033609352e-08,
+      "loss": 0.0003,
+      "step": 19674
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.204091573307355e-08,
+      "loss": 0.0002,
+      "step": 19675
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.2016561130053578e-08,
+      "loss": 0.0002,
+      "step": 19676
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.1992206527033606e-08,
+      "loss": 0.0003,
+      "step": 19677
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.1967851924013638e-08,
+      "loss": 1.811,
+      "step": 19678
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 2.1943497320993667e-08,
+      "loss": 0.0002,
+      "step": 19679
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.1919142717973695e-08,
+      "loss": 0.0002,
+      "step": 19680
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.1894788114953724e-08,
+      "loss": 0.0002,
+      "step": 19681
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.1870433511933752e-08,
+      "loss": 0.0002,
+      "step": 19682
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.1846078908913784e-08,
+      "loss": 0.0001,
+      "step": 19683
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.1821724305893812e-08,
+      "loss": 0.0003,
+      "step": 19684
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.179736970287384e-08,
+      "loss": 0.0004,
+      "step": 19685
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.177301509985387e-08,
+      "loss": 0.0002,
+      "step": 19686
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.1748660496833898e-08,
+      "loss": 0.0004,
+      "step": 19687
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.172430589381393e-08,
+      "loss": 0.0002,
+      "step": 19688
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 2.1699951290793958e-08,
+      "loss": 0.0003,
+      "step": 19689
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1675596687773986e-08,
+      "loss": 0.0002,
+      "step": 19690
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1651242084754015e-08,
+      "loss": 0.0003,
+      "step": 19691
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1626887481734047e-08,
+      "loss": 0.0002,
+      "step": 19692
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.160253287871408e-08,
+      "loss": 0.0003,
+      "step": 19693
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1578178275694107e-08,
+      "loss": 0.0002,
+      "step": 19694
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1553823672674135e-08,
+      "loss": 0.0004,
+      "step": 19695
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1529469069654164e-08,
+      "loss": 0.0002,
+      "step": 19696
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1505114466634196e-08,
+      "loss": 0.0002,
+      "step": 19697
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1480759863614224e-08,
+      "loss": 0.0002,
+      "step": 19698
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.1456405260594253e-08,
+      "loss": 0.0003,
+      "step": 19699
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 2.143205065757428e-08,
+      "loss": 0.0001,
+      "step": 19700
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.140769605455431e-08,
+      "loss": 0.0001,
+      "step": 19701
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.138334145153434e-08,
+      "loss": 0.0001,
+      "step": 19702
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.135898684851437e-08,
+      "loss": 0.0002,
+      "step": 19703
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.13346322454944e-08,
+      "loss": 0.0001,
+      "step": 19704
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.1310277642474427e-08,
+      "loss": 0.0001,
+      "step": 19705
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.1285923039454455e-08,
+      "loss": 0.0002,
+      "step": 19706
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.1261568436434487e-08,
+      "loss": 0.0002,
+      "step": 19707
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.1237213833414516e-08,
+      "loss": 0.0002,
+      "step": 19708
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.1212859230394544e-08,
+      "loss": 0.0002,
+      "step": 19709
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 2.1188504627374573e-08,
+      "loss": 0.0003,
+      "step": 19710
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.11641500243546e-08,
+      "loss": 0.0001,
+      "step": 19711
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.1139795421334633e-08,
+      "loss": 0.0001,
+      "step": 19712
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.111544081831466e-08,
+      "loss": 0.0002,
+      "step": 19713
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.109108621529469e-08,
+      "loss": 0.0002,
+      "step": 19714
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.1066731612274718e-08,
+      "loss": 0.0001,
+      "step": 19715
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.104237700925475e-08,
+      "loss": 0.0004,
+      "step": 19716
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.101802240623478e-08,
+      "loss": 0.0001,
+      "step": 19717
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.0993667803214807e-08,
+      "loss": 0.0001,
+      "step": 19718
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.0969313200194835e-08,
+      "loss": 0.0002,
+      "step": 19719
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 2.0944958597174864e-08,
+      "loss": 0.0002,
+      "step": 19720
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.0920603994154896e-08,
+      "loss": 0.0002,
+      "step": 19721
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.0896249391134924e-08,
+      "loss": 0.0001,
+      "step": 19722
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.0871894788114953e-08,
+      "loss": 0.0004,
+      "step": 19723
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.084754018509498e-08,
+      "loss": 0.0001,
+      "step": 19724
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.082318558207501e-08,
+      "loss": 0.0002,
+      "step": 19725
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.079883097905504e-08,
+      "loss": 0.0002,
+      "step": 19726
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.077447637603507e-08,
+      "loss": 0.0001,
+      "step": 19727
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.0750121773015098e-08,
+      "loss": 0.0003,
+      "step": 19728
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.0725767169995127e-08,
+      "loss": 0.0002,
+      "step": 19729
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.0701412566975155e-08,
+      "loss": 0.0033,
+      "step": 19730
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 2.0677057963955187e-08,
+      "loss": 0.0002,
+      "step": 19731
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.0652703360935215e-08,
+      "loss": 0.0002,
+      "step": 19732
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.0628348757915244e-08,
+      "loss": 0.0003,
+      "step": 19733
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.0603994154895272e-08,
+      "loss": 0.0002,
+      "step": 19734
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.0579639551875304e-08,
+      "loss": 0.0001,
+      "step": 19735
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.0555284948855333e-08,
+      "loss": 0.0002,
+      "step": 19736
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.053093034583536e-08,
+      "loss": 0.0003,
+      "step": 19737
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.050657574281539e-08,
+      "loss": 0.0003,
+      "step": 19738
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.0482221139795418e-08,
+      "loss": 0.0004,
+      "step": 19739
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.045786653677545e-08,
+      "loss": 0.0003,
+      "step": 19740
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 2.043351193375548e-08,
+      "loss": 0.0002,
+      "step": 19741
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0409157330735507e-08,
+      "loss": 0.0003,
+      "step": 19742
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0384802727715535e-08,
+      "loss": 0.0002,
+      "step": 19743
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0360448124695564e-08,
+      "loss": 0.0002,
+      "step": 19744
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0336093521675596e-08,
+      "loss": 0.0005,
+      "step": 19745
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0311738918655624e-08,
+      "loss": 0.0002,
+      "step": 19746
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0287384315635653e-08,
+      "loss": 0.0003,
+      "step": 19747
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0263029712615684e-08,
+      "loss": 0.0002,
+      "step": 19748
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0238675109595713e-08,
+      "loss": 0.0002,
+      "step": 19749
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0214320506575745e-08,
+      "loss": 1.2696,
+      "step": 19750
+    },
+    {
+      "epoch": 19.19,
+      "learning_rate": 2.0189965903555773e-08,
+      "loss": 0.0002,
+      "step": 19751
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.01656113005358e-08,
+      "loss": 0.0002,
+      "step": 19752
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.014125669751583e-08,
+      "loss": 0.0003,
+      "step": 19753
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.011690209449586e-08,
+      "loss": 0.0002,
+      "step": 19754
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.009254749147589e-08,
+      "loss": 0.0004,
+      "step": 19755
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.006819288845592e-08,
+      "loss": 0.0002,
+      "step": 19756
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.0043838285435947e-08,
+      "loss": 0.0004,
+      "step": 19757
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 2.0019483682415976e-08,
+      "loss": 0.0002,
+      "step": 19758
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.9995129079396007e-08,
+      "loss": 0.0001,
+      "step": 19759
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.9970774476376036e-08,
+      "loss": 0.0006,
+      "step": 19760
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.9946419873356064e-08,
+      "loss": 0.0004,
+      "step": 19761
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.9922065270336093e-08,
+      "loss": 0.0001,
+      "step": 19762
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.989771066731612e-08,
+      "loss": 0.0003,
+      "step": 19763
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.9873356064296153e-08,
+      "loss": 0.0002,
+      "step": 19764
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.984900146127618e-08,
+      "loss": 0.0002,
+      "step": 19765
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.982464685825621e-08,
+      "loss": 0.0003,
+      "step": 19766
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.980029225523624e-08,
+      "loss": 0.0002,
+      "step": 19767
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.9775937652216267e-08,
+      "loss": 0.0002,
+      "step": 19768
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.97515830491963e-08,
+      "loss": 0.0002,
+      "step": 19769
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.9727228446176327e-08,
+      "loss": 0.0001,
+      "step": 19770
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.9702873843156356e-08,
+      "loss": 0.0003,
+      "step": 19771
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.9678519240136384e-08,
+      "loss": 0.0001,
+      "step": 19772
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.9654164637116413e-08,
+      "loss": 0.0003,
+      "step": 19773
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.9629810034096444e-08,
+      "loss": 0.0002,
+      "step": 19774
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.9605455431076473e-08,
+      "loss": 0.0035,
+      "step": 19775
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.95811008280565e-08,
+      "loss": 0.0002,
+      "step": 19776
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.955674622503653e-08,
+      "loss": 0.0001,
+      "step": 19777
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.953239162201656e-08,
+      "loss": 0.0001,
+      "step": 19778
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.950803701899659e-08,
+      "loss": 0.0002,
+      "step": 19779
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.948368241597662e-08,
+      "loss": 0.0002,
+      "step": 19780
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.9459327812956647e-08,
+      "loss": 0.0002,
+      "step": 19781
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.9434973209936676e-08,
+      "loss": 0.0003,
+      "step": 19782
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.9410618606916707e-08,
+      "loss": 0.0002,
+      "step": 19783
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.9386264003896736e-08,
+      "loss": 0.0002,
+      "step": 19784
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.9361909400876764e-08,
+      "loss": 0.0001,
+      "step": 19785
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.9337554797856793e-08,
+      "loss": 0.0002,
+      "step": 19786
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.931320019483682e-08,
+      "loss": 0.0003,
+      "step": 19787
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.9288845591816853e-08,
+      "loss": 0.0002,
+      "step": 19788
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.926449098879688e-08,
+      "loss": 0.0001,
+      "step": 19789
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.924013638577691e-08,
+      "loss": 0.0003,
+      "step": 19790
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.921578178275694e-08,
+      "loss": 0.0002,
+      "step": 19791
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.9191427179736967e-08,
+      "loss": 0.0001,
+      "step": 19792
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.9167072576717e-08,
+      "loss": 0.0003,
+      "step": 19793
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.9142717973697027e-08,
+      "loss": 0.0004,
+      "step": 19794
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.9118363370677056e-08,
+      "loss": 0.0002,
+      "step": 19795
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.9094008767657084e-08,
+      "loss": 0.0002,
+      "step": 19796
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.9069654164637113e-08,
+      "loss": 0.0003,
+      "step": 19797
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.9045299561617144e-08,
+      "loss": 0.0002,
+      "step": 19798
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.9020944958597173e-08,
+      "loss": 0.0003,
+      "step": 19799
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.89965903555772e-08,
+      "loss": 0.0002,
+      "step": 19800
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.897223575255723e-08,
+      "loss": 0.0001,
+      "step": 19801
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.8947881149537265e-08,
+      "loss": 0.0006,
+      "step": 19802
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 1.8923526546517293e-08,
+      "loss": 0.0002,
+      "step": 19803
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8899171943497322e-08,
+      "loss": 0.0003,
+      "step": 19804
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.887481734047735e-08,
+      "loss": 0.0004,
+      "step": 19805
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.885046273745738e-08,
+      "loss": 0.0004,
+      "step": 19806
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.882610813443741e-08,
+      "loss": 0.0002,
+      "step": 19807
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.880175353141744e-08,
+      "loss": 0.0003,
+      "step": 19808
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8777398928397468e-08,
+      "loss": 0.0001,
+      "step": 19809
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8753044325377496e-08,
+      "loss": 0.0002,
+      "step": 19810
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8728689722357524e-08,
+      "loss": 0.0002,
+      "step": 19811
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8704335119337556e-08,
+      "loss": 0.0002,
+      "step": 19812
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.8679980516317585e-08,
+      "loss": 0.0002,
+      "step": 19813
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8655625913297613e-08,
+      "loss": 0.0001,
+      "step": 19814
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8631271310277642e-08,
+      "loss": 0.0003,
+      "step": 19815
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.860691670725767e-08,
+      "loss": 0.0002,
+      "step": 19816
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8582562104237702e-08,
+      "loss": 0.0003,
+      "step": 19817
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.855820750121773e-08,
+      "loss": 0.0001,
+      "step": 19818
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.853385289819776e-08,
+      "loss": 0.0002,
+      "step": 19819
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8509498295177787e-08,
+      "loss": 0.0002,
+      "step": 19820
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8485143692157816e-08,
+      "loss": 0.0085,
+      "step": 19821
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8460789089137848e-08,
+      "loss": 0.0003,
+      "step": 19822
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.8436434486117876e-08,
+      "loss": 0.0003,
+      "step": 19823
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8412079883097905e-08,
+      "loss": 0.0001,
+      "step": 19824
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8387725280077933e-08,
+      "loss": 0.0041,
+      "step": 19825
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8363370677057965e-08,
+      "loss": 0.0002,
+      "step": 19826
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8339016074037993e-08,
+      "loss": 0.0004,
+      "step": 19827
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8314661471018022e-08,
+      "loss": 0.0001,
+      "step": 19828
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.829030686799805e-08,
+      "loss": 0.0002,
+      "step": 19829
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.826595226497808e-08,
+      "loss": 0.0005,
+      "step": 19830
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.824159766195811e-08,
+      "loss": 0.0002,
+      "step": 19831
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.821724305893814e-08,
+      "loss": 0.0003,
+      "step": 19832
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.8192888455918167e-08,
+      "loss": 0.0003,
+      "step": 19833
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8168533852898196e-08,
+      "loss": 0.0002,
+      "step": 19834
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8144179249878224e-08,
+      "loss": 0.0003,
+      "step": 19835
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8119824646858256e-08,
+      "loss": 0.0002,
+      "step": 19836
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8095470043838285e-08,
+      "loss": 0.0002,
+      "step": 19837
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.8071115440818313e-08,
+      "loss": 0.0002,
+      "step": 19838
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.804676083779834e-08,
+      "loss": 0.0002,
+      "step": 19839
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.802240623477837e-08,
+      "loss": 0.0001,
+      "step": 19840
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.7998051631758402e-08,
+      "loss": 0.0003,
+      "step": 19841
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.797369702873843e-08,
+      "loss": 0.0006,
+      "step": 19842
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.794934242571846e-08,
+      "loss": 0.0003,
+      "step": 19843
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 1.7924987822698487e-08,
+      "loss": 0.0003,
+      "step": 19844
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.7900633219678516e-08,
+      "loss": 0.0002,
+      "step": 19845
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.7876278616658548e-08,
+      "loss": 0.0002,
+      "step": 19846
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.7851924013638576e-08,
+      "loss": 0.0004,
+      "step": 19847
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.7827569410618604e-08,
+      "loss": 0.0002,
+      "step": 19848
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.7803214807598633e-08,
+      "loss": 0.0002,
+      "step": 19849
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.7778860204578665e-08,
+      "loss": 0.0002,
+      "step": 19850
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.7754505601558693e-08,
+      "loss": 0.0002,
+      "step": 19851
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.7730150998538722e-08,
+      "loss": 0.0002,
+      "step": 19852
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.770579639551875e-08,
+      "loss": 0.0001,
+      "step": 19853
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.768144179249878e-08,
+      "loss": 0.0002,
+      "step": 19854
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.7657087189478814e-08,
+      "loss": 0.0007,
+      "step": 19855
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.7632732586458842e-08,
+      "loss": 0.0002,
+      "step": 19856
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.760837798343887e-08,
+      "loss": 0.0002,
+      "step": 19857
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.75840233804189e-08,
+      "loss": 0.0005,
+      "step": 19858
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.7559668777398928e-08,
+      "loss": 0.0004,
+      "step": 19859
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.753531417437896e-08,
+      "loss": 0.0002,
+      "step": 19860
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.7510959571358988e-08,
+      "loss": 0.0002,
+      "step": 19861
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.7486604968339016e-08,
+      "loss": 0.0002,
+      "step": 19862
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.7462250365319045e-08,
+      "loss": 0.0003,
+      "step": 19863
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.7437895762299073e-08,
+      "loss": 0.0003,
+      "step": 19864
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.7413541159279105e-08,
+      "loss": 0.0002,
+      "step": 19865
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.7389186556259134e-08,
+      "loss": 0.0002,
+      "step": 19866
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.7364831953239162e-08,
+      "loss": 0.0004,
+      "step": 19867
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.734047735021919e-08,
+      "loss": 0.0002,
+      "step": 19868
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.731612274719922e-08,
+      "loss": 0.0002,
+      "step": 19869
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.729176814417925e-08,
+      "loss": 0.0002,
+      "step": 19870
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.726741354115928e-08,
+      "loss": 0.0002,
+      "step": 19871
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.7243058938139308e-08,
+      "loss": 0.0002,
+      "step": 19872
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.7218704335119336e-08,
+      "loss": 0.0002,
+      "step": 19873
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.7194349732099368e-08,
+      "loss": 0.0002,
+      "step": 19874
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.7169995129079396e-08,
+      "loss": 0.0002,
+      "step": 19875
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.7145640526059425e-08,
+      "loss": 0.0002,
+      "step": 19876
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.7121285923039453e-08,
+      "loss": 0.0002,
+      "step": 19877
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.7096931320019482e-08,
+      "loss": 0.0002,
+      "step": 19878
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.7072576716999514e-08,
+      "loss": 0.0002,
+      "step": 19879
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.7048222113979542e-08,
+      "loss": 0.0001,
+      "step": 19880
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.702386751095957e-08,
+      "loss": 0.0002,
+      "step": 19881
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.69995129079396e-08,
+      "loss": 0.0002,
+      "step": 19882
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.6975158304919628e-08,
+      "loss": 0.0001,
+      "step": 19883
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.695080370189966e-08,
+      "loss": 0.0002,
+      "step": 19884
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.6926449098879688e-08,
+      "loss": 0.0002,
+      "step": 19885
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.6902094495859716e-08,
+      "loss": 0.0003,
+      "step": 19886
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.6877739892839745e-08,
+      "loss": 0.0002,
+      "step": 19887
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.6853385289819773e-08,
+      "loss": 0.0002,
+      "step": 19888
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.6829030686799805e-08,
+      "loss": 0.0002,
+      "step": 19889
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.6804676083779833e-08,
+      "loss": 0.0004,
+      "step": 19890
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.6780321480759862e-08,
+      "loss": 0.0004,
+      "step": 19891
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.675596687773989e-08,
+      "loss": 0.0003,
+      "step": 19892
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.6731612274719922e-08,
+      "loss": 0.0002,
+      "step": 19893
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.670725767169995e-08,
+      "loss": 1.6227,
+      "step": 19894
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 1.668290306867998e-08,
+      "loss": 0.0003,
+      "step": 19895
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6658548465660008e-08,
+      "loss": 0.0002,
+      "step": 19896
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6634193862640036e-08,
+      "loss": 0.0002,
+      "step": 19897
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6609839259620068e-08,
+      "loss": 0.0001,
+      "step": 19898
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6585484656600096e-08,
+      "loss": 0.0002,
+      "step": 19899
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6561130053580125e-08,
+      "loss": 0.0002,
+      "step": 19900
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6536775450560153e-08,
+      "loss": 0.3922,
+      "step": 19901
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6512420847540182e-08,
+      "loss": 0.0002,
+      "step": 19902
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6488066244520214e-08,
+      "loss": 0.0002,
+      "step": 19903
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.6463711641500242e-08,
+      "loss": 0.0002,
+      "step": 19904
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.643935703848027e-08,
+      "loss": 0.0004,
+      "step": 19905
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.64150024354603e-08,
+      "loss": 0.0003,
+      "step": 19906
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.6390647832440327e-08,
+      "loss": 0.0002,
+      "step": 19907
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.6366293229420363e-08,
+      "loss": 0.0002,
+      "step": 19908
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.634193862640039e-08,
+      "loss": 0.0003,
+      "step": 19909
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.631758402338042e-08,
+      "loss": 0.0002,
+      "step": 19910
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.6293229420360448e-08,
+      "loss": 0.0003,
+      "step": 19911
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.6268874817340476e-08,
+      "loss": 0.0005,
+      "step": 19912
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.6244520214320508e-08,
+      "loss": 0.0003,
+      "step": 19913
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.6220165611300537e-08,
+      "loss": 0.0002,
+      "step": 19914
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.6195811008280565e-08,
+      "loss": 0.0003,
+      "step": 19915
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.6171456405260594e-08,
+      "loss": 0.0002,
+      "step": 19916
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.6147101802240625e-08,
+      "loss": 0.0003,
+      "step": 19917
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.6122747199220654e-08,
+      "loss": 0.0001,
+      "step": 19918
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.6098392596200682e-08,
+      "loss": 0.0011,
+      "step": 19919
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.607403799318071e-08,
+      "loss": 0.0003,
+      "step": 19920
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.604968339016074e-08,
+      "loss": 0.0003,
+      "step": 19921
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.602532878714077e-08,
+      "loss": 0.0001,
+      "step": 19922
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.60009741841208e-08,
+      "loss": 0.0003,
+      "step": 19923
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.5976619581100828e-08,
+      "loss": 0.0002,
+      "step": 19924
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.5952264978080857e-08,
+      "loss": 0.0004,
+      "step": 19925
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 1.5927910375060885e-08,
+      "loss": 0.0002,
+      "step": 19926
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.5903555772040917e-08,
+      "loss": 0.0003,
+      "step": 19927
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.5879201169020945e-08,
+      "loss": 0.0005,
+      "step": 19928
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.5854846566000974e-08,
+      "loss": 0.0002,
+      "step": 19929
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.5830491962981002e-08,
+      "loss": 0.0002,
+      "step": 19930
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.580613735996103e-08,
+      "loss": 0.0001,
+      "step": 19931
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.5781782756941062e-08,
+      "loss": 0.0002,
+      "step": 19932
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.575742815392109e-08,
+      "loss": 0.0009,
+      "step": 19933
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.573307355090112e-08,
+      "loss": 0.0003,
+      "step": 19934
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.5708718947881148e-08,
+      "loss": 0.0001,
+      "step": 19935
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 1.5684364344861176e-08,
+      "loss": 0.0001,
+      "step": 19936
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.5660009741841208e-08,
+      "loss": 0.0002,
+      "step": 19937
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.5635655138821237e-08,
+      "loss": 0.0004,
+      "step": 19938
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.5611300535801265e-08,
+      "loss": 0.0003,
+      "step": 19939
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.5586945932781297e-08,
+      "loss": 0.0002,
+      "step": 19940
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.5562591329761325e-08,
+      "loss": 0.0001,
+      "step": 19941
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.5538236726741354e-08,
+      "loss": 0.0003,
+      "step": 19942
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.5513882123721382e-08,
+      "loss": 0.0003,
+      "step": 19943
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.548952752070141e-08,
+      "loss": 0.0001,
+      "step": 19944
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.5465172917681443e-08,
+      "loss": 0.0003,
+      "step": 19945
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.544081831466147e-08,
+      "loss": 0.0003,
+      "step": 19946
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 1.54164637116415e-08,
+      "loss": 0.0002,
+      "step": 19947
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5392109108621528e-08,
+      "loss": 0.0003,
+      "step": 19948
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5367754505601556e-08,
+      "loss": 0.0002,
+      "step": 19949
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5343399902581588e-08,
+      "loss": 0.0002,
+      "step": 19950
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5319045299561617e-08,
+      "loss": 0.0004,
+      "step": 19951
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5294690696541645e-08,
+      "loss": 0.0003,
+      "step": 19952
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5270336093521674e-08,
+      "loss": 0.0002,
+      "step": 19953
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5245981490501705e-08,
+      "loss": 0.0002,
+      "step": 19954
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5221626887481734e-08,
+      "loss": 0.0004,
+      "step": 19955
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.5197272284461762e-08,
+      "loss": 0.0002,
+      "step": 19956
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 1.517291768144179e-08,
+      "loss": 0.0002,
+      "step": 19957
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.514856307842182e-08,
+      "loss": 0.0002,
+      "step": 19958
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.512420847540185e-08,
+      "loss": 0.0002,
+      "step": 19959
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.509985387238188e-08,
+      "loss": 0.0001,
+      "step": 19960
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.5075499269361908e-08,
+      "loss": 0.0005,
+      "step": 19961
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.505114466634194e-08,
+      "loss": 0.0003,
+      "step": 19962
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.5026790063321968e-08,
+      "loss": 0.0001,
+      "step": 19963
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.5002435460301997e-08,
+      "loss": 0.0003,
+      "step": 19964
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.4978080857282025e-08,
+      "loss": 0.0003,
+      "step": 19965
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.4953726254262057e-08,
+      "loss": 0.0003,
+      "step": 19966
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 1.4929371651242086e-08,
+      "loss": 0.0004,
+      "step": 19967
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4905017048222114e-08,
+      "loss": 0.0003,
+      "step": 19968
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4880662445202142e-08,
+      "loss": 0.0006,
+      "step": 19969
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4856307842182171e-08,
+      "loss": 0.0003,
+      "step": 19970
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4831953239162201e-08,
+      "loss": 0.0003,
+      "step": 19971
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.480759863614223e-08,
+      "loss": 0.0003,
+      "step": 19972
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4783244033122258e-08,
+      "loss": 0.0002,
+      "step": 19973
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4758889430102288e-08,
+      "loss": 0.0002,
+      "step": 19974
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4734534827082317e-08,
+      "loss": 0.0003,
+      "step": 19975
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4710180224062348e-08,
+      "loss": 0.0003,
+      "step": 19976
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4685825621042377e-08,
+      "loss": 0.0002,
+      "step": 19977
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 1.4661471018022407e-08,
+      "loss": 0.0002,
+      "step": 19978
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4637116415002435e-08,
+      "loss": 0.0002,
+      "step": 19979
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4612761811982464e-08,
+      "loss": 0.0002,
+      "step": 19980
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4588407208962494e-08,
+      "loss": 0.0002,
+      "step": 19981
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4564052605942523e-08,
+      "loss": 0.0001,
+      "step": 19982
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4539698002922553e-08,
+      "loss": 0.0004,
+      "step": 19983
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4515343399902581e-08,
+      "loss": 0.0002,
+      "step": 19984
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.449098879688261e-08,
+      "loss": 0.0001,
+      "step": 19985
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.446663419386264e-08,
+      "loss": 0.0001,
+      "step": 19986
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4442279590842668e-08,
+      "loss": 0.0003,
+      "step": 19987
+    },
+    {
+      "epoch": 19.42,
+      "learning_rate": 1.4417924987822698e-08,
+      "loss": 0.0002,
+      "step": 19988
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4393570384802727e-08,
+      "loss": 0.0001,
+      "step": 19989
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4369215781782757e-08,
+      "loss": 0.0002,
+      "step": 19990
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4344861178762785e-08,
+      "loss": 0.0003,
+      "step": 19991
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4320506575742814e-08,
+      "loss": 0.0001,
+      "step": 19992
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4296151972722844e-08,
+      "loss": 0.0003,
+      "step": 19993
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4271797369702873e-08,
+      "loss": 0.0004,
+      "step": 19994
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4247442766682903e-08,
+      "loss": 0.0002,
+      "step": 19995
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.4223088163662931e-08,
+      "loss": 0.0003,
+      "step": 19996
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.419873356064296e-08,
+      "loss": 0.0001,
+      "step": 19997
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 1.417437895762299e-08,
+      "loss": 0.0004,
+      "step": 19998
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4150024354603018e-08,
+      "loss": 0.0002,
+      "step": 19999
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4125669751583048e-08,
+      "loss": 0.0002,
+      "step": 20000
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4101315148563077e-08,
+      "loss": 0.0002,
+      "step": 20001
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4076960545543109e-08,
+      "loss": 0.0002,
+      "step": 20002
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4052605942523137e-08,
+      "loss": 0.0003,
+      "step": 20003
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4028251339503166e-08,
+      "loss": 0.0003,
+      "step": 20004
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.4003896736483196e-08,
+      "loss": 0.0001,
+      "step": 20005
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.3979542133463224e-08,
+      "loss": 0.0002,
+      "step": 20006
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.3955187530443254e-08,
+      "loss": 0.0002,
+      "step": 20007
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 1.3930832927423283e-08,
+      "loss": 0.0003,
+      "step": 20008
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3906478324403311e-08,
+      "loss": 0.0002,
+      "step": 20009
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3882123721383341e-08,
+      "loss": 0.0002,
+      "step": 20010
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.385776911836337e-08,
+      "loss": 0.0003,
+      "step": 20011
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.38334145153434e-08,
+      "loss": 0.0003,
+      "step": 20012
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3809059912323428e-08,
+      "loss": 0.0003,
+      "step": 20013
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3784705309303459e-08,
+      "loss": 0.0002,
+      "step": 20014
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3760350706283487e-08,
+      "loss": 0.0002,
+      "step": 20015
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3735996103263516e-08,
+      "loss": 0.0001,
+      "step": 20016
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3711641500243546e-08,
+      "loss": 0.0003,
+      "step": 20017
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3687286897223574e-08,
+      "loss": 0.0003,
+      "step": 20018
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 1.3662932294203604e-08,
+      "loss": 0.0002,
+      "step": 20019
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3638577691183633e-08,
+      "loss": 1.0495,
+      "step": 20020
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3614223088163661e-08,
+      "loss": 0.0001,
+      "step": 20021
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3589868485143691e-08,
+      "loss": 0.0001,
+      "step": 20022
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.356551388212372e-08,
+      "loss": 0.0007,
+      "step": 20023
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.354115927910375e-08,
+      "loss": 0.0002,
+      "step": 20024
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3516804676083778e-08,
+      "loss": 0.0003,
+      "step": 20025
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3492450073063809e-08,
+      "loss": 0.0002,
+      "step": 20026
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3468095470043837e-08,
+      "loss": 0.0001,
+      "step": 20027
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3443740867023865e-08,
+      "loss": 0.0003,
+      "step": 20028
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 1.3419386264003897e-08,
+      "loss": 0.0003,
+      "step": 20029
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3395031660983926e-08,
+      "loss": 0.0002,
+      "step": 20030
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3370677057963956e-08,
+      "loss": 0.0003,
+      "step": 20031
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3346322454943984e-08,
+      "loss": 0.0001,
+      "step": 20032
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3321967851924014e-08,
+      "loss": 0.0002,
+      "step": 20033
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3297613248904043e-08,
+      "loss": 0.0002,
+      "step": 20034
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3273258645884071e-08,
+      "loss": 0.0002,
+      "step": 20035
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3248904042864102e-08,
+      "loss": 0.0002,
+      "step": 20036
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.322454943984413e-08,
+      "loss": 0.0002,
+      "step": 20037
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.320019483682416e-08,
+      "loss": 0.0002,
+      "step": 20038
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 1.3175840233804189e-08,
+      "loss": 0.0003,
+      "step": 20039
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.3151485630784217e-08,
+      "loss": 0.0003,
+      "step": 20040
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.3127131027764247e-08,
+      "loss": 0.0003,
+      "step": 20041
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.3102776424744276e-08,
+      "loss": 0.0002,
+      "step": 20042
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.3078421821724306e-08,
+      "loss": 0.0002,
+      "step": 20043
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.3054067218704334e-08,
+      "loss": 0.0002,
+      "step": 20044
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.3029712615684364e-08,
+      "loss": 0.0003,
+      "step": 20045
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.3005358012664393e-08,
+      "loss": 0.0004,
+      "step": 20046
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.2981003409644421e-08,
+      "loss": 0.0002,
+      "step": 20047
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.2956648806624451e-08,
+      "loss": 0.0001,
+      "step": 20048
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.293229420360448e-08,
+      "loss": 0.0002,
+      "step": 20049
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 1.290793960058451e-08,
+      "loss": 0.0002,
+      "step": 20050
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2883584997564539e-08,
+      "loss": 0.0002,
+      "step": 20051
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2859230394544567e-08,
+      "loss": 0.0003,
+      "step": 20052
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2834875791524597e-08,
+      "loss": 0.0001,
+      "step": 20053
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2810521188504626e-08,
+      "loss": 0.0003,
+      "step": 20054
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2786166585484656e-08,
+      "loss": 0.0002,
+      "step": 20055
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2761811982464684e-08,
+      "loss": 0.0002,
+      "step": 20056
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2737457379444716e-08,
+      "loss": 0.0001,
+      "step": 20057
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2713102776424744e-08,
+      "loss": 0.0003,
+      "step": 20058
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2688748173404773e-08,
+      "loss": 0.0001,
+      "step": 20059
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 1.2664393570384803e-08,
+      "loss": 0.0005,
+      "step": 20060
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2640038967364832e-08,
+      "loss": 0.0001,
+      "step": 20061
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2615684364344862e-08,
+      "loss": 0.0004,
+      "step": 20062
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.259132976132489e-08,
+      "loss": 0.0002,
+      "step": 20063
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2566975158304919e-08,
+      "loss": 0.0002,
+      "step": 20064
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2542620555284949e-08,
+      "loss": 0.0004,
+      "step": 20065
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2518265952264977e-08,
+      "loss": 0.0003,
+      "step": 20066
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2493911349245007e-08,
+      "loss": 0.0003,
+      "step": 20067
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2469556746225036e-08,
+      "loss": 0.0002,
+      "step": 20068
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2445202143205066e-08,
+      "loss": 0.0003,
+      "step": 20069
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 1.2420847540185094e-08,
+      "loss": 0.0002,
+      "step": 20070
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2396492937165123e-08,
+      "loss": 0.0002,
+      "step": 20071
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2372138334145153e-08,
+      "loss": 0.0002,
+      "step": 20072
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2347783731125182e-08,
+      "loss": 0.0002,
+      "step": 20073
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2323429128105212e-08,
+      "loss": 0.0002,
+      "step": 20074
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.229907452508524e-08,
+      "loss": 0.0002,
+      "step": 20075
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2274719922065269e-08,
+      "loss": 0.0001,
+      "step": 20076
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2250365319045299e-08,
+      "loss": 0.0002,
+      "step": 20077
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2226010716025327e-08,
+      "loss": 0.0002,
+      "step": 20078
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2201656113005357e-08,
+      "loss": 0.0003,
+      "step": 20079
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 1.2177301509985386e-08,
+      "loss": 0.0003,
+      "step": 20080
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.2152946906965416e-08,
+      "loss": 0.0001,
+      "step": 20081
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.2128592303945444e-08,
+      "loss": 0.0003,
+      "step": 20082
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.2104237700925473e-08,
+      "loss": 0.0001,
+      "step": 20083
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.2079883097905505e-08,
+      "loss": 0.0002,
+      "step": 20084
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.2055528494885533e-08,
+      "loss": 0.0003,
+      "step": 20085
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.2031173891865563e-08,
+      "loss": 0.0002,
+      "step": 20086
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.2006819288845592e-08,
+      "loss": 0.0003,
+      "step": 20087
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.198246468582562e-08,
+      "loss": 0.0002,
+      "step": 20088
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.195811008280565e-08,
+      "loss": 0.0002,
+      "step": 20089
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.1933755479785679e-08,
+      "loss": 0.0002,
+      "step": 20090
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 1.1909400876765709e-08,
+      "loss": 0.0002,
+      "step": 20091
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1885046273745737e-08,
+      "loss": 0.0002,
+      "step": 20092
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1860691670725768e-08,
+      "loss": 0.0002,
+      "step": 20093
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1836337067705796e-08,
+      "loss": 0.0003,
+      "step": 20094
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1811982464685825e-08,
+      "loss": 0.0002,
+      "step": 20095
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1787627861665855e-08,
+      "loss": 0.0002,
+      "step": 20096
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1763273258645883e-08,
+      "loss": 0.0002,
+      "step": 20097
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1738918655625913e-08,
+      "loss": 0.0003,
+      "step": 20098
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1714564052605942e-08,
+      "loss": 0.0001,
+      "step": 20099
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.169020944958597e-08,
+      "loss": 0.0002,
+      "step": 20100
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 1.1665854846566e-08,
+      "loss": 0.5514,
+      "step": 20101
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1641500243546029e-08,
+      "loss": 0.0001,
+      "step": 20102
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1617145640526059e-08,
+      "loss": 0.0002,
+      "step": 20103
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1592791037506087e-08,
+      "loss": 0.0003,
+      "step": 20104
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1568436434486118e-08,
+      "loss": 0.0003,
+      "step": 20105
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1544081831466146e-08,
+      "loss": 0.0003,
+      "step": 20106
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1519727228446174e-08,
+      "loss": 0.0001,
+      "step": 20107
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1495372625426205e-08,
+      "loss": 0.0002,
+      "step": 20108
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1471018022406233e-08,
+      "loss": 0.0002,
+      "step": 20109
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1446663419386265e-08,
+      "loss": 0.0001,
+      "step": 20110
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 1.1422308816366293e-08,
+      "loss": 0.0001,
+      "step": 20111
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1397954213346323e-08,
+      "loss": 0.0001,
+      "step": 20112
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1373599610326352e-08,
+      "loss": 0.0001,
+      "step": 20113
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.134924500730638e-08,
+      "loss": 0.0001,
+      "step": 20114
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.132489040428641e-08,
+      "loss": 0.0003,
+      "step": 20115
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1300535801266439e-08,
+      "loss": 0.0002,
+      "step": 20116
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1276181198246469e-08,
+      "loss": 0.0001,
+      "step": 20117
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1251826595226498e-08,
+      "loss": 0.0003,
+      "step": 20118
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1227471992206526e-08,
+      "loss": 0.0002,
+      "step": 20119
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1203117389186556e-08,
+      "loss": 0.0004,
+      "step": 20120
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1178762786166585e-08,
+      "loss": 0.0002,
+      "step": 20121
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 1.1154408183146615e-08,
+      "loss": 0.0002,
+      "step": 20122
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.1130053580126643e-08,
+      "loss": 0.0002,
+      "step": 20123
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.1105698977106673e-08,
+      "loss": 0.0003,
+      "step": 20124
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.1081344374086702e-08,
+      "loss": 0.0004,
+      "step": 20125
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.105698977106673e-08,
+      "loss": 0.0002,
+      "step": 20126
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.103263516804676e-08,
+      "loss": 0.0001,
+      "step": 20127
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.1008280565026789e-08,
+      "loss": 0.0006,
+      "step": 20128
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.0983925962006819e-08,
+      "loss": 0.0001,
+      "step": 20129
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.0959571358986848e-08,
+      "loss": 0.0002,
+      "step": 20130
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.0935216755966876e-08,
+      "loss": 0.0002,
+      "step": 20131
+    },
+    {
+      "epoch": 19.56,
+      "learning_rate": 1.0910862152946906e-08,
+      "loss": 0.0003,
+      "step": 20132
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.0886507549926935e-08,
+      "loss": 0.0004,
+      "step": 20133
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.0862152946906965e-08,
+      "loss": 0.0004,
+      "step": 20134
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.0837798343886993e-08,
+      "loss": 0.0003,
+      "step": 20135
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.0813443740867023e-08,
+      "loss": 0.0001,
+      "step": 20136
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.0789089137847053e-08,
+      "loss": 0.0002,
+      "step": 20137
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.0764734534827082e-08,
+      "loss": 0.0001,
+      "step": 20138
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.0740379931807112e-08,
+      "loss": 0.0002,
+      "step": 20139
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.071602532878714e-08,
+      "loss": 0.0002,
+      "step": 20140
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.069167072576717e-08,
+      "loss": 0.0003,
+      "step": 20141
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 1.06673161227472e-08,
+      "loss": 0.0002,
+      "step": 20142
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0642961519727228e-08,
+      "loss": 0.0002,
+      "step": 20143
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0618606916707258e-08,
+      "loss": 0.0003,
+      "step": 20144
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0594252313687286e-08,
+      "loss": 0.0002,
+      "step": 20145
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0569897710667316e-08,
+      "loss": 0.0002,
+      "step": 20146
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0545543107647345e-08,
+      "loss": 0.0002,
+      "step": 20147
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0521188504627375e-08,
+      "loss": 0.0003,
+      "step": 20148
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0496833901607403e-08,
+      "loss": 0.0002,
+      "step": 20149
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0472479298587432e-08,
+      "loss": 0.0004,
+      "step": 20150
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.0448124695567462e-08,
+      "loss": 0.0002,
+      "step": 20151
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 1.042377009254749e-08,
+      "loss": 0.0003,
+      "step": 20152
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.039941548952752e-08,
+      "loss": 0.0002,
+      "step": 20153
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0375060886507549e-08,
+      "loss": 0.0002,
+      "step": 20154
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0350706283487578e-08,
+      "loss": 0.0002,
+      "step": 20155
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0326351680467608e-08,
+      "loss": 0.0002,
+      "step": 20156
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0301997077447636e-08,
+      "loss": 0.0002,
+      "step": 20157
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0277642474427666e-08,
+      "loss": 0.0002,
+      "step": 20158
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0253287871407695e-08,
+      "loss": 0.0003,
+      "step": 20159
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0228933268387725e-08,
+      "loss": 0.0002,
+      "step": 20160
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0204578665367753e-08,
+      "loss": 0.0003,
+      "step": 20161
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0180224062347782e-08,
+      "loss": 0.0003,
+      "step": 20162
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 1.0155869459327812e-08,
+      "loss": 0.0002,
+      "step": 20163
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.0131514856307842e-08,
+      "loss": 0.0003,
+      "step": 20164
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.0107160253287872e-08,
+      "loss": 0.0004,
+      "step": 20165
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.00828056502679e-08,
+      "loss": 0.0002,
+      "step": 20166
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.005845104724793e-08,
+      "loss": 0.0005,
+      "step": 20167
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.003409644422796e-08,
+      "loss": 0.0002,
+      "step": 20168
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 1.0009741841207988e-08,
+      "loss": 0.0002,
+      "step": 20169
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 9.985387238188018e-09,
+      "loss": 0.0002,
+      "step": 20170
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 9.961032635168046e-09,
+      "loss": 0.0002,
+      "step": 20171
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 9.936678032148077e-09,
+      "loss": 0.0004,
+      "step": 20172
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 9.912323429128105e-09,
+      "loss": 0.0001,
+      "step": 20173
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.887968826108134e-09,
+      "loss": 0.0001,
+      "step": 20174
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.863614223088164e-09,
+      "loss": 0.0003,
+      "step": 20175
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.839259620068192e-09,
+      "loss": 0.0002,
+      "step": 20176
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.814905017048222e-09,
+      "loss": 0.0003,
+      "step": 20177
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.79055041402825e-09,
+      "loss": 0.0002,
+      "step": 20178
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.76619581100828e-09,
+      "loss": 0.0002,
+      "step": 20179
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.74184120798831e-09,
+      "loss": 0.0004,
+      "step": 20180
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.717486604968338e-09,
+      "loss": 0.0001,
+      "step": 20181
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.693132001948368e-09,
+      "loss": 0.0004,
+      "step": 20182
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 9.668777398928396e-09,
+      "loss": 0.0001,
+      "step": 20183
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.644422795908427e-09,
+      "loss": 0.0004,
+      "step": 20184
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.620068192888455e-09,
+      "loss": 0.0002,
+      "step": 20185
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.595713589868483e-09,
+      "loss": 0.0003,
+      "step": 20186
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.571358986848514e-09,
+      "loss": 0.0004,
+      "step": 20187
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.547004383828542e-09,
+      "loss": 0.0005,
+      "step": 20188
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.522649780808572e-09,
+      "loss": 0.0003,
+      "step": 20189
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.4982951777886e-09,
+      "loss": 0.0003,
+      "step": 20190
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.473940574768632e-09,
+      "loss": 0.0002,
+      "step": 20191
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.449585971748661e-09,
+      "loss": 0.0002,
+      "step": 20192
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.42523136872869e-09,
+      "loss": 0.0002,
+      "step": 20193
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 9.40087676570872e-09,
+      "loss": 0.0002,
+      "step": 20194
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.376522162688748e-09,
+      "loss": 0.0002,
+      "step": 20195
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.352167559668778e-09,
+      "loss": 0.0003,
+      "step": 20196
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.327812956648807e-09,
+      "loss": 0.0003,
+      "step": 20197
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.303458353628835e-09,
+      "loss": 0.0002,
+      "step": 20198
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.279103750608865e-09,
+      "loss": 0.0005,
+      "step": 20199
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.254749147588894e-09,
+      "loss": 0.0002,
+      "step": 20200
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.230394544568924e-09,
+      "loss": 0.0001,
+      "step": 20201
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.206039941548952e-09,
+      "loss": 0.0003,
+      "step": 20202
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.181685338528982e-09,
+      "loss": 0.0002,
+      "step": 20203
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.157330735509011e-09,
+      "loss": 0.0002,
+      "step": 20204
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 9.13297613248904e-09,
+      "loss": 0.0003,
+      "step": 20205
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 9.10862152946907e-09,
+      "loss": 0.0002,
+      "step": 20206
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 9.084266926449098e-09,
+      "loss": 0.0003,
+      "step": 20207
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 9.059912323429128e-09,
+      "loss": 0.0001,
+      "step": 20208
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 9.035557720409157e-09,
+      "loss": 0.0001,
+      "step": 20209
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 9.011203117389185e-09,
+      "loss": 0.0002,
+      "step": 20210
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 8.986848514369215e-09,
+      "loss": 0.0002,
+      "step": 20211
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 8.962493911349244e-09,
+      "loss": 0.0001,
+      "step": 20212
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 8.938139308329274e-09,
+      "loss": 0.0002,
+      "step": 20213
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 8.913784705309302e-09,
+      "loss": 0.0001,
+      "step": 20214
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.889430102289332e-09,
+      "loss": 0.0002,
+      "step": 20215
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.865075499269361e-09,
+      "loss": 0.0002,
+      "step": 20216
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.84072089624939e-09,
+      "loss": 0.0001,
+      "step": 20217
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.816366293229421e-09,
+      "loss": 0.0002,
+      "step": 20218
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.79201169020945e-09,
+      "loss": 0.0004,
+      "step": 20219
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.76765708718948e-09,
+      "loss": 0.0003,
+      "step": 20220
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.743302484169508e-09,
+      "loss": 0.0002,
+      "step": 20221
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.718947881149537e-09,
+      "loss": 0.0003,
+      "step": 20222
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.694593278129567e-09,
+      "loss": 0.0002,
+      "step": 20223
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 8.670238675109595e-09,
+      "loss": 0.0003,
+      "step": 20224
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.645884072089625e-09,
+      "loss": 0.0002,
+      "step": 20225
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.621529469069654e-09,
+      "loss": 0.0002,
+      "step": 20226
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.597174866049684e-09,
+      "loss": 0.0004,
+      "step": 20227
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.572820263029712e-09,
+      "loss": 0.0002,
+      "step": 20228
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.548465660009741e-09,
+      "loss": 0.0001,
+      "step": 20229
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.524111056989771e-09,
+      "loss": 0.0002,
+      "step": 20230
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.4997564539698e-09,
+      "loss": 0.0002,
+      "step": 20231
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.47540185094983e-09,
+      "loss": 0.0002,
+      "step": 20232
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.451047247929858e-09,
+      "loss": 0.0002,
+      "step": 20233
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.426692644909887e-09,
+      "loss": 0.0003,
+      "step": 20234
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 8.402338041889917e-09,
+      "loss": 0.0002,
+      "step": 20235
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.377983438869945e-09,
+      "loss": 0.0002,
+      "step": 20236
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.353628835849975e-09,
+      "loss": 0.0003,
+      "step": 20237
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.329274232830004e-09,
+      "loss": 0.0003,
+      "step": 20238
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.304919629810034e-09,
+      "loss": 0.0001,
+      "step": 20239
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.280565026790062e-09,
+      "loss": 0.0002,
+      "step": 20240
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.256210423770091e-09,
+      "loss": 0.0001,
+      "step": 20241
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.231855820750121e-09,
+      "loss": 0.0001,
+      "step": 20242
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.20750121773015e-09,
+      "loss": 0.0002,
+      "step": 20243
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.183146614710181e-09,
+      "loss": 0.0003,
+      "step": 20244
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 8.15879201169021e-09,
+      "loss": 0.0002,
+      "step": 20245
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 8.134437408670238e-09,
+      "loss": 0.0002,
+      "step": 20246
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 8.110082805650268e-09,
+      "loss": 0.0001,
+      "step": 20247
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 8.085728202630297e-09,
+      "loss": 0.0005,
+      "step": 20248
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 8.061373599610327e-09,
+      "loss": 0.0001,
+      "step": 20249
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 8.037018996590355e-09,
+      "loss": 0.0002,
+      "step": 20250
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 8.012664393570386e-09,
+      "loss": 0.0004,
+      "step": 20251
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 7.988309790550414e-09,
+      "loss": 0.0004,
+      "step": 20252
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 7.963955187530443e-09,
+      "loss": 0.0002,
+      "step": 20253
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 7.939600584510473e-09,
+      "loss": 0.0004,
+      "step": 20254
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 7.915245981490501e-09,
+      "loss": 0.0003,
+      "step": 20255
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.890891378470531e-09,
+      "loss": 0.0003,
+      "step": 20256
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.86653677545056e-09,
+      "loss": 0.0002,
+      "step": 20257
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.842182172430588e-09,
+      "loss": 0.0001,
+      "step": 20258
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.817827569410618e-09,
+      "loss": 0.0002,
+      "step": 20259
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.793472966390648e-09,
+      "loss": 0.0002,
+      "step": 20260
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.769118363370677e-09,
+      "loss": 0.0004,
+      "step": 20261
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.744763760350705e-09,
+      "loss": 0.0002,
+      "step": 20262
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.720409157330736e-09,
+      "loss": 0.0003,
+      "step": 20263
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.696054554310764e-09,
+      "loss": 0.0002,
+      "step": 20264
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.671699951290794e-09,
+      "loss": 0.0002,
+      "step": 20265
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 7.647345348270823e-09,
+      "loss": 0.0002,
+      "step": 20266
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.622990745250853e-09,
+      "loss": 0.0002,
+      "step": 20267
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.598636142230881e-09,
+      "loss": 0.0002,
+      "step": 20268
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.57428153921091e-09,
+      "loss": 0.0004,
+      "step": 20269
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.54992693619094e-09,
+      "loss": 0.0002,
+      "step": 20270
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.52557233317097e-09,
+      "loss": 0.0002,
+      "step": 20271
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.501217730150998e-09,
+      "loss": 0.0002,
+      "step": 20272
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.476863127131029e-09,
+      "loss": 0.0004,
+      "step": 20273
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.452508524111057e-09,
+      "loss": 0.0003,
+      "step": 20274
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.4281539210910855e-09,
+      "loss": 0.0003,
+      "step": 20275
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 7.403799318071115e-09,
+      "loss": 0.0003,
+      "step": 20276
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.379444715051144e-09,
+      "loss": 0.0002,
+      "step": 20277
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.355090112031174e-09,
+      "loss": 0.0004,
+      "step": 20278
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.3307355090112035e-09,
+      "loss": 0.0001,
+      "step": 20279
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.306380905991232e-09,
+      "loss": 0.0002,
+      "step": 20280
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.282026302971261e-09,
+      "loss": 0.0002,
+      "step": 20281
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.257671699951291e-09,
+      "loss": 0.0001,
+      "step": 20282
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.23331709693132e-09,
+      "loss": 0.0002,
+      "step": 20283
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.208962493911349e-09,
+      "loss": 0.0002,
+      "step": 20284
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.1846078908913785e-09,
+      "loss": 0.0002,
+      "step": 20285
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 7.160253287871407e-09,
+      "loss": 0.0002,
+      "step": 20286
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 7.135898684851436e-09,
+      "loss": 0.0008,
+      "step": 20287
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 7.1115440818314656e-09,
+      "loss": 0.0002,
+      "step": 20288
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 7.087189478811495e-09,
+      "loss": 0.0002,
+      "step": 20289
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 7.062834875791524e-09,
+      "loss": 0.0004,
+      "step": 20290
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 7.038480272771554e-09,
+      "loss": 0.0002,
+      "step": 20291
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 7.014125669751583e-09,
+      "loss": 0.0001,
+      "step": 20292
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 6.989771066731612e-09,
+      "loss": 0.0003,
+      "step": 20293
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 6.965416463711641e-09,
+      "loss": 0.0003,
+      "step": 20294
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 6.941061860691671e-09,
+      "loss": 0.0003,
+      "step": 20295
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 6.9167072576717e-09,
+      "loss": 0.0002,
+      "step": 20296
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 6.892352654651729e-09,
+      "loss": 0.0005,
+      "step": 20297
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.867998051631758e-09,
+      "loss": 0.0002,
+      "step": 20298
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.843643448611787e-09,
+      "loss": 0.0002,
+      "step": 20299
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.819288845591816e-09,
+      "loss": 0.0002,
+      "step": 20300
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.794934242571846e-09,
+      "loss": 0.0002,
+      "step": 20301
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.770579639551875e-09,
+      "loss": 0.0002,
+      "step": 20302
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.746225036531904e-09,
+      "loss": 0.0003,
+      "step": 20303
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.721870433511933e-09,
+      "loss": 0.0004,
+      "step": 20304
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.697515830491963e-09,
+      "loss": 0.0003,
+      "step": 20305
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.673161227471992e-09,
+      "loss": 0.0004,
+      "step": 20306
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 6.6488066244520215e-09,
+      "loss": 0.0001,
+      "step": 20307
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.624452021432051e-09,
+      "loss": 0.0002,
+      "step": 20308
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.60009741841208e-09,
+      "loss": 0.0002,
+      "step": 20309
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.5757428153921085e-09,
+      "loss": 0.0003,
+      "step": 20310
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.551388212372138e-09,
+      "loss": 0.0001,
+      "step": 20311
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.527033609352167e-09,
+      "loss": 0.0002,
+      "step": 20312
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.5026790063321964e-09,
+      "loss": 0.0004,
+      "step": 20313
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.478324403312226e-09,
+      "loss": 0.0004,
+      "step": 20314
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.453969800292255e-09,
+      "loss": 0.0002,
+      "step": 20315
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.4296151972722835e-09,
+      "loss": 0.0003,
+      "step": 20316
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 6.405260594252313e-09,
+      "loss": 2.3944,
+      "step": 20317
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.380905991232342e-09,
+      "loss": 0.0002,
+      "step": 20318
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.356551388212372e-09,
+      "loss": 0.0003,
+      "step": 20319
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.3321967851924015e-09,
+      "loss": 0.0003,
+      "step": 20320
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.307842182172431e-09,
+      "loss": 0.0002,
+      "step": 20321
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.283487579152459e-09,
+      "loss": 0.0001,
+      "step": 20322
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.259132976132489e-09,
+      "loss": 0.0001,
+      "step": 20323
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.234778373112518e-09,
+      "loss": 0.0002,
+      "step": 20324
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.210423770092547e-09,
+      "loss": 0.0001,
+      "step": 20325
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.1860691670725765e-09,
+      "loss": 0.0003,
+      "step": 20326
+    },
+    {
+      "epoch": 19.75,
+      "learning_rate": 6.161714564052606e-09,
+      "loss": 0.0003,
+      "step": 20327
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 6.137359961032634e-09,
+      "loss": 0.0002,
+      "step": 20328
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 6.113005358012664e-09,
+      "loss": 0.0003,
+      "step": 20329
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 6.088650754992693e-09,
+      "loss": 0.0004,
+      "step": 20330
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 6.064296151972722e-09,
+      "loss": 0.0002,
+      "step": 20331
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 6.039941548952752e-09,
+      "loss": 0.0001,
+      "step": 20332
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 6.015586945932782e-09,
+      "loss": 0.0001,
+      "step": 20333
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 5.99123234291281e-09,
+      "loss": 0.0003,
+      "step": 20334
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 5.966877739892839e-09,
+      "loss": 0.0001,
+      "step": 20335
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 5.942523136872869e-09,
+      "loss": 0.0002,
+      "step": 20336
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 5.918168533852898e-09,
+      "loss": 0.0002,
+      "step": 20337
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 5.893813930832927e-09,
+      "loss": 0.0004,
+      "step": 20338
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.869459327812957e-09,
+      "loss": 0.0001,
+      "step": 20339
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.845104724792985e-09,
+      "loss": 0.3662,
+      "step": 20340
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.820750121773014e-09,
+      "loss": 0.0002,
+      "step": 20341
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.796395518753044e-09,
+      "loss": 0.0003,
+      "step": 20342
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.772040915733073e-09,
+      "loss": 0.0002,
+      "step": 20343
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.747686312713102e-09,
+      "loss": 1.3593,
+      "step": 20344
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.7233317096931324e-09,
+      "loss": 0.0004,
+      "step": 20345
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.698977106673162e-09,
+      "loss": 0.0005,
+      "step": 20346
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.67462250365319e-09,
+      "loss": 0.0002,
+      "step": 20347
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 5.6502679006332195e-09,
+      "loss": 0.0002,
+      "step": 20348
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.625913297613249e-09,
+      "loss": 0.0002,
+      "step": 20349
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.601558694593278e-09,
+      "loss": 0.0003,
+      "step": 20350
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.577204091573307e-09,
+      "loss": 0.0002,
+      "step": 20351
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.552849488553337e-09,
+      "loss": 0.0001,
+      "step": 20352
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.528494885533365e-09,
+      "loss": 0.0003,
+      "step": 20353
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.5041402825133945e-09,
+      "loss": 0.0002,
+      "step": 20354
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.479785679493424e-09,
+      "loss": 0.0001,
+      "step": 20355
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.455431076473453e-09,
+      "loss": 0.0002,
+      "step": 20356
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.431076473453482e-09,
+      "loss": 0.0004,
+      "step": 20357
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 5.406721870433512e-09,
+      "loss": 0.0002,
+      "step": 20358
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.382367267413541e-09,
+      "loss": 0.0004,
+      "step": 20359
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.35801266439357e-09,
+      "loss": 0.0002,
+      "step": 20360
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.3336580613736e-09,
+      "loss": 0.0003,
+      "step": 20361
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.309303458353629e-09,
+      "loss": 0.0002,
+      "step": 20362
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.284948855333658e-09,
+      "loss": 0.0003,
+      "step": 20363
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.2605942523136875e-09,
+      "loss": 0.0004,
+      "step": 20364
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.236239649293716e-09,
+      "loss": 0.0239,
+      "step": 20365
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.211885046273745e-09,
+      "loss": 0.0001,
+      "step": 20366
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.1875304432537746e-09,
+      "loss": 0.0002,
+      "step": 20367
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.163175840233804e-09,
+      "loss": 0.0002,
+      "step": 20368
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 5.138821237213833e-09,
+      "loss": 0.0002,
+      "step": 20369
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 5.1144666341938625e-09,
+      "loss": 0.0004,
+      "step": 20370
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 5.090112031173891e-09,
+      "loss": 0.0003,
+      "step": 20371
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 5.065757428153921e-09,
+      "loss": 0.0001,
+      "step": 20372
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 5.04140282513395e-09,
+      "loss": 0.0004,
+      "step": 20373
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 5.01704822211398e-09,
+      "loss": 0.0004,
+      "step": 20374
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 4.992693619094009e-09,
+      "loss": 0.0002,
+      "step": 20375
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 4.968339016074038e-09,
+      "loss": 0.0001,
+      "step": 20376
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 4.943984413054067e-09,
+      "loss": 0.0003,
+      "step": 20377
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 4.919629810034096e-09,
+      "loss": 0.0001,
+      "step": 20378
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 4.895275207014125e-09,
+      "loss": 0.0001,
+      "step": 20379
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.870920603994155e-09,
+      "loss": 0.0003,
+      "step": 20380
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.846566000974184e-09,
+      "loss": 0.0002,
+      "step": 20381
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.822211397954213e-09,
+      "loss": 0.0003,
+      "step": 20382
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.797856794934242e-09,
+      "loss": 0.0002,
+      "step": 20383
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.773502191914271e-09,
+      "loss": 0.0002,
+      "step": 20384
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.7491475888943e-09,
+      "loss": 0.0001,
+      "step": 20385
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.7247929858743305e-09,
+      "loss": 0.0002,
+      "step": 20386
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.70043838285436e-09,
+      "loss": 0.0004,
+      "step": 20387
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.676083779834389e-09,
+      "loss": 0.0005,
+      "step": 20388
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 4.6517291768144175e-09,
+      "loss": 0.0002,
+      "step": 20389
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.627374573794447e-09,
+      "loss": 0.0002,
+      "step": 20390
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.603019970774476e-09,
+      "loss": 0.0003,
+      "step": 20391
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.5786653677545054e-09,
+      "loss": 0.0001,
+      "step": 20392
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.554310764734535e-09,
+      "loss": 0.0003,
+      "step": 20393
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.529956161714564e-09,
+      "loss": 0.0004,
+      "step": 20394
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.5056015586945925e-09,
+      "loss": 0.0001,
+      "step": 20395
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.481246955674622e-09,
+      "loss": 0.0003,
+      "step": 20396
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.456892352654651e-09,
+      "loss": 0.0003,
+      "step": 20397
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.4325377496346804e-09,
+      "loss": 0.0002,
+      "step": 20398
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 4.4081831466147106e-09,
+      "loss": 0.0004,
+      "step": 20399
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.38382854359474e-09,
+      "loss": 0.0004,
+      "step": 20400
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.359473940574768e-09,
+      "loss": 0.0003,
+      "step": 20401
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.335119337554798e-09,
+      "loss": 0.0001,
+      "step": 20402
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.310764734534827e-09,
+      "loss": 0.0001,
+      "step": 20403
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.286410131514856e-09,
+      "loss": 0.0001,
+      "step": 20404
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.2620555284948855e-09,
+      "loss": 0.0002,
+      "step": 20405
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.237700925474915e-09,
+      "loss": 0.0004,
+      "step": 20406
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.213346322454943e-09,
+      "loss": 0.0002,
+      "step": 20407
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.188991719434973e-09,
+      "loss": 0.0003,
+      "step": 20408
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.164637116415002e-09,
+      "loss": 0.0003,
+      "step": 20409
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 4.140282513395031e-09,
+      "loss": 0.0001,
+      "step": 20410
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.1159279103750605e-09,
+      "loss": 0.0003,
+      "step": 20411
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.091573307355091e-09,
+      "loss": 0.0002,
+      "step": 20412
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.067218704335119e-09,
+      "loss": 0.0002,
+      "step": 20413
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.0428641013151484e-09,
+      "loss": 0.0002,
+      "step": 20414
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 4.018509498295178e-09,
+      "loss": 0.0001,
+      "step": 20415
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 3.994154895275207e-09,
+      "loss": 0.0001,
+      "step": 20416
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 3.969800292255236e-09,
+      "loss": 0.0002,
+      "step": 20417
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 3.945445689235266e-09,
+      "loss": 0.0003,
+      "step": 20418
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 3.921091086215294e-09,
+      "loss": 0.0004,
+      "step": 20419
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 3.896736483195324e-09,
+      "loss": 0.0002,
+      "step": 20420
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.872381880175353e-09,
+      "loss": 0.0001,
+      "step": 20421
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.848027277155382e-09,
+      "loss": 0.0003,
+      "step": 20422
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.823672674135411e-09,
+      "loss": 0.0002,
+      "step": 20423
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.799318071115441e-09,
+      "loss": 0.0004,
+      "step": 20424
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.77496346809547e-09,
+      "loss": 0.0003,
+      "step": 20425
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.750608865075499e-09,
+      "loss": 0.0002,
+      "step": 20426
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.7262542620555285e-09,
+      "loss": 0.0002,
+      "step": 20427
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.7018996590355574e-09,
+      "loss": 0.0002,
+      "step": 20428
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.677545056015587e-09,
+      "loss": 0.0001,
+      "step": 20429
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 3.653190452995616e-09,
+      "loss": 0.0003,
+      "step": 20430
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.6288358499756453e-09,
+      "loss": 0.0001,
+      "step": 20431
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.6044812469556746e-09,
+      "loss": 0.0025,
+      "step": 20432
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.5801266439357035e-09,
+      "loss": 0.0003,
+      "step": 20433
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.5557720409157328e-09,
+      "loss": 0.0001,
+      "step": 20434
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.531417437895762e-09,
+      "loss": 0.0001,
+      "step": 20435
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.5070628348757914e-09,
+      "loss": 0.0004,
+      "step": 20436
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.4827082318558207e-09,
+      "loss": 0.0003,
+      "step": 20437
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.45835362883585e-09,
+      "loss": 0.0003,
+      "step": 20438
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.433999025815879e-09,
+      "loss": 0.0002,
+      "step": 20439
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.409644422795908e-09,
+      "loss": 0.0003,
+      "step": 20440
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 3.3852898197759375e-09,
+      "loss": 0.0002,
+      "step": 20441
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.3609352167559664e-09,
+      "loss": 0.0003,
+      "step": 20442
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.336580613735996e-09,
+      "loss": 0.0001,
+      "step": 20443
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.3122260107160254e-09,
+      "loss": 0.0002,
+      "step": 20444
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.2878714076960543e-09,
+      "loss": 0.0002,
+      "step": 20445
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.2635168046760836e-09,
+      "loss": 0.0002,
+      "step": 20446
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.239162201656113e-09,
+      "loss": 0.0002,
+      "step": 20447
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.2148075986361418e-09,
+      "loss": 0.0002,
+      "step": 20448
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.190452995616171e-09,
+      "loss": 0.0004,
+      "step": 20449
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.1660983925962008e-09,
+      "loss": 0.0002,
+      "step": 20450
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 3.1417437895762297e-09,
+      "loss": 0.0003,
+      "step": 20451
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 3.117389186556259e-09,
+      "loss": 0.0028,
+      "step": 20452
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 3.0930345835362883e-09,
+      "loss": 0.0002,
+      "step": 20453
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 3.068679980516317e-09,
+      "loss": 0.0004,
+      "step": 20454
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 3.0443253774963465e-09,
+      "loss": 0.0003,
+      "step": 20455
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 3.019970774476376e-09,
+      "loss": 0.0004,
+      "step": 20456
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.995616171456405e-09,
+      "loss": 0.0014,
+      "step": 20457
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.9712615684364344e-09,
+      "loss": 1.711,
+      "step": 20458
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.9469069654164637e-09,
+      "loss": 0.0005,
+      "step": 20459
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.9225523623964925e-09,
+      "loss": 0.0003,
+      "step": 20460
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 2.898197759376522e-09,
+      "loss": 0.0003,
+      "step": 20461
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.873843156356551e-09,
+      "loss": 0.0003,
+      "step": 20462
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.849488553336581e-09,
+      "loss": 0.0002,
+      "step": 20463
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.8251339503166098e-09,
+      "loss": 0.0005,
+      "step": 20464
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.800779347296639e-09,
+      "loss": 0.0003,
+      "step": 20465
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.7764247442766684e-09,
+      "loss": 0.0003,
+      "step": 20466
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.7520701412566972e-09,
+      "loss": 0.0002,
+      "step": 20467
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.7277155382367265e-09,
+      "loss": 0.0002,
+      "step": 20468
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.703360935216756e-09,
+      "loss": 0.0003,
+      "step": 20469
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.679006332196785e-09,
+      "loss": 0.0004,
+      "step": 20470
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 2.6546517291768144e-09,
+      "loss": 0.0002,
+      "step": 20471
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.6302971261568437e-09,
+      "loss": 0.0003,
+      "step": 20472
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.6059425231368726e-09,
+      "loss": 0.0002,
+      "step": 20473
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.581587920116902e-09,
+      "loss": 0.0004,
+      "step": 20474
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.5572333170969312e-09,
+      "loss": 0.0003,
+      "step": 20475
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.5328787140769605e-09,
+      "loss": 0.0001,
+      "step": 20476
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.50852411105699e-09,
+      "loss": 0.0001,
+      "step": 20477
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.484169508037019e-09,
+      "loss": 0.0002,
+      "step": 20478
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.459814905017048e-09,
+      "loss": 0.0003,
+      "step": 20479
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.4354603019970773e-09,
+      "loss": 0.0002,
+      "step": 20480
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.4111056989771066e-09,
+      "loss": 0.0002,
+      "step": 20481
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 2.3867510959571355e-09,
+      "loss": 0.0002,
+      "step": 20482
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.3623964929371652e-09,
+      "loss": 0.0003,
+      "step": 20483
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.3380418899171945e-09,
+      "loss": 0.0002,
+      "step": 20484
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.3136872868972234e-09,
+      "loss": 0.0004,
+      "step": 20485
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.2893326838772527e-09,
+      "loss": 0.0001,
+      "step": 20486
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.264978080857282e-09,
+      "loss": 0.0002,
+      "step": 20487
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.240623477837311e-09,
+      "loss": 0.0002,
+      "step": 20488
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.2162688748173402e-09,
+      "loss": 0.0002,
+      "step": 20489
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.19191427179737e-09,
+      "loss": 0.0004,
+      "step": 20490
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.167559668777399e-09,
+      "loss": 0.0002,
+      "step": 20491
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 2.143205065757428e-09,
+      "loss": 0.0002,
+      "step": 20492
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 2.1188504627374574e-09,
+      "loss": 0.0002,
+      "step": 20493
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 2.0944958597174863e-09,
+      "loss": 0.0002,
+      "step": 20494
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 2.0701412566975156e-09,
+      "loss": 0.0002,
+      "step": 20495
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 2.0457866536775453e-09,
+      "loss": 0.0004,
+      "step": 20496
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 2.0214320506575742e-09,
+      "loss": 0.0002,
+      "step": 20497
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.9970774476376035e-09,
+      "loss": 0.0003,
+      "step": 20498
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.972722844617633e-09,
+      "loss": 0.0001,
+      "step": 20499
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.948368241597662e-09,
+      "loss": 0.0002,
+      "step": 20500
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.924013638577691e-09,
+      "loss": 0.0001,
+      "step": 20501
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.8996590355577203e-09,
+      "loss": 0.0003,
+      "step": 20502
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.8753044325377496e-09,
+      "loss": 0.0002,
+      "step": 20503
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.8509498295177787e-09,
+      "loss": 0.0001,
+      "step": 20504
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.826595226497808e-09,
+      "loss": 0.0002,
+      "step": 20505
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.8022406234778373e-09,
+      "loss": 0.0004,
+      "step": 20506
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.7778860204578664e-09,
+      "loss": 0.0003,
+      "step": 20507
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.7535314174378957e-09,
+      "loss": 0.0001,
+      "step": 20508
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.729176814417925e-09,
+      "loss": 0.0002,
+      "step": 20509
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.704822211397954e-09,
+      "loss": 0.0003,
+      "step": 20510
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.6804676083779832e-09,
+      "loss": 0.0002,
+      "step": 20511
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.6561130053580127e-09,
+      "loss": 0.0002,
+      "step": 20512
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 1.6317584023380418e-09,
+      "loss": 0.0003,
+      "step": 20513
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.6074037993180709e-09,
+      "loss": 0.0002,
+      "step": 20514
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.5830491962981004e-09,
+      "loss": 0.0002,
+      "step": 20515
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.5586945932781295e-09,
+      "loss": 0.0003,
+      "step": 20516
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.5343399902581586e-09,
+      "loss": 0.0002,
+      "step": 20517
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.509985387238188e-09,
+      "loss": 0.0002,
+      "step": 20518
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.4856307842182172e-09,
+      "loss": 0.0003,
+      "step": 20519
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.4612761811982463e-09,
+      "loss": 0.0002,
+      "step": 20520
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.4369215781782756e-09,
+      "loss": 0.0002,
+      "step": 20521
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.4125669751583049e-09,
+      "loss": 0.0004,
+      "step": 20522
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.3882123721383342e-09,
+      "loss": 0.0002,
+      "step": 20523
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.3638577691183633e-09,
+      "loss": 0.0001,
+      "step": 20524
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.3395031660983926e-09,
+      "loss": 0.0003,
+      "step": 20525
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.3151485630784219e-09,
+      "loss": 0.0002,
+      "step": 20526
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.290793960058451e-09,
+      "loss": 0.0002,
+      "step": 20527
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.2664393570384803e-09,
+      "loss": 0.0002,
+      "step": 20528
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.2420847540185096e-09,
+      "loss": 0.0003,
+      "step": 20529
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.2177301509985387e-09,
+      "loss": 0.0002,
+      "step": 20530
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.1933755479785678e-09,
+      "loss": 0.0002,
+      "step": 20531
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.1690209449585973e-09,
+      "loss": 0.0007,
+      "step": 20532
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 1.1446663419386264e-09,
+      "loss": 0.0002,
+      "step": 20533
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.1203117389186555e-09,
+      "loss": 0.0001,
+      "step": 20534
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.095957135898685e-09,
+      "loss": 0.0003,
+      "step": 20535
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.071602532878714e-09,
+      "loss": 0.0002,
+      "step": 20536
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.0472479298587432e-09,
+      "loss": 0.0002,
+      "step": 20537
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 1.0228933268387727e-09,
+      "loss": 0.0004,
+      "step": 20538
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 9.985387238188018e-10,
+      "loss": 0.0002,
+      "step": 20539
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 9.74184120798831e-10,
+      "loss": 0.0002,
+      "step": 20540
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 9.498295177788601e-10,
+      "loss": 0.0002,
+      "step": 20541
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 9.254749147588893e-10,
+      "loss": 0.0002,
+      "step": 20542
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 9.011203117389186e-10,
+      "loss": 0.0006,
+      "step": 20543
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 8.767657087189478e-10,
+      "loss": 0.0001,
+      "step": 20544
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 8.52411105698977e-10,
+      "loss": 0.0002,
+      "step": 20545
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 8.280565026790063e-10,
+      "loss": 0.0004,
+      "step": 20546
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 8.037018996590354e-10,
+      "loss": 0.0004,
+      "step": 20547
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 7.793472966390647e-10,
+      "loss": 0.0002,
+      "step": 20548
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 7.54992693619094e-10,
+      "loss": 0.0002,
+      "step": 20549
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 7.306380905991231e-10,
+      "loss": 0.0001,
+      "step": 20550
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 7.062834875791524e-10,
+      "loss": 0.0002,
+      "step": 20551
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.819288845591816e-10,
+      "loss": 0.0002,
+      "step": 20552
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.575742815392109e-10,
+      "loss": 0.0002,
+      "step": 20553
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 6.332196785192401e-10,
+      "loss": 0.0003,
+      "step": 20554
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 6.088650754992693e-10,
+      "loss": 0.0002,
+      "step": 20555
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 5.845104724792986e-10,
+      "loss": 0.0003,
+      "step": 20556
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 5.601558694593277e-10,
+      "loss": 0.0001,
+      "step": 20557
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 5.35801266439357e-10,
+      "loss": 0.0002,
+      "step": 20558
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 5.114466634193863e-10,
+      "loss": 0.0002,
+      "step": 20559
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.870920603994155e-10,
+      "loss": 0.0001,
+      "step": 20560
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.6273745737944467e-10,
+      "loss": 0.0002,
+      "step": 20561
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.383828543594739e-10,
+      "loss": 0.0002,
+      "step": 20562
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 4.1402825133950317e-10,
+      "loss": 0.0002,
+      "step": 20563
+    },
+    {
+      "epoch": 19.98,
+      "learning_rate": 3.8967364831953237e-10,
+      "loss": 0.0005,
+      "step": 20564
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.6531904529956157e-10,
+      "loss": 0.0004,
+      "step": 20565
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.409644422795908e-10,
+      "loss": 0.0001,
+      "step": 20566
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.1660983925962007e-10,
+      "loss": 0.0002,
+      "step": 20567
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.922552362396493e-10,
+      "loss": 0.0002,
+      "step": 20568
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.679006332196785e-10,
+      "loss": 0.0002,
+      "step": 20569
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.4354603019970776e-10,
+      "loss": 0.0002,
+      "step": 20570
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 2.1919142717973696e-10,
+      "loss": 0.0004,
+      "step": 20571
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.9483682415976619e-10,
+      "loss": 0.0003,
+      "step": 20572
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.704822211397954e-10,
+      "loss": 0.0004,
+      "step": 20573
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 1.4612761811982466e-10,
+      "loss": 0.0002,
+      "step": 20574
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.2177301509985388e-10,
+      "loss": 0.006,
+      "step": 20575
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 9.741841207988309e-11,
+      "loss": 0.0003,
+      "step": 20576
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 7.306380905991233e-11,
+      "loss": 0.0003,
+      "step": 20577
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 4.8709206039941546e-11,
+      "loss": 0.0001,
+      "step": 20578
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 2.4354603019970773e-11,
+      "loss": 0.0003,
+      "step": 20579
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.0,
+      "loss": 0.0001,
+      "step": 20580
+    },
+    {
+      "epoch": 20.0,
+      "eval_accuracy": 0.9990277102576568,
+      "eval_loss": 0.013907751068472862,
+      "eval_runtime": 2.821,
+      "eval_samples_per_second": 729.18,
+      "eval_steps_per_second": 11.698,
+      "step": 20580
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 20580,
+  "num_train_epochs": 20,
+  "save_steps": 500,
+  "total_flos": 2144145774438696.0,
+  "trial_name": null,
+  "trial_params": null
+}